LegisTech Library | Bússola Tech

Beyond National Experimentation: Inter-Parliamentary Diplomacy and the Dissemination of Artificial Intelligence in Parliaments

Luís Kimaid — Tue, 03 Mar 2026 19:47:45 GMT

Parliaments across jurisdictions are exploring the potential of artificial intelligence to enhance transparency, improve internal workflows, expand citizen accessibility, and strengthen oversight functions. Yet the implementation of artificial intelligence in parliaments does not unfold in isolation. It is embedded within pre-existing institutional architectures, asymmetries in digital maturity, and longstanding traditions of inter-parliamentary diplomacy. As a result, the diffusion of artificial intelligence in parliaments is not simply a technological trajectory, it is a cooperative one.

A high-level panel discussion held on 6 November 2025, convened by Bússola Tech and moderated by Luís Kimaid, brought together Eric Janse, the Clerk of the House of Commons of Canada, Grant Vergottini, CEO of Xcential Legislative Technologies, José Pedro Montero, President of the Association of Secretaries-General of Parliaments, Julius Kampamba, Lead of the Technical Team of the Society of Clerks-at-the-Table (Africa), and Miguel Landeros, Secretary-General of the Cámara de Diputadas y Diputados de Chile. The exchange focused on the instruments through which inter-parliamentary cooperation can advance the implementation and understanding of artificial intelligence in legislative institutions.

Building on that discussion, this article advances an analytical account of inter-parliamentary cooperation as an emerging infrastructure for artificial intelligence governance in parliaments. It first identifies the typologies of cooperation through which knowledge and practice circulate across jurisdictions. It then examines the institutional preconditions that determine whether such cooperation translates into meaningful implementation. Finally, it considers how the impact of cooperative initiatives may be assessed in the absence of fully developed evaluative frameworks. Through this analysis, the article situates artificial intelligence not as a stand-alone innovation but as a catalyst operating within a layered architecture of diplomatic engagement, administrative capacity, and institutional design.

Typologies of Inter-Parliamentary Cooperation on Artificial Intelligence

Inter-parliamentary cooperation on artificial intelligence emerges through distinct but interconnected modalities that operate at multilateral, bilateral, technical and professional levels. These modalities differ in structure, duration, and depth of engagement, yet collectively constitute the operative architecture through which knowledge on artificial intelligence circulates across legislative institutions.

One prominent modality is cooperation through formal multilateral associations and inter-parliamentary bodies. Organisations such as the Society of Clerks-at-the-Table, the Association of Secretaries-General of Parliaments, the Commonwealth Parliamentary Association, the Organization of American States, ParlAmericas, the Inter-Parliamentary Union and various bilateral parliamentary associations provide structured forums in which artificial intelligence is introduced as a subject of deliberation. Within these platforms, artificial intelligence has been incorporated into general debates, thematic sessions, professional development seminars, and annual meetings. In these settings, senior parliamentary administrators and officials present communications detailing how their respective parliaments are approaching digital transformation and artificial intelligence. The function of these forums is not technological development per se, but the exchange of institutional experience, the articulation of practical challenges, and the collective reflection on governance implications. Through repeated meetings over several cycles, artificial intelligence shifts from a peripheral innovation topic to a central administrative concern embedded within the routine agenda of parliamentary diplomacy.

A second typology is bilateral technical cooperation based on direct institutional engagement. This form of cooperation involves one parliament supporting another in the development of digitalisation processes, data structuring, or artificial intelligence applications, taking into account differences in technical capacity and institutional maturity. Such cooperation may include the transfer of technological solutions, adaptation of tools originally developed in one jurisdiction for use in another, and iterative processes whereby the receiving parliament modifies or improves upon the shared initiative. The exchange is not unidirectional; technology and experience circulate in multiple directions, particularly where development capacities differ. In this typology, cooperation is framed not merely as dissemination but as co-development, in which innovation in one parliament becomes a resource for others, and improvements made by the recipient may subsequently be reintegrated into the originating institution’s practice.

A third typology consists of professional development mechanisms and structured learning exchanges. Annual seminars, sub-regional workshops, and exchange visits are designed to facilitate both awareness and practical engagement. These mechanisms include the organisation of thematic workshops on artificial intelligence, the cascading of topics from regional to sub-regional levels, and the inclusion of both technical staff and, in some instances, members of parliament. The objective is not confined to conceptual sensitisation; it extends to enabling technical personnel to observe and engage directly with operational systems in more advanced institutions. Learning is framed as extending beyond successful practices to include challenges and implementation difficulties, thereby allowing other parliaments to navigate adoption processes with greater institutional awareness. Within this typology, reporting mechanisms provide a degree of continuity and follow-up, even where formalised metrics remain limited.

A fourth typology concerns the development and dissemination of standards and shared infrastructures. Cooperation in this domain is oriented towards the creation of common data standards and structured formats for legislative information. Rather than producing isolated, jurisdiction-specific solutions, the emphasis lies in establishing interoperable frameworks capable of supporting artificial intelligence applications. The development of standards in collaboration with academic institutions and supranational bodies reflects an understanding that artificial intelligence relies on structured, machine-readable information rather than paper-based or static document formats, such as Akoma Ntoso. The publication of specific frameworks, such as the Guidelines for AI in Parliaments developed under the leadership of Dr. Fotios Fitsilis, further illustrates how normative principles for ethical governance, transparency, and human oversight are being articulated at the inter-parliamentary level to support responsible implementation.

A fifth typology is the creation of shared repositories, case databases, and knowledge platforms. Databases collecting use cases, toolkits, and documented experiences transform isolated initiatives into collective parliamentary knowledge. Such repositories allow parliaments to identify peers experimenting with specific technologies and to initiate targeted exchanges. The inclusion of experiences that did not achieve intended outcomes is framed as equally valuable, permitting institutions to learn from implementation challenges. This form of cooperation converts episodic dialogue into cumulative institutional memory and facilitates sustained engagement beyond single conferences or events. Research initiatives dedicated to systematically documenting parliamentary artificial intelligence cases further reinforce this dynamic. For example, Bússola Tech’s repository of AI cases illustrates how dispersed experimentation can be consolidated into structured institutional knowledge rather than remaining anecdotal.

A sixth typology involves cooperative pilot projects and joint development initiatives. In this configuration, two or more parliaments formally commit to addressing a common artificial intelligence challenge through a shared team or coordinated experimentation. The logic governing this model is that parliaments are not in competitive relations, therefore, they may collaborate in testing tools, sharing code, and distributing technical expertise. Shared ownership of innovation reduces duplication and reinforces collective capacity. The objective is not simply to discuss artificial intelligence, but to embed cooperative experimentation within institutional workflows.

Across these typologies, cooperation is shaped by recognition of asymmetries in digital maturity. Consequently, cooperative modalities are adapted to varying levels of development, ranging from support for small-scale applications using existing large language models to more complex cloud-based implementations. The typology of cooperation thus incorporates differentiated pathways rather than uniform solutions.

Taken together, these modalities illustrate that inter-parliamentary cooperation around artificial intelligence operates simultaneously as diplomatic engagement, technical collaboration, professional socialisation, standards development, and shared experimentation. Each element addresses a distinct dimension of artificial intelligence adoption, thereby constituting a layered architecture of cooperation through which parliaments seek to navigate this transition.

Institutional Preconditions for Effective AI Cooperation

The cooperative modalities previously identified depend upon internal conditions within parliaments that determine whether artificial intelligence initiatives can be meaningfully exchanged, adapted, and sustained. Inter-parliamentary cooperation does not substitute for institutional readiness. Differences in digital maturity, organisational capacity, political prioritisation, and information architecture directly shape the scope and depth of collaborative engagement .

A primary condition concerns the state of digitisation and data availability. Legislative information is not uniformly digitised across jurisdictions and databases are not always complete or structured. In such environments, advanced artificial intelligence applications cannot simply be transferred from one parliament to another. Cooperation must instead be calibrated to the technical baseline of the receiving institution. Where resources, engineering capacity, and infrastructure are limited, smaller-scale tools using existing language models may constitute the only feasible entry point. By contrast, parliaments with stronger technical foundations may proceed directly to cloud-based implementations and more complex architectures. The feasibility of cooperation therefore rests upon a realistic appraisal of internal technological conditions.

Closely related to digitisation is the question of structured information. The transition from publishing documents in paper or PDF format to publishing structured data suitable for computational consumption represents a foundational shift. Artificial intelligence systems rely on consistency, standardisation, and interpretability. The development and adoption of shared standards, undertaken in collaboration with academic institutions, are central to enabling cross-jurisdictional interoperability. Without convergence in how legislative information is structured, the accuracy and reliability of artificial intelligence tools cannot be ensured across contexts. Cooperation in this domain thus requires not only technical exchange but agreement on underlying publication practices.

Institutional governance structures constitute a further precondition. The Association of Secretaries-General of Parliaments, for example, provides platforms in which administrators present communications on their respective approaches to digital transformation and artificial intelligence. These forums create continuity across meetings and embed artificial intelligence within routine administrative deliberation. Although formalised metrics may not always exist, these reporting structures function as mechanisms of follow-up and collective awareness. Where governance arrangements enable regular reporting and review, cooperative initiatives acquire durability, where they do not, engagement risks remaining episodic.

Political endorsement also emerges as decisive. Efforts to integrate artificial intelligence often require deliberate mobilisation of members of parliament and senior officials. In some instances, MPs have been included in exchange visits so that exposure to operational systems can generate internal support upon return. Institutional uptake therefore depends not solely on technical viability but on whether leadership recognises artificial intelligence as aligned with parliamentary objectives, including transparency, accessibility, and efficiency.

Human capital is an additional structural factor. Immersive exchanges, staff attachments, and joint teams presuppose the availability of personnel capable of engaging substantively with artificial intelligence systems questions. Practical learning involves observing not only technical deployment but also management decisions, procurement processes, regulatory considerations, and responses to implementation challenges. Such exchanges require parliaments to allocate time and expertise. Cooperation, in this sense, demands sustained investment in professional capacity rather than attendance at isolated events. Strategic documentation further conditions the effectiveness of cooperation. The development of staff guidance on artificial intelligence transforms dialogue into transferable instruments. Where such documentation is absent, cooperation lacks concrete reference points for adaptation.

Measuring Impact of AI Cooperation in Parliaments

Having identified the typologies of cooperation and the institutional conditions that enable inter-parliamentary cooperation initiatives for the promotion of artificial intelligence in parliaments, the question of impact becomes unavoidable. Its impact must be evaluated by the extent to which it produces durable institutional change within participating parliaments.

One dimension of impact concerns participation and continuity. The breadth of engagement across jurisdictions, political traditions, and administrative levels provides an initial indicator of relevance. Where cooperation is sustained across multiple meetings and bilateral follow-ups, artificial intelligence shifts from a peripheral topic to a recurring institutional concern. Continuity beyond a single conference cycle indicates that cooperation has moved from episodic discussion to structured engagement. The sustained recurrence of thematic events accompanied by published analytical outputs offers an additional qualitative indicator of consolidation. For example, the sequence of conferences, webinars, and research publications dedicated to artificial intelligence in parliaments organised by Bússola Tech exemplifies how iterative dialogue combined with structured documentation can deepen institutional learning over time.

A second dimension relates to practical localisation. Exchanges are not impactful merely because they transmit models, they are impactful when receiving institutions adapt tools to their own technological and political environments. The differentiation between jurisdictions with advanced infrastructure and those with limited digitisation underscores that impact cannot be uniform. In some cases, it will take the form of incremental applications using existing language models, while in others, it may involve cloud-based deployments or structured data reforms. The metric is not technological sophistication per se, but the alignment between cooperative input and local capacity.

A third dimension concerns reporting and follow-up. In some arrangements, representatives report on developments within their respective parliaments, providing a mechanism for tracking progress. While formal quantitative metrics remain limited, structured reporting to membership bodies functions as a qualitative monitoring instrument. Such mechanisms allow the cooperative network to observe whether artificial intelligence initiatives are expanding, stagnating, or encountering obstacles.

At the same time, limitations in measurement need to be recognised. There is no comprehensive parametric tool currently in place to assess the effectiveness of inter-parliamentary cooperation on disseminating artificial intelligence expertise for parliaments. Evaluation often relies on indicators such as continued interest, allocation of resources, and willingness to pursue follow-up projects. This absence of systematic metrics does not negate impact, it reveals an area requiring further institutional development.

Finally, impact must be considered in normative terms. Artificial intelligence is an instrument to enhance parliamentary openness and efficiency. Where cooperative initiatives contribute to improved access to legislative information, strengthened oversight capabilities, or more inclusive digital interfaces, they produce value beyond technical deployment. The ultimate measure of cooperation lies in whether parliamentary institutions are better equipped to exercise their work and fulfil their functions.

Conclusion

This article demonstrates that inter-parliamentary cooperation on artificial intelligence cannot be understood as a peripheral diplomatic activity or as an accessory to technological experimentation. It constitutes an emerging layer of institutional infrastructure through which parliaments collectively navigate a structural transformation in how legislative information is produced, organised, and interpreted.

Cooperation alone does not guarantee transformation. The analysis of institutional preconditions shows that artificial intelligence adoption remains contingent upon internal readiness. Digitisation, structured data, governance continuity, political endorsement, human capital, and strategic documentation form the substrate upon which cooperative exchange can produce tangible effects. Where these conditions are uneven, cooperation must assume a differentiated character, adapting its instruments to varying levels of digital maturity.

The question of impact further clarifies the structural character of cooperation. Impact is not reducible to the frequency of conferences or the symbolic inclusion of artificial intelligence on the agenda. It becomes visible when dialogue crystallises into internal sustained follow-up mechanisms. Yet the absence of systematic metrics reveals a paradox: while cooperation is increasingly central to artificial intelligence in parliaments, the evaluation of such cooperation initiatives remains under-institutionalised.

Artificial intelligence should not be treated as an autonomous driver of institutional change but as an instrument for enhancing openness, accessibility, and administrative effectiveness. Cooperation, in this sense is a collective effort to reduce asymmetries in capacity so that parliaments, irrespective of size or resources, may strengthen their functions. The non-competitive character of parliamentary institutions enables this model of shared advancement. In this respect, cooperation is the mechanism through which parliaments construct a shared pathway into the knowledge-intensive phase of legislative administration.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Before the Technology: Rethinking Parliamentary Modernisation

Luís Kimaid — Wed, 25 Feb 2026 15:36:05 GMT

Introduction

Parliamentary modernisation is too often reduced to a technological procurement problem. Institutions are encouraged to adopt digital platforms, artificial intelligence, automation tools, and new service interfaces, as though innovation were primarily a matter of acquisition. Yet modernisation, properly understood, is a question of institutional sequencing, organisational maturity, governance design, and strategic endurance. Technology is an accelerant; it is not the origin.

A recent Bússola Tech discussion held on February 11th 2026 brought together Ciaran Doyle of the Oireachtas of Ireland, Fotis Fitsilis of the Hellenic Parliament, Rida Qazi of the Senate of Pakistan, and Marty Bickford of Xcential Legislative Technologies, under the moderation of Luís Kimaid from Bússola Tech. The exchange revolved precisely around this issue: what must come first, and why? Its substance provides a structured lens through which to examine the deeper logic of parliamentary transformation.

This article advances a coherent analysis of parliamentary modernisation as a matter of institutional sequencing and strategic design. It examines the foundational conditions that must be consolidated before technological intervention becomes meaningful, identifies the categories of investment capable of generating durable structural leverage, explores the governance arrangements required to ensure continuity across political cycles, and assesses how present architectural choices will condition the responsible integration of artificial intelligence into legislative work in the years ahead.

The Institutional Preparation

The first investment in parliamentary modernisation is neither software nor hardware. It is institutional self-understanding.

Modernisation begins with a mental shift at executive level. Parliamentary leadership must internalise that transformation is not episodic but permanent. Once an institution embarks upon systematic change, the only constant becomes adaptation itself. This recognition alters governance behaviour, with an organisational culture conditioned to accept continual recalibration.

This cognitive preparation is followed by diagnostic clarity. An institution must understand how it currently operates, its workflows, its procedural dependencies, its formal and informal power structures, its data flows, and its public interfaces. Enterprise-wide mapping of business processes, applications, data architectures, and inter-unit relationships is not a technical luxury but a structural prerequisite. Without such mapping, technological intervention risks reinforcing fragmentation rather than resolving it.

Equally important is institutional identity. In some contexts, parliaments must first consolidate clarity regarding their constitutional role, sovereignty, and operational legitimacy before attempting technological reform. Where internal responsibilities are diffuse, workflows ambiguous, or authority poorly codified, digitalisation simply accelerates confusion. Standardisation of processes and clarification of responsibilities constitute foundational work that may appear invisible but are indispensable to transformation.

Thus, the pre-technical phase consists of three elements: executive mental shift, structural self-mapping, and institutional role clarity. Only upon these foundations can strategy become meaningful.

Strategy, Frameworks, and Organisational Architecture

Modernisation requires formalisation within a long-term strategic framework. However, strategy must not be confused with static planning. It establishes direction, implementation remains iterative.

Structured frameworks provide coherence and survivability. They define objectives, prioritisation mechanisms, governance layers, and evaluation criteria. Crucially, they create institutional memory beyond individual office-holders. In environments characterised by leadership rotation, frameworks function as stabilising artefacts. They embed logic into the organisation itself rather than into personalities.

Prioritisation mechanisms are particularly critical. Resources in Parliaments are inherently limited, and not all reforms can proceed simultaneously. Structured exercises that evaluate initiatives according to urgency and relevance enable leadership to visualise trade-offs. Such methodologies prevent reform from being captured by external agendas or technological fashion cycles.

Governance architecture must also be explicit. Implementation teams, decision authorities, and oversight mechanisms need clear delineation. Without governance symmetry between innovation and compliance, reform oscillates between paralysis and overreach. Modernisation therefore depends upon governance structures capable of sustaining both discipline and agility.

High-Leverage Early Investments

When institutions ask which investments unlock multiple downstream gains, the answer is not singular but layered. First, investment in people generates compounding returns. Transformation requires multidisciplinary capacity, such as project management, enterprise architecture, business analysis, solution design, and digital facilitation. Internal capability ensures strategic autonomy and prevents dependency upon vendors for vision-setting. Training and experimentation cultivate adaptive resilience within the institution.

Secondly, data architecture constitutes structural leverage. Digitisation of records, standardisation of metadata, interoperability of systems, and consolidation of institutional memory create the substrate upon which subsequent innovation rests. Without structured data, neither efficiency nor analytical sophistication is attainable. Data strategy therefore precedes advanced technological application.

Thirdly, identity management and authentication systems provide foundational integrity. Weakness in these areas creates cascading constraints later. Early investment in robust authentication, standards adherence, and interoperable API gateways prevents costly retrofitting and enhances system coherence.

Fourthly, delivery governance must be institutionalised. A programme management structure that maintains execution discipline ensures that strategy translates into outcomes. Without delivery architecture, modernisation remains rhetorical.

Finally, public-facing systems often generate visible legitimacy gains. Transparent access to legislative data strengthens institutional credibility. These areas may not be technically simple, but they create perceptible value that reinforces political support for continued reform.

High-leverage investment, therefore, is characterised not by technological novelty but by systemic multiplier effects.

Roadmaps and Political Turnover

Parliaments operate within inherently political environments. Leadership changes, shifting priorities, and electoral cycles impose volatility upon administrative agendas. Modernisation roadmaps must therefore be designed to survive discontinuity.

Three mechanisms emerge as stabilising factors.

First, moral and technical authority derived from deep system understanding enables reform leaders to defend continuity. When transformation proposals are anchored in documented workflows, strategic alignment, and benchmarking evidence, they become harder to dismiss as personal initiatives.

Secondly, alignment with national and supranational strategic directives embeds reform within broader policy ecosystems. When parliamentary modernisation resonates with wider digital or governance strategies, it acquires external reinforcement.

Thirdly, modular reform design enhances resilience. Incremental building blocks embedded in clear systems reduce vulnerability to wholesale reversal. When processes become institutionally normalised rather than politically branded, they outlast administrations.

Endurance, therefore, depends upon embedding reform into organisational logic rather than into leadership charisma.

Conclusion

Parliamentary modernisation is not an act of technological adoption but of institutional sequencing. It begins with mental recalibration and structural clarity. It advances through strategic frameworks and prioritisation discipline. It leverages early investments in people, data, authentication, and governance. It survives political turnover through institutional roots.

Technology amplifies whatever foundations exist. Where governance is coherent, data structured, and identity clear, it enhances capability. Where these are absent, it magnifies disorder. The decisive focus, therefore, should be the institutional foundations that must be strengthened before any tool can meaningfully function.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Highlights of the in-person Event about Artificial Intelligence and Parliamentary Accountability

Bússola Tech — Fri, 30 Jan 2026 16:02:53 GMT

Bússola Tech, in collaboration with the Organization of American States (OAS) and the Federal Senate of Brazil, was honoured to host the LegisTech: Artificial Intelligence in Parliaments for Enhanced Accountability. It took place on 27th and 28th of January 2026 at the Senate of Brazil, in Brasília, this high-level convening brought together parliamentary authorities, technical leaders, and specialists from across the world to explore how Artificial Intelligence is transforming the transparency and accountability functions of modern legislatures.

Building on the continuity of previous strategic dialogues, this edition focused on the concrete institutional, procedural, and technological conditions required for parliaments to integrate AI responsibly and effectively. From structuring legislative data for intelligent systems and enhancing digital public interfaces, to modernising workflow management and fortifying cybersecurity, the programme examined the evolving intersection between innovation, governance, and democratic legitimacy.

Throughout the two-day programme, participants shared their experiences, compared emerging practices, and analysed strategic pathways for strengthening the use of AI in support of more open, efficient, and citizen-centred parliaments. The conference highlighted a wide range of practical implementations, lessons learned, and institutional innovations adopted across different regions.

By convening a diverse set of perspectives and cases, the event demonstrated the transformative potential of AI in reinforcing parliamentary transparency, improving access to information, and enabling more dynamic forms of engagement with citizens. The discussions supported parliaments in becoming more resilient, data-driven, and future-ready, while ensuring that innovation is anchored in accountability and public trust.

The recordings and the publication about the topic of the event will be made available soon.

Photos of the event

Partners of the Event

Book about "Artificial Intelligence in Legislative Services: Principles for Effective Implementation"

Bússola Tech — Thu, 27 Nov 2025 20:12:58 GMT

Partners

This book, developed by Bússola Tech in partnership with the Organization of American States (OAS), the Commonwealth Parliamentary Association (CPA), and the Legislative Assembly of Rio Grande do Norte (Brazil), is the culmination of collaboration among parliamentary staff, academics, and AI experts. Their collective insights guide an exploration of how Artificial Intelligence (AI) can significantly enhance parliamentary services.

It provides a comprehensive look at AI applications in legislative drafting, procedural guidance, historical archiving, and citizen engagement, supported by real-world case studies illustrating practical implementations. The publication emphasizes critical considerations such as ethical deployment, data integrity, and transparent communication to ensure AI enhances parliamentary work without compromising core democratic values.

The book also addresses the complexities of integrating AI into parliamentary procedures, stressing the need for human oversight, cross-institutional cooperation, and robust staff training to maximize AI’s benefits. Through careful and ethical implementation, parliaments can leverage AI to strengthen their services, inclusivity, and public trust, fostering democratic governance in the digital age.

Coordinator: Luís Kimaid, Executive Director at Bússola Tech.

Authors:

André Freire - Câmara dos Deputados of Brazil
Brahma Mishra - MeetMonk
Ciarán Doyle - Oireachtas of Ireland
Fotios Fitsilis - Hellenic Parliament
German Tarasewiez - Cámara de Diputados de la Nación of Argentina
Grant Vergottini - Xcential Legislative Technologies
James Pinnell - Commonwealth Parliamentary Association (CPA)
João Lima - Senado Federal of Brazil
Juan de Dios Cincunegui - Universidad Austral
Luís Kimaid - Bússola Tech
Moises Benamor - Organization of American States (OAS)
Monica Palmirani - Università di Bologna
Márcio Tancredi and Adriano Torres Ribeiro de Castro - Senado Federal of Brazil
Prince Kenneh - Parliament of Sierra Leone
Rodolfo Vaz - Câmara dos Deputados of Brazil
Wade Ballou - U.S. House of Representatives

Partners:

From Citizen to Senator: Artificial Intelligence and the Reinvention of Citizen Lawmaking in Brazil

Beth Noveck — Fri, 15 Aug 2025 17:03:15 GMT

Co-authors:

Luís Kimaid, Executive-Director at Bússola Tech;
Alisson Bruno de Queiroz, e-citizenship coordinator at the Senado Federal of Brazil;
Beth Simone Noveck, Director at the Governance Lab.

When an 82-year-old retiree from the coastal city of Maceió called Brazil's Senate hotline to suggest that medication labels needed larger font sizes, he had no idea that a senator from Amazonas would champion his simple suggestion. Across the country, Alessandro from Minas Gerais proposed legislation to prevent internet service providers from throttling speeds for services like Netflix—five years later, his net neutrality idea became reality. These citizens are not anomalies but participants in Brazil's systematic approach to democratic engagement.

Imagine a democracy where ordinary people have influenced 46 legislative bills through more than 120,000 submitted ideas and 11 million votes.

From a proposal to end housing subsidies for deputies and judges and terminate perks for former presidents to the idea to legalize marijuana and ban straws, the public has had a significant impact on lawmaking.

While none of the resident proposals have become law as is (citizen proposals are only 140 characters), their input has helped to shape the legislative process. Unlike most nations where citizen participation remains limited to periodic voting or the occasional town hall, Brazil has built four robust, integrated participatory channels that invite citizens into every stage of governance. Through these channels, Brazilians can propose laws, question witnesses, participate in educational workshops to learn how to craft legislation, and vote on pending bills—creating an ecosystem of engagement unseen elsewhere.

This isn't a theoretical vision—it's Brazil's current reality. In most democracies, the public exercises political power primarily at the ballot box every few years before returning to the sidelines. Brazil's Federal Senate, however, has reimagined this relationship, transforming citizens from occasional voters into active participants in lawmaking.

The system faces challenges—it is time consuming, often duplicative, hard to manage, and produces primarily indirect legislative outcomes. But now Brazil has the opportunity to use artificial intelligence to supercharge this participation and make it more directly relevant. As LinkedIn co-founder Reid Hoffman writes in his new book Superagency: “Instead of thinking of AI first and foremost as a mechanism that could be harnessed for command and control governance, through applications like facial recognition, predictive policing, and algorithmic surveillance, we can choose a future where AI is used to connect citizens more substantively to legislative processes.”

The Brazilian Senate’s established participatory framework provides the perfect foundation for AI enhancement, allowing Brazil to overcome existing limitations while building on citizen engagement mechanisms already deeply integrated into its democratic institutions.

The Legislative Idea Platform: The People’s Proposals

Through the Senate's e-Citizenship website, any Brazilian can propose new laws directly to their Senators. The process, known as the Legislative Idea, launched in 2012 and is remarkably accessible—citizens submit a brief tweet-length proposal with their first name and state of residence. For those who prefer speaking or are less comfortable with technology, a toll-free number connects them with Senate staff who capture and submit their ideas online. Additionally, people who are deaf can submit ideas in sign language by sharing a video with the e-Citizenship team.

The Senate's 15-person e-Citizenship team reviews submissions—over a thousand each month—to ensure they comply with constitutional requirements. Once approved, ideas remain open for public support for four months. The three most voted-up ideas from the last 24 hours are displayed on the portal's homepage, encouraging public review and participation.

If a proposal collects 20,000 signatures, it is sent to a committee where it can be developed into either a formal bill or a proposal for constitutional amendments. While Senators don't always wait for these thresholds—they can champion promising ideas earlier in the process—the Human Rights and Participatory Legislation Committee plays a key role in transforming qualified proposals into formal legislation. The system has proven remarkably engaging, with over 11 million signatures recorded across various legislative ideas.

Recently, a suggestion to "Use the billion-dollar budget approved for electoral campaigns to rebuild Rio Grande do Sul, specifically the areas affected by the biggest environmental and humanitarian disaster in the region” was approaching the 20,000 up-vote threshold.

Challenges and Limitations of the Legislative Idea Process

While the Legislative Idea process has successfully brought citizen voices into Brazil's lawmaking, several challenges remain. Achieving tens of thousands of up-votes is a substantial hurdle, especially since only the most popular proposals appear on the home page. With over a thousand ideas submitted monthly, promising proposals can be overlooked. Additionally, similar ideas often split support across multiple proposals, diluting their impact.

Upgrading the Legislative Idea with AI

Generative AI (GenAI), a form of machine learning which recognizes patterns in language, could significantly reduce duplication. GenAIcan analyze new submissions against existing ones, flagging overlapping content, allowing Senate staff to automate the process of inviting submitters to join existing proposals before posting their own. This step could reduce fragmentation of support and streamline the review process.

Beyond reducing duplication, AI could also help Senate staff to organize proposals by theme, giving political leaders insight into public concerns categorized by geography and topic. GoVocal, a citizen engagement company in Belgium and Remesh, another engagement platform from the United States, have both developed clustering algorithms for grouping related ideas that, if used, could facilitate a more holistic approach to addressing public issues.

The current system's display of only the three most popular ideas on the front page limits exposure for newer submissions. Iceland's Better Reykjavik platform, developed by the nonprofit technology provider Citizens Foundation, randomizes the display of public submissions to ensure all ideas get public exposure.

Brazil could implement AI-driven rotation that's more sophisticated than simple randomization. Algorithms could prioritize ideas based on engagement metrics and relevance to current events, ensuring that trending proposals are highlighted while giving less popular submissions fair visibility. Such a dynamic system could adjust featured submissions based on community interests, fostering inclusive discourse by highlighting resonant ideas.

One persistent challenge in public consultation is the overwhelming volume of diverse comments that parliamentary officials must process. Brazil needs a system that can categorize feedback based on key themes, sentiments, and patterns, helping to distill complex inputs into concise summaries while ensuring diverse viewpoints aren’t lost. Fortunately, they can build on promising research in this area. Google DeepMind's recent "Habermas Machine" experiment involving over 5,000 participants demonstrated that AI could effectively synthesize diverse viewpoints into consensus statements that participants found clearer and more representative than those created by human mediators. The system was particularly adept at incorporating minority perspectives rather than simply reflecting majority opinions. By adapting similar technology, Brazil could more effectively integrate the full spectrum of citizen perspectives into the legislative process, making the overwhelming task of processing public input both more manageable and more inclusive.

The Interactive Event: Citizen Participation in Senate Hearings

Unlike in-person attendees who must remain silent in the Senate gallery, Brazilian citizens can actively participate in committee hearings through the e-Citizenship portal or a toll-free hotline. In 2024 alone, the public submitted 69,000 questions across 440 committee hearings. Senators read or responded to 90% of these questions.

The e-Citizenship team selects and relays the most relevant citizen questions to committee chairs. The process has become so integral that Senators now expect these public inputs—as one staff member notes, “If we are late, which is rare, the Senator asks, 'where's the list of questions?' Public comments have become the rule, rather than the exception.”

Challenges and Limitations

Despite its success, the Interactive Event faces significant challenges. Processing tens of thousands of questions requires substantial staff resources. Ensuring selected questions are relevant, constructive, and representative of diverse viewpoints requires careful curation. The current manual process may not always capture the most impactful or pertinent questions. Additionally, staff must maintain parliamentary decorum by filtering inappropriate content.

Despite efforts to notify a wide range of potentially interested parties, there's a risk that certain demographics or interest groups may be overrepresented in the question pool, potentially skewing the focus of hearings.

Upgrading the Interactive Event with AI

AI is already beginning to transform the hearing process. The e-Citizenship team currently uses AI to select the best and most relevant questions from citizens, but the technology offers far more potential. The team is turning to generative AI to write brief summaries of events to help the public understand the topics of congressional hearings.

With text-to-video and audio capabilities, AI could transform written backgrounders into short videos, slide decks, or podcasts, making information more intelligible to those without policy backgrounds. For example, NotebookLM, a free tool from Google, creates podcasts from documents. While currently available only in English, a Portuguese version could generate conversational discussions of upcoming hearings, making complex topics more accessible to average citizens.

Currently, submissions are only accepted in Portuguese. In the future, AI translation could accommodate questions in other languages and handle transcription and organization of voice or video submissions. These content sorting tools would reduce staff workload, route queries to the appropriate committee chairs more efficiently, and ensure diverse topics are addressed during hearings.

For citizens, AI-powered chatbots could help navigate the process of submitting questions, understanding hearing procedures, and accessing information about discussion topics. This would lower barriers to participation, especially for first-time users or those less familiar with legislative processes.

Integrating AI-driven transcription into committee meetings could further enhance efficiency. By transcribing discussions in real-time, AI could compare these transcripts with submitted questions to identify whether certain topics have already been addressed and suggest which citizen the Senator should recognize next. The system could also recommend follow-up questions that logically extend the current discussion, ensuring more coherent and productive dialogue. This would streamline moderation and maintain hearing flow, enabling Senators to focus on relevant and timely queries.

AI could play a critical role in parliamentary committee engagements by moderating citizen interactions during public hearings and consultations. Beyond simply filtering inappropriate content, AI tools could organize citizen questions to ensure the most relevant and constructive issues are addressed while maintaining focused discussion. AI-powered transcription could compare citizen inputs across sessions, preventing repetitive discussions and highlighting unresolved topics. Instead of simple yes/no votes, AI could facilitate online deliberation forums where citizens discuss bills in real-time, with the system analyzing these discussions to provide lawmakers with nuanced insights into public sentiment, concerns, and suggestions.

The e-Citizenship team has already experimented with using AI to generate comprehensive reports summarizing hearings and plans to make this a standard practice. These reports would provide valuable records for both Senators and the public in understanding hearing outcomes and impacts.

Early experiments with immersive virtual worlds have been attempted but haven't persisted due to the expense and distraction of headsets and goggles. As technology evolves to become more accessible, AI could create virtual environments for hearings, allowing citizens to feel present and engaged when participating remotely. Instead of typing a question and watching on YouTube, participants could experience being in the virtual gallery.

By leveraging these AI-driven enhancements, the Brazilian Senate could significantly improve the efficiency, accessibility, and effectiveness of the Interactive Event process, fostering even greater public participation in the democratic process.

Legislative Workshop: Young Minds Shaping Policy

To train the next generation of citizens, the Brazilian Senate has developed a curriculum for elementary, secondary, and higher education that enables schools and universities to teach young people how to develop legislative ideas under teacher guidance. The e-Citizenship team conducts regular outreach to university leadership to promote youth involvement and foster civic engagement among Brazilian young people.

This investment in civic education has yielded results: 10% of submissions to the legislative ideas site now come from educational institutions, with students having developed two thousand different legislative proposals. In 2023 alone, more than 1,500 teachers registered to conduct Legislative Workshops in a hundred cities across every state in the country.

Students learn about governance while digging deeply into national problems to develop legislative solutions, which they discuss in the classroom. In universities, this exercise is often integrated into political science or public administration courses, where students learn about the branches of government, institutional jurisdiction, and the lawmaking process before developing ideas to improve legislation.

The program has been inclusive, with workshops conducted in schools for those with intellectual disabilities. One compelling example came from Marcelo Siqueira from the Federal District, who proposed introducing hanging vegetable gardens to promote healthy eating in centers that educate those with disabilities. While the proposal received few up-votes, Senator Paulo Paim of Rio Grande do Sul found merit in the idea and introduced it as legislation. As Senator Paim noted, "Participation in this type of activity has widely recognized benefits in terms of mental health, strengthening community ties and the comprehensive development of children and young people, in addition to contributing to the sense of belonging of the elderly and people with disabilities."

Challenges and Limitations

Despite its success in engaging young people in the legislative process, the Legislative Workshop program faces several hurdles. The quality and feasibility of student proposals vary widely, requiring significant resources for review and refinement. Ensuring equitable access across diverse school systems remains a challenge, potentially limiting the program's reach. Moreover, maintaining student engagement beyond the workshop and translating enthusiasm into long-term civic participation poses ongoing difficulties.

Upgrading the Legislative Workshop with AI

By incorporating AI technologies, the Legislative Workshop could become even more widespread and impactful, revolutionizing how young people learn about and participate in the legislative process.

One of the most significant improvements would be personalization. AI could analyze each student's interests, learning style, and progress to create tailored learning paths within the curriculum. This would ensure students remain engaged and challenged regardless of their starting knowledge level. For students with disabilities, AI could provide real-time captioning, text-to-speech, or other assistive technologies to ensure equal participation, making the program truly inclusive.

Beyond individual learning, AI could foster unprecedented collaboration. By registering their initial interests, AI could suggest potential partnerships between students or classes working on similar topics, fostering cross-school collaboration and idea exchange. Classrooms could organize online discussions around their projects and work on joint proposals, rather than duplicating efforts and overwhelming Senate staff with numerous underdeveloped ideas.

This collaborative approach could extend internationally as the Legislative Workshop concept spreads to other countries. Brazilian students could collaborate with peers from the United States, India, or Nigeria on global issues like climate change or digital privacy rights. AI-powered translation tools would break down language barriers, enabling truly global civic engagement among youth.

To enhance the learning experience, AI could power virtual simulations of the legislative process. Using AI-driven virtual or augmented reality, students could participate in lifelike recreations of parliamentary debates or committee hearings. These simulations could adapt in real-time based on students' decisions, providing a dynamic learning experience that brings the legislative process to life more effectively than textbooks.

AI could also provide instant, constructive feedback on students' legislative proposals, highlighting areas for improvement and suggesting resources for further learning. This would complement teacher guidance and allow for more rapid iteration of ideas, leading to better proposals and a higher likelihood of student ideas becoming actual legislation. The AI could be trained on successful past proposals and current legislative priorities, ensuring relevant and actionable feedback.

Finally, AI could help students navigate the complex ethical considerations inherent in lawmaking. An AI system could prompt students to consider various ethical implications of their proposed legislation, encouraging critical thinking about the broader impacts of laws on diverse populations. This could include simulations of how a proposed law might affect different demographic groups or regions over time, giving students a deeper understanding of the far-reaching consequences of legislation and the importance of thoughtful lawmaking.

The Legislative Workshop program represents a powerful investment in Brazil's democratic future. By engaging young people directly in the legislative process, it not only educates them about governance mechanics but also instills civic responsibility and empowerment. The program's success in generating thousands of student-led proposals, some influencing actual legislation, demonstrates its potential to shape a more engaged and informed citizenry.

As these young participants grow into voting adults, they carry with them not just knowledge of how laws are made, but the confidence and experience that has come from participation and that might blossom into even more active citizenship.

Public Consultation: A Citizen Referendum

The Federal Senate has also created an online system of public consultation that allows citizens to voice their opinions on pending legislation. Over thirty million yes or no “votes” on bills have been recorded over the last decade.

The process, formalized by a 2013 resolution, opens all bills to public scrutiny. The law stipulates that “The Federal Senate website will house a mechanism that allows citizens to express their opinion on any legislative proposal” and that “any citizen, through a single registration with their personal identification data, may support or reject legislative proposals being processed in the Federal Senate.”

This type of non-representative plebiscite provides a snapshot of public opinion rather than an occasion for deliberation. Citizens are not asked to support their votes with data or evidence, and there's no opportunity for debate among those with differing opinions. While limited in its scope, the system at least offers a small way for people to participate in governance.

The aggregated voting results are public, and the E-Citizenship team regularly shares them with parliamentary offices to ensure that citizen voices reach policymakers. To be clear, Senators are not obligated to vote in line with public consultation results. Instead, the system provides lawmakers with a pulse check from among the 15 million registered participants and the 34 million public votes

Challenges and Limitations

While the Public Consultation system has successfully engaged millions, the yes/no voting format can oversimplify complex issues. There's no opportunity to provide data, evidence, or informed opinions, and citizens may vote without understanding proposed legislation. The system also risks influence from coordinated campaigns by interest groups encouraging supporters to vote one way or another. Moreover, the non-binding nature of consultations means politicians can ignore results, potentially causing citizen frustration.

Upgrading Public Consultation with AI

AI and advanced technologies could dramatically improve citizen participation in the lawmaking process, enabling Brazil to enhance its system.

Complex legislative language presents a barrier to participation. AI is already generating easy-to-understand bill summaries. Going further, AI could develop balanced explainers discussing potential impacts, pros, and cons in plain language. AI image generation could make this content more engaging through infographics or short videos, making legislation accessible to broader audiences.

AI models could simulate potential effects of proposed legislation on different demographics, regions, and sectors. Citizens could interact with these models to understand how a bill might affect them personally, their community, or the country as a whole.

A critical challenge identified by Jigsaw, Google's unit focusing on technology and human rights, is what they call the "last mile problem"—transforming citizen participation into actionable insights for institutions. They have been experimenting with using generative AI to write one-pagers with concrete takeaways about areas of agreement that could inform legislation and points of contention requiring further deliberation. Imagine if Brazil's Senate expanded beyond yes/no votes, using AI to synthesize citizen comments for Senators.

Brazil has an opportunity to redesign public consultation around more targeted questions that would generate directly useful input for lawmakers. By combining AI-powered analysis with purposeful conversation design, Brazil could create a virtuous cycle where citizen input more directly informs legislative priorities.

The e-Citizenship portal already requires registration, though voting remains anonymous. With AI, citizens could register their interests, and the system could match bill content to those interests, notifying people of pending legislation relevant to their concerns or communities. This personalized approach could increase engagement by connecting people with issues that matter to them.

For issues that transcend national boundaries—like climate change, internet governance, or global health—AI could facilitate cross-border citizen deliberation and collaboration. The system could aggregate global opinions, allowing Brazilian citizens to collaborate with counterparts in other countries on shared challenges. This acknowledges that many pressing issues require coordinated transnational responses informed by diverse global perspectives.

Through these enhancements, Brazil's already impressive citizen engagement system could become even more deliberative and impactful, setting an example for participatory democracy worldwide.

The Way Forward: Reimagining Democracy for the Digital Age

As we’ve explored the Federal Senate of Brazil's four innovative mechanisms for online citizen engagement—the Legislative Idea, the Interactive Event, the Legislative Workshop, and Public Consultation—all poised to benefit from the integration of AI to improve their efficiency and effectiveness.

Brazil's model offers valuable lessons for democracies worldwide. In an era where many nations grapple with political polarization, declining trust in institutions, and a lack of political accountability, Brazil's approach provides a roadmap for revitalizing democratic participation.

It shows that with careful design, strong institutional support, and thoughtful integration of technology, it is possible to create more participatory and responsive democratic institutions and foster the kind of informed, engaged citizenry that healthy democracies require.

Modernizing Legislative Drafting and Publishing: How to Construct a Tender or Request for Proposal (RFP)

Marty Bickford — Tue, 05 Aug 2025 20:01:31 GMT

Recommendations from Xcential for procuring a new system for legislative and regulatory drafting solution

Executive Summary: The Vision for Modern Legislative Technology

The Problem

Parliaments and legislatures around the globe have a problem with existing systems used to draft and publish both legislation and regulation. Existing systems make it difficult to create or capture value relative to efficiency, precision, and integration.

Procurement and implementation of a new system that solves for efficiency, precision, integration, and future-proofing comes with a different set of problems. These problems are often addressed through a government Request for Proposal (RFP) or Tender (note to reader: we will use these term interchangeably). To get the most from a new system, the RFP/Tender will include a list of requirements and an expected cost.

These RFPs/Tenders often fail. Between 2020 and 2025, many RFPs/Tenders were issued that did not attract acceptable bids. There are several reasons:

High cost of bids. Either vendors cannot deliver requirements within a fixed price bid or the estimated cost on a time and materials basis is unacceptable.
Low vendor participation. With complexity of requirements, many vendors cannot fulfill enough of the requirements or they believe there is too much risk (especially on fixed price bids.
Reduced innovation. When an RFP/Tender has an exhaustive list of requirements that are both specialized and specific, there is no room for discovery and innovation. Innovation requires trust and collaboration. Innovation also requires a look at first principle problems. So ironically, the volume of specifics directly removes room for innovation which is part of the value governments are seeking to capture.

The result is unmet needs. Time is wasted by valuable government staff, and the organization is disappointed by continuing to work with a legacy system.

The Opportunity

Software and standards designed specifically for legislative and regulatory drafting, management, and publication have advanced considerably in recent years.

These systems provide several advantages to parliaments, regulatory agencies, economic development, and citizens.

Modernization & Future Proofing
Modernization of government technology is happening across parliaments, legislatures, and agencies. These new systems bring integration and interconnection between systems, creating efficiencies for users and consolidation of system administration. New systems can be deployed and used on-premises and in the cloud, on the desktop and within the browser. New systems can be updated more easily and managed by government staff. New systems should be future-proof, with a path forward to future updates, integrations, and UX. Artificial intelligence (AI) may not be necessary today but your new system should accommodate it in the future.

For example: You can prepare for Rules as Code. The combination of LegisPro from Xcential and the UNISYS business rules engine can extract business rules from legal and regulatory text in preparation for Rules as Code. UNISYS is an Xcential partner and describes this as “yyperautomate”. This rule extraction could be used to answer questions from citizens while avoiding AI hallucinations.

Standards and Interoperability
Standards in the legislative and regulatory space have advanced. LegalDocML and Akoma Ntoso create precision in drafting and consolidation, they enable the building of modular software systems (instead of monolithic systems) for enhancement and maintenance, and they enable transparency for rule-makers and citizens. Moving to LegalDocML and Akoma Ntoso is a meaningful achievement and creates value.

Automation
Automation is available to create efficiency and precision. New systems can automate the creation of required documents such as lists of amendments. Page and line numbers can be automated. Advanced tracking of changes can automate the application and merging of amendments to bills. Producing versions for publication and printing can be automated. Status reporting is reported automatically as drafting crosses defined stages. The management of citations and sections (even as they move) is automated. Automation not only reduces time spent on routine activities, but it reduces the opportunity for error which leads to greater precision.

Integration
Integration is expected in new and modern systems. System management is simplified and secure when user roles and permissions are managed through integration. Data sharing between systems creates the opportunity for better management and reporting. But most importantly, integrated systems allow for the advancement of individual components in a targeted manner.

Finally, a new system that is (a) based on LegalDocML and Akoma Ntoso, (b) integration ready, and (c) prioritizes automation and precision will create more transparency and a platform for improved citizen engagement.

Every parliament (with an older system) desires a modern system that delivers on the fundamental charter to support the rule of law through transparency and precision.

This goal is within reach.

The Xcential Blueprint

Realizing the opportunity above requires great procurement.

Xcential has seen many RFP/tenders for legislative drafting and publishing systems. Below are some recommendations on constructing your RFP/tender. We hope it is useful.

Introduction: Why Rethink Procurement?

Evolving Legislative Landscape

The demands on legislative and regulatory drafters, publishers, and management are increasing. These demands include: volume of bills and acts, complexity, expectations on government efficiency, and the role of legislative drafting services.

The sheer volume of bills and acts has increased over the past years. While the numbers are expected to go up and down from year to year, the numbers appear to be going up.

For example in the United Kingdom the number of bills that became law increased:

2020 = 29
2021 = 35
2022 = 48
2023 = 57

Another example is the volume of bills introduced by state legislatures in the United States:

2020 = 73,250
2021 = 117,188
2022 = 75,675
2023 = 121,537

Complexity is ever increasing. Our democracies tend to address problems when citizens demand action. Governments put efforts into healthcare, then climate change, then global trade, and then government efficiency. These efforts are related and are never resolved. Each issue is important and intertwined with others, so a change to climate change law may affect global trade and perhaps even healthcare.

We don’t take laws off the books very often, so in each legislative or regulatory session the complexity continues to grow.

Government efficiency has become a prominent issue in the United States. As citizens, we should expect the efficient and effective use of public funds. We should always expect that funds will become less available.

Finally, one area Xcential sees evolving globally is the role of a professional, non-partisan group of drafters. We often refer to this group as a Legislative Counsel. They draft bills and amendments, prepare required documentation, and ensure elected officials can introduce high-quality legislation (independent of political affiliation). Most national and sub-national governments employ a group to serve elected representatives in this way. It is typically optional for elected representatives to use these groups, but their use is becoming more common and the volume is increasing. When you add complexity and efficiency, the strain on these professional drafters is ever increasing.

Governments' desire for technology transformation, modernization projects, and especially to demonstrate a use of artificial intelligence, is another meaningful change in the landscape.

Finally, legaltech is evolving rapidly. This makes it important to position the RFP/tender to capture new ideas from new vendors. Make room for new ideas and new technologies.

Pitfalls of Traditional RFPs

Everyone who invests the time to issue an RFP or Tender hopes and expects it to be successful.

Xcential has recognized some key pitfalls that can lead to poor results:

Overcustomization‍

It is important to create an RFP that covers the legislative and regulatory process in full. It is also important to separate the process required by law from the processes that evolved around systems to get the work done. RFPs/tenders that require customization to solve for individual drafter preferences will result in complication, costs, and unneccessary risk.

SaaS software proposes that software can be less expensive (total cost of ownership) if you subscribe to the way that particular SaaS software has been implemented. The users are asked to compromise how they work. Custom software does not ask for this compromise and instead is built to match existing ways of working (at a high cost).

It is common for a seemingly small customization to result in disproportionate costs and complications. Customization is required, but be cautious.

Lack of modular thinking

Systems are defined by what connects them. One path to a legislative system is to connect related actions within one large blob of customized code, this appears to solve all your requirements in one big upgrade. But the blob makes it expensive to maintain, locks you into a specific vendor going forward, and reduces your ability to upgrade specific parts of the system. It is better to create a modern, modular system where components are integrated but can be updated or enhanced without disrupting the entire system.

Ask vendors or technologists about the value of components versus a monolithic system. Modern engineering teams will promote Agile as a method for rapid prototyping and the incorporation of user feedback to get to a great system. Learn from this approach and why components create agility and long-lasting systems.

Market realities are often overlooked

Software vendors need to make a profit. Profit is related to risk reduction and economic development. When requiring a fixed bid, risk will be built into price. Smaller (often more innovative) software firms will either not bid or account for low-risk tolerance into the price. Smaller companies may not survive a fixed bid that is incorrect. The market demands a reasonable risk versus reward calculation in favor of the contractor. And the level of risk will directly affect who bids and who the government has an opportunity to work with.

In 2024, there were several tenders for legislative and regulatory systems that failed. This is because they asked for very complex systems that could not be delivered successfully by vendors.

Focusing on How instead of What

The “what” is the problem. For example, “We need a way for customers to easily see their order status.”

The “how” is the specific technical solution. For example, “We need a mobile application for customers to run an order status report.”

In the above example, the “what” leaves options open for building a web application, a responsive website, SMS updates, or even integration with natural language processing chatbots. By defining the “how” we are asking for proposals that eliminate available options and future innovation.

Gathering a collection of “how” statements from current users can result in a long list of features without understanding their true necessity or priority. This creates situations where a more elegant and cost-sensitive approach could solve all the “what” statements, but a less ideal solution will be chosen to accommodate “how” statements.

When “how” statements are exhaustive, you can almost guarantee that costs will be high and innovation will be low.

Carefully consider requirements that address legislative rules instead of the way staff is accustomed to working. Expect change management to be challenging, but alignment against “what” instead of expectations of “how” will make change management easier.

Paradigm Shift

If we want to design for a successful system and a successful procurement, we need to move away from “build it all at once” to “integrate smartly” or “build modular components in an agile methodology”.

We need to abandon monolithic and highly customized systems to component-based systems that are interconnected and future-proof. We need to foster innovation and better ways of working, instead of exhaustive lists of specific, user-defined “hows”.

The Xcential blueprint for successful RPFs/Tenders is to (a) build with LegalDocML or Akoma Ntoso to create a component-based system, (b) define requirements based on what is needed, not how it will be implemented, (c) leverage integrations with other systems to improve workflow, management and total cost of ownership, (d) consider the market realities of large and small software vendors, and (d) leave room for innovation, agile, and changing requirements as prototypes and MVPs come online.

Phase 1: Strategic Planning (Before the RFP)

Internal Alignment and Stakeholders

Internal alignment within a legislative drafting office, legis counsel, or regulatory agency is strong. It is a part of the cultures of these institutions. In practice, alignment is required to get through procurement.

Alignment is easy when your message to everyone in the organization is “you’ll get what you ask for and this will be painless.” Alignment is challenging when your message is “we will get a better system, but it may require you to change.”

No one likes change. But change is why we write software. So we plan for change and we manage change. Change management begins when we first consider the project, how we specify the project, and what expectations we set early.

Who needs to be Involved

Before the RFP/Tender is constructed and published, it is important to carefully consider who is contributing to requirements.

Focusing on “what” instead of “how” is not always easy for everyone, we all think differently. Take everyone’s feedback and include what you believe make for good RFP/tender requirements.

Leadership needs to be involved. Who is leadership? Those who shape the thinking of your organization. Understand what is critical to leadership in terms of major initiatives, budgets, timing, and reportable outcomes. If you can deliver reportable outcomes (positive achievements for their stakeholders) for major initiatives on time and on budget, then you have large objects to construct your project around.

IT or Information Technology needs to be involved early. IT is moving in a direction. That direction might be to control all user access from a central control panel. Another might be to move towards vendor managed software on a sovereign cloud infrastructure. Learn how they deploy today and how they want to deploy in the near future. Talk to individuals who have had large-scale, complex system deployments achieved recently; learn why their projects worked and solicit their advice.

Individuals within your legislative drafting or legislative counsel organization are vital. Talk with people who understand what is done on a daily basis. Managers who are accountable to other departments (like publishing and parliamentarians) can help you understand the requirements and flexibility required to integrate with other functions.

Ask leadership and IT questions like:

What does success look like?
What existing systems could this new system replace, augment, and integrate with?
What is in and what is out of this RFP?
Is there an MVP that we can define and accept?

Now we need to ask questions that will highlight functional requirements. These questions should be asked of individual users, managers, and members of your drafting and publishing organizations.

What are the document journeys? What are the user “jobs to be done” during each of these journeys?
How is work allocated, communicated, and reported on?
What information and documents need to be created, stored, and used during these journeys?
How is data gathered and used? Ask about inputs, tasks completed, states of documents, and necessary outputs.
How do you find information you need in various stages of a drafting project?
How are mistakes handled? What events tend to be error prone?
What processes do you feel are inefficient? What processes seem like they could be automated with technology?
What are the most time consuming jobs to be done?
What absolutely has to be included in a new legislative drafting and publishing system?

None of the feedback you get will go directly into your RFP or tender. The skill is to listen closely and learn what you believe makes for good requirements. Be careful to limit expectations on items that we call “nice to haves” as expectations lead to difficult change management and staff being unhappy.

Listening for good requirements can be difficult, so we suggest the following.

Assessing Needs is about Pain

Smart folks will want to explain to you what the new solution should be. Let them. But we have to focus on the pain of the existing process. Pain says there is opportunity to improve over the existing system. Maybe you have to work overnight when a bill is introduced, this tells you there is an opportunity for automation to make staff more efficient and happier.

Ask about bottlenecks and inefficiencies. You’ll find some of these occur infrequently. Dig deep because often bottlenecks and inefficiencies that occur regularly may be overlooked; they become part of the normal, expected process.

Desired outcomes are another way to discover pain points. Staff may be asked to do things that are not quite in reach, but believe they should be. Compliance with other departments and legal rules are desired outcomes, so discover if there is any pain here as well. It is critical for change management, that you solve pain when rolling out a new system.

Future-Proofing

Leadership and IT are concerned with future-proofing. You should consider this as critical as well.

First, do your own research on legislative document standards. Research LegalDocML, Akoma Ntoso, and LegalRuleML. Understand for yourself, if standards matter or not. Do you agree with the perspective of component-based systems? Will you defend the value of standards?

Cloud versus On-Premises is straightforward. Learn from IT what they prefer and align to the future of your organization. Learn why IT is going this direction.

APIs and integration with other systems will be a hard requirement, no system lives alone. Again it is IT that can provide the right answers on preferred methods. For example, IT may prefer that users are authenticated with a common service and roles managed from a centralized panel. Business process triggers and expected outcomes from other departments and should be reviewed with IT.

Xcential does not recommend hard requirements for artificial intelligence (AI) and natural language processing (NLP) in your RFP/tender. However, leadership and innovative IT staff may push you in this direction. There are great opportunities with these technologies, like conversational search that makes searching for documents much more enjoyable. Our recommendation is to understand how AI could be used in the future and ensure that your new system is able to extend and adopt new services.

Realistic Budgeting and Phased Implementation

Budgeting and phased implementation are items you must define. Construct them with gathered feedback.

Total Cost of Ownership (TCO) means how much does this system cost to implement and operate. It includes purchase price, taxes and fees, installation and setup, configuration, customization, training, financing, and hosting to get the software. TCO also includes licenses, maintenance, support, and head count required with a new system.

Component-ortientation or a modular approach supports a phased approach to implementation. Define what you need to see in a proof of concept before awarding a contract. Define what you need in an MVP which allows staff to complete end-to-end work, before implementing automations and “nice to haves”.

All software vendors can do continuous integration and continuous delivery; this is “how” they deploy solutions. But not all are great at delivering MVPs, because this requires a deep understanding of journeys and needs. Define the MVP yourself (with input from others), it will ensure you are prepared to make decisions about tradeoffs and priorities. It also puts you in control of timelines.

MVP is directly related to the budget. You know the budget. The challenge is creating an RFP/tender that virtually guarantees a successful project on budget.

One budget issue is how you will pay for the project, fixed cost bid or time & materials. Vendor gouging on time & materials is real; but it is not hard to see and correct. The problem with a fixed cost bid is risk. Risk limits global trade and risk limits vendor interest in your project. This is especially true of smaller software vendors who know that one or two bad decisions can be an existential threat. Time & materials contracts create agility, flexibility to adjust requirements and budget, and will solicit innovation from the best vendors.

Phase 2: RFP Blueprint (sections and content)

After you’ve set your strategic framework and gathered valuable input on what should be included, it is time to draft the RFP or tender.

Below we’ll list some basic components of a good RFP. Most importantly, we’ll provide a vendor perspective on these components.

Then we’ll provide some example requirements that you are free to use for a real RFP for a new legislative drafting and publishing solution. These sections can be combined, reordered, or reframed to match your style, but each has a specific purpose.

Project Overview and Narrative

Executive Summary

Purpose of Executive Summary

Summarize the RFP for the reader with a clear and concise description describing the overall project. This is an opportunity to highlight perspectives and facts that will be considered critical to your evaluation.

As a vendor it is easy to get overly focused on details of features, hosting, pricing, and other important parts of the RFP. Pull those out and summarize what is critical to the success of a deployed legislative drafting solution.

We recommend drafting a summary early and then redraft the executive summary as one of your last items. Remember the Executive Summary can be followed by the Purpose which will dive deeper into what is important about the RFP, so consider how these sections read together. But you may also combine Executive Summary, Purpose and other sections to create a larger narrative section.

Example Executive Summary

"We seek to identify and appoint a suitable service provider to customize, implement, and maintain (for a period) a legislative drafting and publishing system to be used in the drafting, amending, and processing of bills and other legal instruments in the parliamentary legislative process. We require a legislative drafting and publishing solution that is built on modern technology and open standards.

It is expected that vendor will transfer skills. Our staff should be trained to set up, operate, modify, provide support and maintain the legislative drafting and publishing system."

Purpose of the RFP = State What You are Trying to Achieve

Purpose of Purpose

Explain in this section “why this RFP exists”. It is to select a vendor to build your solution. Or at least to narrow options down to three.

You are selecting a vendor to build the solution. So the RFP is meant to find someone you can pay to build your system.

Next we explain what we will look for in both a vendor and a solution. We are telling the vendors that we want great software, but that we also want to work with a great company.

There are great companies who don’t make a good match for a legislative or regulatory solution. Here are some characteristics of a good vendor for this project:

Mature project management processes and manager(s)
Have delivered similar systems
Stable in finances and staffing
Practice agile and consider both hosting and CI/CD from the start
Understand amending, precision, and parliamentary processes
Dedicate themselves to open standards (LegalDocML) and open integration with others

Things to consider positive, when evaluating technical responses to requirements:

Coverage of most important requirements and ability to provide custom solutions. It is important that the planned solution has the flexibility to be built upon, when desired.
How much of the software will be configured versus developed by requirement. We want a solution that is “out of the box”, as much as possible. In the legislative and regulatory space, each jurisdiction has unique requirements that will require customization, but we want to minimize customization as much as possible.
Integration will be necessary for authentication, workflow, reporting, document storage, hosting services, databases, and more. How does the solution’s components communicate with one another (drafting versus workflow) and how do components communicate with non-solution components (authentication). What APIs, webhooks, endpoints, microservices, and serverless functions are available?
The cost is a critical component. There are two factors that will drive up your development costs: large consulting corporations and custom development. Both have benefits, which is why they cost extra.

Example Purpose

"We have chosen this RFP process to identify an industry vendor to build a LEGISLATIVE AND PUBLISHING SOLUTION. By reviewing responses to this, we will evaluate vendor characteristics and proposals required to successfully deliver this complicated solution.

Vendor characteristics evaluated will include project management, history of related work, financial stability, technical perspectives, maturity of technical processes, and legislative understanding.

Proposals will be evaluated on requirement availability, custom development required to complete solution, integration options, project management, and total cost of ownership.

This RFP may result in the selection of a vendor; or it may result in the selection of two or three vendors to proceed to a proof of concept. The proof of concept will be designed in communication with the selected vendors."

Organization Overview

Purpose of Organization Overview

In this section, explain what your organization does. Explain how you serve the parliament and how you serve other organizations. Explain the roles and responsibilities of who use the solution. Vendors will want to know the number of users as well.

Example Organization Overview

"We are responsible for the critical and complex process of drafting bills and other legislative and regulatory instruments. We are a professional and non-political department created to serve a bicameral legislature and draft documents for them at their request.

Primary roles and users of the new solution include:

Elected Official or Staff - These users will request drafting projects. They may also review drafts during drafting projects.
Drafters (10) - These users draft bills, amendments to bills, amendments to amendments, and much more. Drafters manage metadata, versions, lists, sponsors, reviews, comments, and more. They prepare documents for use by parliament and publication.
Reviewers (50) - These users review drafts and final documents for content, procedural and legal compliance, and format. Their feedback is provided to drafters.
Publishers - These users will receive output from the solution in expected formats and publish according to jurisdiction procedure.
Consolidators (3) - These users will take enacted bills (or other documents) and consolidate them into new authoritative versions of law.

Key Goals and Desired Outcomes

Purpose of Key Goals and Desired Outcomes

Describe what success looks like. These are not requirements but outcomes.

Technical requirements explain the details required to perform jobs to be done in the expected manner. Key Goals and Desired Outcomes provide a vision for what you expect to be true after a successful RFP/Tender and a successful project.

Example Key Goals and Desired Outcomes

Key goals of this RFP/Tender include:

Find suitable partners to execute the vision for a Legislative Drafting and Publishing Solution
Find a suitable partner who can execute this project within an acceptable timeframe
Find a suitable partner who can execute this project within an acceptable budget
Find a suitable partner who can execute this project with technical expertise and vision
Find a suitable partner who can execute this project with excellent communication, partnership and innovation skills

Desired outcomes of this project:

Drafters can create required legislative and regulatory instruments required to support the Parliament and do so according to process requirements
Publishers can publish legislative and regulatory instruments in a manner that supports existing process and tools, supports expectations of transparency for citizens, and creates efficiency of time and money
A modern legislative drafting and publishing solution that is more efficient in time and in total cost of ownership
A modern legislative drafting and publishing solution that supports open standards and open programmable interfaces to create immediate and future opportunities for integration and interoperability
IT staff and users that are trained to use the system; and have the materials required to train future users"

Scope Summary

Purpose of Scope Summary

Give vendors a high-level vision of the solution. Answer the question, “what is a legislative drafting and publishing system?”

Scope sets the boundaries for the solution. It includes what will be included and what is not included. Requirements will flush out scope, so in the Scope Summary we are drawing large, bold boundaries. It is often helpful to describe what is out of scope to make those boundaries clear.

Scope is not just the left and right sides of a user journey, but includes specifics about how jobs are done, technical formats, integration, and the like. Consider which boundaries are important and highlight them in the Scope Summary. For example, the use of Microsoft Word vs XML editors for drafting may be an important boundary that will help vendors know early if they can fulfill your RFP/Tender or not.

Example Scope Summary

"We will select a partner who can:

Deliver the legislative drafting and publishing solution that is implemented, configured, and customized to (a) draft legislative and regulatory instruments such as bills, resolutions, and amendments, (b) enact amendments, (c) manage all required meta data, (d) enables reviews and collaboration, (e) print documents in required formats for the parliament, (f) automate processes such as amendment generation, numbering, and document synthesis, (g) process bills, (h) publish legislation and regulation in coordination with parliament publishers, and (i) provide the management of documents as projects.
Work with staff to configure templates and workflow for use by staff
Deliver the required knowledge transfer for staff to operate the system and train new staff to operate the system
Integrate with existing systems to create efficiency for users and IT staff; we use Microsoft products and integration with these products will be weighed as positive
Deliver the solution in a secure hosting environment that meets security requirements
Base a system on drafting, publishing, and management of documents in XML and full support of LegalDocML to a component-based system”

Timeline and Budget

Purpose of Timeline and Budget

Your future partners are looking a few things:

Can we deliver this specific solution with unlimited time and budget?
Can we make a profit on this project?
Can we deliver this project on time?

Answers will affect their interest in bidding and should affect how they respond to requirements.

We all want successful projects, which does require a successful partner. We want partners to make a reasonable profit. We want them to be eager to support us post deployment.

In any project the dimensions of success include: work to be done, time to do it, money available to acquire time, and efficiency of time. Give your future partner enough information for them to make good decisions.

Example Time and Budget

"We expect to complete the RFP/Tender process and select a partner by [DATE]. At our discretion, this date may be extended for further evaluation of partners or at will.

The current expectation that the full legislative drafting and publishing solution will be in its final state of production (end state) by [DATE].

We abide by a legislative calendar. It is our goal to have the new system in place and in production for the session that begins in [MONTH] of [YEAR]. There users must be trained and testing completed (including a full mock session) by [MONTH] of [YEAR].

We will evaluate a total cost of ownership over a 5 year period beginning at the awarding of this contract. It is the responsibility of vendors submitting to this RFP/Tender to calculate portions of total cost of ownership within their control. Factors to be considered include: hosting costs, all required software licenses, software development and configuration costs, support costs, and project management costs.

Vendors may propose blended proposals which include fixed costs along with time and materials. We will give preference to proposals that relate fixed costs to requirement delivery.

We will evaluate success bids relative to time, money, quality of vendor, quality of solution, and risk. Bids which are not within an acceptable range on both time and money will not be considered. Bids that are not within an acceptable range of money, may be asked to reconsider costs at our discretion."

Functional Requirements

So far, we have outlined meaningful approaches to setting-up the RFP and customer expectations. Now we use the priorities and stakeholder feedback to construct quality requirements.

Our requirements will paint the picture of the solution and challenge vendors to provide a way to create the best solution. Your requirements should focus on the what, not the how.

The volume of requirements matters, therefore prioritize core functionality and non-negotiables. 10 requirements will not provide enough specificity for evaluation by either party. 400 requirements may overcomplicate responses and tax the human resources of small, innovative vendors.

Do focus on legislative rules and business processes. Be willing to accept flexibility on business processes. Be specific and informative on legislative rules. Legislative and regulatory processes are similar across most democratic parliamentary systems, but jurisdiction rules are often unique and specialized. Remember that many quality vendors who can build a final solution are not experts in your legislative processes.

Requirement Sections

Your vision was maybe covered in the above sections and you should consider short summaries of sections. Short summaries above requirement sections help vendors who must distribute requirements within their company.

Summaries are a great place to provide Use Cases. Use Cases explain how a user accomplishes a job with a system. Use Cases that include diagrams are great, if complex and available.

Here are some sections we feel are unique to an RFP/Tender for a legislative and drafting solution:

Drafting
Amendment Management
Publication
Document Automation and Archiving
Reviews and Collaboration
Legislative Project Management
System

Drafting Requirements

Amendment Management

Publishing

Document Automation and Archiving

Reviews and Collaboration

Legislative Project Management (LPM)

Systems

Technical and Integration Requirements

Your IT group and future direction will determine requirements in this section. Xcential will point out meaningful considerations, but does not provide sample requirements.

Open Standards Mandate

A modern legislative drafting and publishing system must be based around LegalDocMl and Akoma Ntoso. This is an immediate and long-term decision that pays benefits in money (component oriented system and no vendor lock-in) and long-term flexibility.

These standards create interoperability between documents and applications. The main reason to do all editing in XML is because of the value of the metadata, structure, and validation that makes interoperability real.

Migrate away from proprietary document formats when it is practical. Xcential is happy to consult on XML and legislative standards.

Security and Compliance

Give these requirements to IT and they may already have standard answers. Do not compromise on these relative to the end state of the solution.

But don’t miss out on small, high-quality partners if IT believes they could get through security and compliance requirements.

Scalability and Performance

Give these to IT as scalability and performance should be standardized requirements for applications and UX.

Infrastructure (Cloud vs On-Premises)

Hosting is not complicated. The recommendation on hosting should be driven by your strategic modernization plan.

If the goal is to move into a secure government cloud, then make that a requirement.

Integration Points (where integration is a must)

Integration is a must, we want all solutions to be interoperable. This connects our systems, which helps us automate mundane tasks and the connection itself creates data hygiene.

Consider authentication, roles and users, backups, document storage, training programs, search, privacy, pushing and getting metadata, pushing and getting documents, updating reporting data across systems. The list can get longer.

Talk with IT and ask where data comes from and where it goes.

Vendor Capabilities and Experience

The best resources to inform vendor capabilities and experience are:

Internal Guideline - Your procurement team may have a guide that will help with requirements.
Past Successful Large System Procurements - Talk with people who were deeply involved in successful, complicated projects. You may learn small things and very big things. Learn the pitfalls and the qualities of vendors that match your teams.
Past RFPs - Read through other RFPs for complicated solutions done to replace functioning systems.

Relevant experience

Legislative experience is important. You want a partner that understands legislation and regulation. If you have a solution partner without a great subject matter expert, then you have to provide the subject matter expertise.

Work with vendors who have referencable customers in this space.

Work with vendors who are skilled at project management and integration. This might mean finding a team of partners.

Team Expertise

Modern software development is critical. This is a modernization project. Work with teams that use modern architectures and open technologies.

If you are going to use LegalDocML, Akoma Ntoso, or LegalRuleML, vendor experience should contain XML engineers.

Methodology

Two meaningful methodologies to utilize are:

Agile - Important agile methods are deploy often, rapid prototyping, user stories, acceptance criteria, daily stand-up, two-week sprints, sprint planning, sprint retrospectives, and constant customer feedback.
CI/CD - Continuous Integration / Continuous Deployment. CI means connect all the systems as you go; this includes integration into final hosting. When this is done, problems shrink instead of grow and development velocity increases. CD means to deploy all the way from development to production on a regular basis. This includes test scripts, merging code, regression testing and user testing. The value is that code it tested in components and there are fewer bugs later.

Training and Support

Start by considering new hire training, not your current staff. If you have a training portal or a handbook, then training for your new system should be designed to fit that format from the very start. This will ensure you get the final product desired and training for current staff will be based on new hire training.

Proposal Submission Requirements and Evaluation Criteria

Most procurement organizations will help with submission requirements and evaluation criteria.

Recommendations:

Clarity - Provide clear and simple instructions. Simplify.
Pricing Structures - If you have defined structures that limit vendor proposals, then be clear about them. If you do not, then allow vendors to propose pricing structures with their reasoning before making a decision. Do not tell vendors how much money you have in the budget until negotiations.
Evaluation Rubrics - They are popular and add value. Always have a clause in the RFP that says you may override the rubric with approval from procurement.
Demonstrations and Proofs of Concept - Require simple demonstrations with finalists. Ask them to show you something with no preparation (if they are a product vendor). Proof of Concepts are good, but they can often be wasted work. Expect to pay vendors for real proofs of concept and carefully consider what is being proved.
MVPs - Require a Minimally Viable Product (MVP) to be produced at a reasonable spot on your timeline. This should be a version of the final solution that can be used end-to-end to do all the jobs of a legislative drafting and publishing solution.

Beyond the RFP - Successful Partnerships

Here are some final recommendations for working with vendors. There are traits of companies that can be discovered during these activities:

Pre-Bid Conferences and Questions - There are restrictions on communication during periods around RFPs/Tenders. So plan communication before or during periods of allowed communication. Better communication and understanding by the vendors, leads to better bids.
Negotiation and Contract Management - Talk money and contracts openly; this is not a topic to be avoided. Vendors are in business and need clarity to make decisions. If you negotiate hard, expect to give in on features. And understand that bids are guesses, so expect things to change and be a partner in finding the right solution.
Relationship Management - Business is people. Talk early and often with partners. If a vendor is difficult to communicate with, beware.
Continuous Improvement - Listen closely to vendors and decide if they care about improving themselves.

This section was purposefully brief. We've taken enough of your time already and we don't feel like this is a major blocker to successful projects (with most organizations). But if you fail to invest in the partnership with your vendors, failure becomes a real risk.

Conclusion: Building the Future of Legislative Work

A well done RFP or tender is the result of great preparation. Xcential hopes this document was helpful in your preparation.

Deciding to modernize software that is used for the most important tasks in a democracy is meaningful. The Rule of Law is the basis of democracy and a legislative drafting and publishing system is responsible for transparency and change to the law.

This is an important system.

Focus on open standards, modern technology, reasonable and clear expectations, and best-in-class partners.

Xcential hopes our blueprint was useful. You are welcome to use any of the content without reference.

Xcential has built systems for the U.S. Congress, State of California, and parliaments around the world. We are an active member of the OASIS working group on Akoma Ntoso and our LegisPro product supports these standards.

We are experts on legislative and regulatory systems. Xcential works with fantastic partners to deliver modern systems around the world.

This post was written by Marty Bickford. Any comments, suggestions, or objections are welcome at marty@xcential.com

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Publication about "Key considerations of Artificial Intelligence in parliaments"

Bússola Tech — Thu, 10 Jul 2025 11:01:51 GMT

Partners

Download the Publication

This report, developed by Bússola Tech in partnership with the Association of Secretaries-General of Parliaments (ASGP), the Organization of American States (OAS), the Society of Clerks-at-the-Table (Socatt) Africa Region, and the Hellenic OCR Team, highlights how Artificial Intelligence (AI) can enhance parliamentary operations.

It explores various AI applications, such as legislative drafting, historical archiving, procedural guidance, and constituent relations, while emphasizing the importance of ethical implementation. The report discusses the significance of user experience, robust data governance, continuous human oversight, and transparent communication to build trust in AI systems.

The document stresses the need for political support, comprehensive staff training, and inter-parliamentary cooperation for AI adoption. It also calls for a concerted institutional effort to maximize the benefits of AI, enabling parliaments to improve their services, accountability, and inclusiveness, ultimately strengthening democratic governance.

Editors: Luís Kimaid from Bússola Tech, Fotios Fitsilis from the Hellenic Parliament and Hellenic OCR Team and João Lima from the Senado Federal of Brazil.

Contributors: Ahto Saks from the Riigikogu of Estonia, Brahma Mishra from MeetMonk, Christopher Henry from the House of Commons of Canada, Claudio Nazareno from the Câmara dos Deputados of Brazil, Esteban Sanchez from the Cámara de Diputadas y Diputados of Chile, Jessica Smith from the U.S. House of Representatives, Jonathan Ruckert from NovaWorks Australia, José Pedro Montero from the Cámara de Senadores of Uruguay and the Association of Secretaries-General of Parliaments (ASGP), Ludovic Delepine from the EU Parliament, Moises Benamor from the Organization of American States, Natalie Foster from the House of Commons of Canada, Orlando Neto from the Senado Federal of Brazil, Sebastian Palacios from Microsoft, Victor Marcel Pinheiro from the Senado Federal of Brazil, Virginia Carmone from the Cámara de Diputadas y Diputados of Chile, Widjai Jaisri from the National Assembly of Suriname.

Download the Publication

Partners:

Evolution in the Usage of AI in Parliaments

Monica Palmirani — Mon, 30 Jun 2025 11:00:42 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction

Artificial intelligence is transforming our societies and everyday life, and its influence is rapidly extending to government institutions, including parliaments. Since most discussions have focused on AI’s role in administrative and judicial decision-making, its use in parliaments remains largely under-analysed, despite its potentially more disruptive impact.

However, the Parliaments are moving some steps forward using proofs-of-concept for experimenting, with new AI solutions, operative tools for main three purposes:

Administrative tasks: to minimise the burden, optimise efficiency, and improve the quality of administrative tasks of the civil servants. Using AI many repetitive jobs could be automatised (e.g., transcript of Assembly), some very time-consuming activities could be delegated (e.g., summarising and ordering the amendments), and some intellectual tasks can pre-process (e.g., consolidation of the bill);
Legislative tasks: to support the decision-maker and the members of parliament's legislative initiative. The AI helps to elicit hidden legal knowledge that the complex legal system doesn’t permit to inspect and manage easily and clearly. We could use AI to retrieve pertinent legal sources (e.g., eDiscovery), simulate the application of the new bill and so to evaluate the possible effects, check the compliance with existing norms, or measure the adherence of the norms with some policies (e.g., Climate Change, gender balance);
Participation tasks: to help the citizens, companies, and institutions to easily access and understand the legislative norms. Using a chatbot or conversational interface portal the end-users can interview the normative system database according to their needs. The query-answering system can support the information retrieval and some legal reasoning portals can return useful information concerning a specific case.

Considering the fundamental role of the Parliaments in the Democratic system, it is crucial to analyse the innovation of AI state of the art through the lens of general principles of law, with a particular emphasis on the theory of law and constitutional theory. If we let technology shape the legal domain without a solid grounding in legal theory, we risk weakening the values that support democracy and the rule of law. By placing AI’s role in parliamentary processes within a constitutional framework, we ensure that technological progress strengthens, rather than undermines, the foundation of our legal systems.

In this scenario, AI systems applied in the parliamentary domain need to follow guidelines for monitoring and governing all the production workflow, from the dataset selection to the evaluation of the results by legal experts. For this reason, the parliaments with a strong digital transformation roadmap are more equipped in terms of infrastructure, competencies, culture, and organisation to face this disruptive momentum of the AI age. Secondly, to make the AI effective, a solid digitalisation of the documentary system should be in a very advanced status, using advanced LegalXML standards and Semantic Web technology for annotating legal knowledge. With a Parliament based only on paper processes is impossible to catch the benefit of the AI revolution.

Consequently, the qualified and annotated legal datasets coming from previous research and technologies applied in the last decades in the Parliaments addressed to the digitalisation of some tasks are now pillars for achieving relevant results for enhancing the law-making process with assistive AI tools. To jump directly to the generative AI tools risks not capitalising on the enormous assets of legal knowledge embedded inside the digital documents. Hence, the work conducted by the AI and law community in the last thirty years is fundamental for preparing the necessary ground for the correct application of AI in Parliament.

After all, we have witnessed an evolution in the digitisation of legal sources, in particular inside of the parliaments or in the legislative drafting offices (e.g., government) in the last ten years oriented to improve not only the publication online of legal sources, but also to redesign the workflow in the lens of the digitisation of the documents flow.

From the digitisation of Parliaments to the AI through Open Data

The process of documentary digitisation was the necessary forerunner of the current hype of AI in Parliaments. It has gone through several phases:

Figure 1 – Evolution in the digitisation of legal sources.

In the 90s’ the Official Journals focused on publishing the legislation online on the Web to implement the principle of accessibility and also the Parliaments followed this trend for enhancing the transparency of the parliamentary activities. Later, open access to law online using the Web was the main goal for sharing legal sources in the ‘Web of Data ‘approach (Filtz 2020, 2021, Livermore2019) creating Open Data collections of documents and legal ontologies. Consequently, we applied the same standard for a deep digital transformation of the law-making process, while enhancing workflows between institutions. The fourth step is to use all the open data and the legal document structure represented in open standards to enable legal data analytics, create new AI applications by using these legal big data, and also to transform portions of procedural rules into smart contracts that are immediately executable and enforceable (see https://dati.senato.it/sito/home that provide RDF format, ontology and AKN bulk in github).

Several official journals, national archives, and parliaments have sought to manage legal sources within legal corpora with the use of technologies like databases, XML, RDF-metadata, and logic formulas. Subsequently, they also set out to provide updated versions of the law at any moment in time (the so-called point-in-time mechanism). Using Formex, an SGML data standard now translated into XML (Formex v4) and integrated with ELI and ECLI, Eur-Lex began to consolidate the database of European Legislation in 1999 and now is moving the whole document collection to Akoma Ntoso using AKN4EU. Today the Publication Office of European Union is experimenting with the LLM for marking up the AKN4EU and for producing the consolidated versions (AI4XML project). The Emilia-Romagna Region (Italy) started to consolidate regulations back in 2003 using the NormeInRete (NIR) XML schema and now it promotes an important project of LLM called SAVIA for supporting the information retrieval of the relevant legislation using a conversational chatbot. The Senate of Italy adopted the Akoma Ntoso standard for bills, transcripts, and other kinds of documents also provided in Open Government Data and now is promoting the use of AI inside of the Parliament for extracting the amendments during the bill changes. On 30 June 2009, the Senate of Brazil launched the parliamentary consolidated database (LexMLBrazil) with a point-in-time function based on a customisation of the XML Akoma Ntoso schema and now some experiments with ChatGPT have been developed. The National Archives of the UK have progressively been transforming all UK legislation into XML, RDF, and Akoma Ntoso since 2012. Recently also the case-law is managed in AKN. The legislation in XML format permits to take advantage during the new bill drafting from large annotated corpora and in this way to favouring the consolidation of the amending draft bill. The AI in general and the LLM can improve this task that is under experimenting in UK. In 2017, the United Nations approved Akoma Ntoso as the official standard for their documentation (AKN4UN), and the EU institutions launched a similar project in 2018 with the AKN4EU initiative. WHO has been transforming all the workflow processes using AKN4UN to detect the “obligation to make periodical reports”. Especially inside of the WHO Assembly to have AKN documents improve interoperability between institutions, member states, and stakeholders and permit to take under hidden valuable knowledge. Moreover, LegalXML modelling provides a robust and solid digital serialisation of legal documents by applying principles of legal theory to annotate legal knowledge (e.g., temporal parameters, semantic web annotation, document structure, normative citations). The Chamber of Deputies of Italy on February 2024 launched a Manifestation of Interest in experimenting with generative AI to support the legislative initiative by the deputies and the operative tasks of the legislative offices. Now the Chamber of Deputies and the Italian Senate have been developing proof-of-concept of the use of LLM for extracting automatically the amendments from the modified bill, for ordering and classifying the amendments.

AI and Legislative Process in Parliaments

AI and Law communities over the last thirty years have developed widely shared theories and models that are capable of managing norms, values, principles, beliefs, interpretation, and argumentation (Sartor, Prakken, Rotolo, Boella, van der Torre). Other scholars use ML/NLP/AI non-symbolic techniques for extracting, classifying, and analysing legal knowledge and legal norms starting from the text (Ashely 2017). Many members of the AI&Law community have developed logic theories and methods for modelling norms in legal formulas and have also developed tools for managing the legal reasoning interaction with legal experts (Governatori, Palmirani, Boella). Some research projects started important investigations in the direction of “Law as Data” to extract data from legal texts and to improve information retrieval based on semantic and legal ontologies (Palmirani 2018, 2018a 2020,2022, Liga 2019, 2019a). The Lynx project aims to translate a legal system into a knowledge graph; ManyLaws mixes different legal metadata to improve searchability. What is new in these research endeavours is the aim to codify normative thought directly using programming languages without passing through any legal language. OpenFisca undertakes to codify significant fragment of legal system with programming; and Marcell, for example, uses AI to improve multilingualism in legal documents. These projects are currently isolated and not well integrated in a unique research vision, and especially they do not fully include the philosophy of law, legal theory analysis, or constitutional law, and do not create a robust legal framework for digital a legal system that is dynamic and diachronic in multilingual perspectives with multiple interpretations and meanings. The ERC project CompuLaw is one of the more advanced projects that take an interdisciplinary approach including legal-techno-social aspects. However, it is more focused on logic-symbolic and non-symbolic modelling of norms integrated with AI techniques like predictive law and eJustice. The ERC project HyperModeLex is focused on the legislation domain. The communication of the AI results is a fundamental task in light of the transparency and explicability principles contained in the Artificial Intelligence Act (AIA) (see Arts. 13 and 14 – Human oversight). It is important to mitigate the ‘black box’ effect (Pasquale 2015, Sovrano 2021, 2024). The Legal Design approach and methodology (Hagan 2020), based on graphic arts and human-computer interaction pillars, are oriented toward favouring the communication of legal knowledge using visualisation and dynamic interfaces. This visual approach permits a better understanding of the AI results in the light of human oversight and human-in-control principles and guarantees the autonomy of the decision-maker.

‘Rules as Code’, ‘Law as Code’,‘Digital-ready policy’

We need to go beyond the state of the art of the current theory of law in parliaments and take the challenge of ‘coding of the law’ by building a solid IT-based legal framework supported by the philosophy of law, theory of law, ethics, constitutional law, parliamentary regulation (Micklitz) and a clear roadmap (Von Lucke 2023). However, there are at least three different methodologies nowadays:

‘Rules as Code’: the legislative process is modelled with flow-chart analysis and coding for a better understanding of the implications and to produce ‘computable law’ easy to integrate into the information systems (e.g., eGov services);
‘Law as Code’: the legislative process starts with drafting legal text for later converting the norms into machine-computable format (e.g., Akoma Ntoso, USML, etc.);
‘Digital-ready policymaking’: the legislative process is modelled for checking compliance against some policies and forecasting the effects in society, it is designed using human-computer interaction principles and a legal design approach simplifying the legal language.

The Parliaments should decide which approach to adopt according to the type of law to model. Each approach has pros and cons. ‘Rule as Code’ is very computable in automatised software but it is also less flexible to the changes of the law; ‘Law as Code’ is flexible and oriented to model norms and principles for admitting multiple interpretations; ‘Digital-ready’ is a managerial and service-oriented approach good for technical norms (e.g., bank law, school law).

Hybrid AI approach in Parliaments

The above-mentioned approaches could take great advantages and rush the use of AI, especially non-symbolic AI (Machine Learning / Deep Learning/ Large Language Model). AI is rapidly evolving, and it is becoming evident that a hybrid technical framework combining machine learning, deep learning based on stochastic technologies, LLM, with semantic knowledge modelling, legal reasoning, and a symbolic rule-based approach (Palmirani 2020, Ashley 2020, Verhij) could produce better results in the legal domain. A retrieval augmented generation (RAG) definitely supports the updating and training task of LLM with the most relevant documents in the legislative system. However, the main problem with current applications of non-symbolic AI (ML/DL/LLM) in the legal domain is the lack of contextual information. This affects the ability to create useful relationships between different annotations, classifications, clustering, correlations, and regressions.

In more detail, the main problems in the current state of the art in ML/DL applications for legal documents include the following:

ML/DL/GenAI works without semantics, and much contextual information included in the legislative document is neglected, with an evident lower capacity for interpretation of similar concepts (e.g., palliative care, hospice care).
Legal citations are a consolidated best practice in legal disciplines, which entrust some important meta-role to external textual resources (e.g., definitions, derogations, modifications, integration of prescriptiveness, penalties, and conditions). This means that ML/DL/LLM should also consider the cited text, especially considering that some algorithms (e.g., similitude, grouping) can find similarities in texts (e.g., “art. 3” and “art. 13”) when the content is completely different. For this reason, the network of norms through citations should be included in the baseline of the experiments.
Temporal parameters are fundamental to creating a robust ML/DL dataset. The repealed acts should have less importance in the probabilistic model rather than the updated codifications. So, frequency, probabilistic calculus, and temporal series should be mitigated with criteria of relevance and legal validity (e.g., enter into force, enter into operational date).
Logic and semantic web annotation should also be integrated with ML/DL in order to understand the type and meaning of relationships that connect different sentences in the text (e.g., obligation and penalty, obligation, and derogation).
Context in legislation domain is fundamental. We find in the legislative provisions’ frequent conditions (e.g., jurisdiction exclusion like in the United Kingdoms) or logic operators (e.g., Boolean connectors in and, or, xor in the legislative definitions) that are neglected by LLM.
Legal language is different from the common language and the LLM/GENAI technique depends by the tokenization process that is different language by language (e.g., in German the word Rindfleischetikettierungsüberwachungsaufgabe is a composition of several words). The legislature language changes from one legislature to another legislature, including political narrative tone. For this reason, the LLM could not have the same accuracy from one legislation to another.
Implicit legal rules are not embedded in the legal language but in the institutions rules, in the competencies power written in the Constitution Law or in the Assembly regulation, in the hierarchy of legal sources coming from the theory of law. These rules should be added to the ML/DL/LLM using rule-based approach.

For these reasons, a hybrid architecture, one that includes symbolic and non-symbolic AI, is strongly advocated in integrating ML/DL legal knowledge with Semantic Web annotation and legal deontic logic modelling (Deakin 2020). Akoma Ntoso as a common interchange LegalXML standard could be a good bridge for creating a common annotated digital corpus for robust AI applications.

Figure 2 – AI and the legal domain.

Conclusions

AI is growing rapidly in many human activities and also inside of the Parliaments for improving administrative, legislative, and communication tasks. The Parliaments need to have a roadmap for deciding which methodology to adopt (e.g., ‘Digital-Ready policies) according to the legal tradition and the type of scenario. Additionally, they need to have clean technical frameworks (eg., Hybrid AI) for mitigating several risks (e.g., bias, discrimination, poisoning data, manipulation of the AI models, lack of explicability). Finally, guidelines for conducting the different projects should be followed for respecting the ethical, legal, and democratic principles (e.g., autonomy of the legislative decisions, legitimacy of the process, sovereignty of the infrastructure).

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

From Paper to Tokens: Transforming Legislative Services in the Chamber of Deputies of Chile

Luís Kimaid — Wed, 11 Jun 2025 14:02:57 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Acknowledgements

The modernisation of the Chamber of Deputies of Chile through the CAMINAR project required remarkable dedication, coordination, and expertise across multiple areas to bring this innovative vision to fruition. I would like to recognise Mr. Miguel Landeros, Secretary-General of the Chamber of Deputies of Chile, Mr. Luis Rojas, Deputy Secretary-General, Mr. Hernan Figueroa, IT Director, Mr. Esteban Sánchez Rivera, Head of Information Systems Development, Ms. Virginia Carmona, Deputy Director of Parliamentary Assignments, and all their staff for their unwavering leadership, technical expertise, and commitment throughout the development of CAMINAR. Their collective efforts were crucial in transforming the Chamber's legislative and administrative processes.

I. Introduction

The Chamber of Deputies of Chile has undergone a significant transformation in recent years, moving from a largely paper-based legislative process to a fully digital system. This shift was driven by the need to address inefficiencies in document management, information retrieval, and legislative transparency. The latest stage of this transformation came in the form of the CAMINAR project, a comprehensive Artificial Intelligence initiative designed to modernise both legislative and administrative functions within the Chamber of Deputies.

CAMINAR introduced advanced artificial intelligence and digital tools aimed at streamlining key processes. Modules like CAMINAR-L, focused on legislative tasks, and CAMINAR-A, addressing administrative needs, have significantly enhanced the Chamber's ability to manage legal texts, financial oversight, and parliamentary assignments. By transitioning to a paperless workflow and embracing AI-powered solutions, the Chamber has improved efficiency, transparency, and accountability in its operations.

The following sections provide an in-depth exploration of this transformation, detailing the challenges faced by the Chamber, the development and implementation of CAMINAR, and the resulting impact on legislative and administrative functions.

II. Background

The Chamber of Deputies of Chile has encountered a multitude of challenges over the years, driven by both historical events and evolving demands for legislative efficiency and transparency. Historically, the Chamber operated in a predominantly analogue environment where legislative activities were heavily reliant on paper documents, face-to-face interactions, and traditional methods of information management. This setup, while functional in earlier times, became increasingly inefficient and cumbersome in the face of rapid technological advancements and the growing complexity of legislative work.

Throughout much of the 20th century, the Chamber's operations were characterised by manual record-keeping and limited technological integration. This not only slowed down the legislative process and administration but also made it difficult to manage and retrieve information promptly. The need for more efficient processes became particularly apparent during periods of intense political activity and reform, where the volume of legislative work increased significantly.

The advent of the digital age brought about new expectations for efficiency. The public and Members of Congress began to demand greater insight into legislative processes and quicker access to information. Additionally, the complexity of modern governance required more sophisticated tools to handle the vast amounts of data generated by legislative activities.

An important moment highlighting the need for digital transformation was the COVID-19 pandemic. The pandemic disrupted traditional legislative procedures and administration, necessitating a swift adaptation to remote work. This sudden shift underscored the Chamber's technological shortcomings and the urgent need for a more robust digital infrastructure. The inability to conduct business as usual due to lockdowns and social distancing measures forced the Chamber to reconsider its reliance on physical documents and in-person interactions.

The need for a comprehensive digital transformation became undeniable. The Chamber of Deputies responded to this need by investing in a robust digital infrastructure, establishing a paperless workflow, and embracing a digital-first approach to legislative and administrative processes. The institution required a solution that would not only modernise its legislative process and administration but also enhance efficiency. This is where the CAMINAR project, and specifically CAMINAR-L and CAMINAR-A, comes into play.

As Esteban Sánchez Rivera, Head of Information Systems Development at the Chamber of Deputies of Chile, noted, "The AI projects have been the result of the Chamber of Deputies' constant trajectory in the search for implementing innovative, efficient, and useful technological tools that benefit the parliament, the parliamentarians, and the citizens."

CAMINAR-L (Legislative) is designed to address these needs by leveraging advanced digital technologies and artificial intelligence to streamline legislative processes. One of the core components of CAMINAR-L is the use of vectorised legal databases. These databases allow for the precise and efficient management of legal texts, ensuring that data is both accurate and easily retrievable. By vectorising the legal databases, the Chamber can quickly access relevant information, reducing the time and effort required to sift through vast amounts of legislative documents.

Another critical aspect of CAMINAR-L is human validation. While AI and digital tools can significantly enhance efficiency, the accuracy and reliability of legislative work still require human oversight. By integrating human validation into the process, CAMINAR-L ensures that the outputs generated by AI tools are checked for errors and relevance, maintaining the integrity of the legislative process.

Bias regulation is also a fundamental principle guiding CAMINAR-L. In legislative work, the presence of bias can significantly impact the outcomes and fairness of the process. CAMINAR-L incorporates mechanisms to modulate and reduce bias, ensuring that the AI tools used provide balanced and impartial support to Members of Congress.

The concept of normative hierarchy is crucial in the legislative context, where different laws and regulations hold varying degrees of importance. CAMINAR-L prioritises the most significant legal norms in its responses, ensuring that the advice and information provided to Members of Congress reflect the correct legal framework and hierarchies.

By addressing these core needs—efficient information management, reliable human oversight, bias regulation, and respect for normative hierarchies—CAMINAR-L transforms the legislative process within the Chilean Chamber of Deputies. It empowers Members of Congress with tools that enhance their ability to draft, debate, and pass legislation in a manner that is both timely and informed.

Complementing CAMINAR-L, CAMINAR-A (Administrative) is focused on modernising the Chamber’s administrative functions, particularly those related to the management of parliamentary assignments and financial oversight. CAMINAR-A leverages artificial intelligence to automate and enhance various aspects of administrative processes, ensuring that parliamentary resources are allocated and utilised in compliance with the stringent regulations governing public funds.

Through CAMINAR-A, the Chamber has implemented tools that allow for real-time access to financial data, automatic extraction and validation of receipt information, and clear guidance on the complex regulatory frameworks surrounding parliamentary expenses. This ensures that all expenditures are managed efficiently, with a high degree of transparency and accountability. By automating these processes, CAMINAR-A not only reduces the administrative burden on staff but also mitigates the risk of errors, thereby strengthening the Chamber’s overall governance.

CAMINAR is not a static solution but a continuously evolving project aimed at continually adapting to the changing needs of the legislative activities and incorporating new advancements in artificial intelligence and additional legislative databases. Additionally, the Chamber is exploring the creation of a public-facing version of CAMINAR, known as CAMINAR-T, which would provide citizens with access to these powerful tools, further democratising the legislative process.

Together, CAMINAR-L and CAMINAR-A represent a comprehensive approach to modernising the Chamber of Deputies of Chile. While CAMINAR-L focuses on enhancing the legislative process through advanced digital tools, CAMINAR-A ensures that the administrative side of parliamentary work is conducted with the same level of efficiency and transparency.

III. CAMINAR-L

CAMINAR-L1: Transcription Assistant

The CAMINAR-L1 module is a component of Project CAMINAR, designed specifically to automate the transcription of parliamentary debates. This innovation addresses the need for efficient, accurate, and timely documentation of legislative proceedings.

Image 1: CAMINAR-L1

The primary objective of CAMINAR-L1 is to facilitate the automatic transcription of parliamentary debates. The integration of Whisper in CAMINAR-L1 facilitates two distinct and independent transcription processes. The online transcription process is used to insert subtitles into the streaming signals of plenary and commission sessions in real-time, ensuring accessibility during live broadcasts. Meanwhile, the offline transcription process generates literal transcriptions of the debates, which are later used as key inputs for the creation of the Diario de Sesiones. This offline process relies on Whisper's AI capabilities to produce accurate, detailed records of the parliamentary debates.

By leveraging advanced speech recognition technologies, this module captures spoken words in real-time, converting them into text. This automation significantly reduces the dependency on manual transcription, which is not only time-consuming but also prone to human error. Leveraging advanced speech recognition technologies like OpenAI's Whisper model, CAMINAR-L1 provides real-time transcription and AI-powered summaries of parliamentary debates, significantly enhancing the efficiency and accessibility of these records.

CAMINAR-L1 is equipped with robust features that enhance its functionality. The real-time transcription feature ensures that every word spoken in parliamentary sessions is captured instantaneously, allowing for immediate access and review. This capability is crucial for maintaining an accurate and up-to-date record of legislative activities.

Another significant feature of CAMINAR-L1 is its ability to generate summaries of the debates. These summaries are powered by AI algorithms, which are based on the depurated literal transcriptions processed by stenographers. After these transcriptions are cleaned and verified, the language model is used to produce concise summaries of committee sessions. This ensures that the summaries maintain a high level of accuracy and relevance, aiding in the efficient documentation of parliamentary activities. These summaries provide a concise overview of the discussions, making it easier for Members of Congress, researchers, and the public to understand the key points and decisions made during the sessions. This feature not only saves time but also enhances the accessibility of the legislative process.

In terms of usage, CAMINAR-L1 is utilised extensively in both plenary sessions and commissions. Its deployment across these settings ensures that the transcription and summarization processes are consistent and comprehensive. By integrating CAMINAR-L1 into these legislative environments, the Chilean Congress can maintain a high standard of documentation. By streamlining these tasks, CAMINAR-L1 allows legislative staff to focus on more strategic and analytical aspects of their work, thereby improving overall productivity and effectiveness.

CAMINAR-L1 also represents a significant stride in the modernization of the Chilean Congress. The project emerged as an innovative solution to the challenges faced by the legislative body, emphasising the importance of digital tools in maintaining legislative functions efficiently. The collaboration between the Senate, the Chamber of Deputies, and the Library of Congress has been crucial in this modernization process, enabling the integration of comprehensive legal databases essential for the effective functioning of CAMINAR-L1 and other modules within the project.

CAMINAR-L2: Semantic Search Assistant

The CAMINAR-L2 module serves as a sophisticated tool designed to enhance the efficiency and accuracy of legal research within the Chilean Chamber of Deputies. This module's primary purpose is to facilitate semantic searches across extensive legal databases, significantly aiding Members of Congress in their quest to find relevant legal precedents and regulations. The system’s ability to comprehend context and meaning in semantic searches is rooted in the vectorisation of the legal database. By using the XML versions of Chile's legal framework, the vectorisation process was refined to the level of the legislative clause, providing a minimum useful detail for semantic understanding. Unlike traditional keyword-based searches, semantic search technology comprehends the context and meaning behind queries, thereby delivering more pertinent and precise results.

This capability is crucial in the legal domain, where the exact phrasing of legislation and precedents can have profound implications. The features of CAMINAR-L2 include advanced search algorithms that utilise natural language processing (NLP) to interpret and respond to queries in a contextually appropriate manner. To improve searches in the vectorised databases, hybrid processes were implemented, integrating semantic searches with keyword-based searches. This means that the system does not merely look for keywords but understands the context in which the words are used, providing results that are significantly more aligned with the user’s intent. This context-based approach is particularly beneficial in legislative research, where the precise interpretation of laws and regulations is critical. One of the key challenges in implementing semantic search within the legal context was ensuring that the system properly understood the hierarchy of the legal norms. Establishing this hierarchy allowed CAMINAR-L2 to deliver coherent and relevant results, prioritising constitutional laws, statutory laws, and regulatory norms appropriately in response to user queries.

In terms of usage, CAMINAR-L2 is employed extensively in both the plenary sessions and the various commissions within the Congress. Members of Congress often need to reference previous laws, amendments, and related legal texts to draft new legislation or to support their arguments in debates. CAMINAR-L2 facilitates this by providing comprehensive search results that include relevant legal texts, historical legislative documents, and pertinent legal precedents. This not only saves time but also ensures that the legislative work is grounded in robust legal research. By enabling precise and contextually relevant searches, the module significantly enhances the accessibility of the legislative process. Members of Congress can base their decisions and legislative drafts on comprehensive and precise legal information, thereby improving the quality and relevance of the laws being crafted.

CAMINAR-L3: Argumentation Assistant

The CAMINAR-L3 module is a groundbreaking tool designed to enhance the quality and depth of legislative debates by providing argumentation assistance tailored to various political biases. The primary purpose of this module is to support Members of Congress by generating arguments that align with specific political stances, whether conservative, liberal, or somewhere in between. This functionality ensures that discussions and legislative proposals are well-rounded and consider multiple perspectives, thereby enriching the legislative process.

Image 2: CAMINAR-L3

This is achieved through a sophisticated understanding of legislative information, enabling the assistant to produce arguments that resonate with the intended legislative position. The key features of CAMINAR-L3 include its ability to customise arguments based on the user's specified political stance and its capacity to provide thorough, well-researched points that can be used in debates and legislative proposals.

One of the standout features of CAMINAR-L3 is its ability to generate arguments that reflect a specific political perspective. Members of Congress can specify whether they need arguments that lean conservative, liberal, or another stance, and the assistant will provide points that align with that ideology. This customization is crucial in legislative debates, where different political viewpoints need to be represented and considered. The assistant ensures that each argument is not only aligned with the chosen political bias but is also grounded in legal and factual accuracy, enhancing the credibility of the debate.

In practical terms, CAMINAR-L3 is used extensively by Members of Congress and their aides during the preparation and execution of debates and legislative drafting. By providing well-structured arguments that reflect the desired political bias, the assistant aids Members of Congress in formulating their points more effectively. This is particularly useful in high-stakes debates where the quality and persuasiveness of arguments can significantly influence the outcome. To ensure factual and legal accuracy, the system limits the domain of information it accesses to verified legal sources. All AI-generated content undergoes validation by legal experts, whose feedback is systematically incorporated into the system, improving the precision and relevance of future outputs.

The impact of CAMINAR-L3 on the legislative process is substantial. By ensuring that debates are enriched with well-researched and politically aligned arguments, the assistant contributes to more informed and balanced legislative discussions. This not only improves the quality of the debates but also enhances the overall legislative process by ensuring that all viewpoints are adequately considered. Moreover, the ability to generate arguments that are politically aligned helps in fostering a more dynamic and representative legislative environment, where different ideologies can be effectively articulated and debated.

CAMINAR-L4: Legislative Tracking Assistant

CAMINAR-L4 is designed to enhance the legislative operations by providing comprehensive tracking of similar bills within the legislative process. This tool addresses a critical need in legislative operations, ensuring that lawmakers are informed about the status of current bills in relation to existing or past legislation.

The primary purpose of CAMINAR-L4 is to track and compare current bills with those that are either already in the legislative process or have been processed in the past. This feature is essential for promoting legislative efficiency, as it helps Members of Congress identify redundant or conflicting bills early in the legislative process. The system periodically vectorises all bills in progress within Congress. Semantic searches are then conducted using the AI model to compare these vectorised bills, identifying similarities and ensuring that redundant or conflicting legislation is flagged. By doing so, CAMINAR-L4 aids in avoiding unnecessary duplication of efforts and ensures coherence in the legislative agenda.

One of the key features of CAMINAR-L4 is its ability to perform detailed comparisons of current bills with existing or historical legislation. This comparison capability is not limited to the text of the bills but extends to their thematic content, legislative intent, and potential implications. The system leverages advanced algorithms to identify similarities and differences, providing Members of Congress with a clear understanding of how a proposed bill aligns or conflicts with the legislative corpus. The ChatGPT-4o model, along with its default features, was used for thematic content analysis.

This information empowers Members of Congress to make informed decisions about whether to proceed with the bill, amend it to resolve conflicts, or collaborate with colleagues to consolidate similar efforts. The impact of CAMINAR-L4 on the legislative process is profound. By promoting efficiency through the early identification of redundant or conflicting bills, it streamlines legislative workflows and reduces the time and resources spent on duplicative efforts. Additionally, the coherence of the legislative framework is enhanced, as lawmakers can ensure that new bills are harmonised with existing laws. This contributes to a more organised and effective legislative process, ultimately benefiting the governance and legal landscape of Chile.

CAMINAR-L5: Regulatory Impact Assessment Assistant

CAMINAR-L5, the Regulatory Impact Assessment Assistant, is an advanced tool designed to evaluate the regulatory impact of proposed laws. The system uses prompts specifically designed for regulatory impact evaluations, with explicit instructions in the prompts about the type of analysis required. This assistant is important in ensuring that new legislative proposals are thoroughly vetted for their potential effects on the existing legal framework, enhancing the overall quality and viability of legislative outputs.

Image 3: CAMINAR-L5

Image 4: CAMINAR-L5 - User feedback mechanism

This includes a comprehensive examination of how a new bill interacts with the current body of law, identifying any potential conflicts, areas where it may complement existing laws, and gaps that it may fill. By providing a detailed impact assessment, CAMINAR-L5 ensures that new legislative initiatives are well-conceived and harmonious with the existing legal structure.

One of the standout features of CAMINAR-L5 is its ability to identify conflicts, complements, and gaps within the existing legislative framework. When a new bill is proposed, the assistant scrutinises it against the backdrop of existing laws. It highlights any conflicts where the new proposal may contradict or undermine current laws, points out how the proposal may complement or enhance the existing legal structure, and identifies gaps in the current legislation that the new bill could potentially fill. This feature ensures that Members of Congress have a clear and comprehensive understanding of the regulatory landscape they are working within. An essential feature is the determination of the hierarchy within the legal framework, which helps resolve conflicts. The model is instructed to consider the relevance of results according to the legal hierarchy. Members of Congress can use this information to refine and adjust their proposals, ensuring that the final legislative product is robust, coherent, and effective.

The impact of CAMINAR-L5 on the legislative process is substantial. By providing a thorough regulatory impact assessment, it enhances the quality of legislative proposals. Members of Congress are equipped with detailed insights that enable them to craft well-informed and technically sound laws. This not only improves the viability of the proposals but also contributes to a more cohesive and effective legislative framework. The use of CAMINAR-L5 ensures that new laws are not only innovative but also practical and aligned with the existing legal system.

CAMINAR-L6: Constitutional Support Assistant

CAMINAR-L6, the Constitutional Support Assistant, is an essential component of the Project CAMINAR suite. Its primary role is to ensure that legislative proposals comply with constitutional standards, thereby safeguarding the legal integrity of the legislative process.

Image 5: CAMINAR-L6

Image 6: CAMINAR-L6

The primary purpose of CAMINAR-L6 is to check the constitutionality of proposed bills. It meticulously reviews each legislative proposal against the constitution, ensuring that the bill adheres to the constitutional framework of the nation. This preemptive check is crucial in preventing the passage of laws that might later be deemed unconstitutional, which can save time, resources, and avoid potential legal challenges. The system ensures the constitutionality of legislative proposals by instructing the model, through specific prompt commands, to analyse the proposed text against the exclusive chapters of the Chilean Constitution. These chapters specifically define legislative matters and the powers of Congress, ensuring that the proposal is constitutionally compliant.

One of the notable features of CAMINAR-L6 is its ability to highlight the required quorums for passing specific types of legislation and to provide relevant jurisprudence. The assistant identifies whether a simple majority, a qualified majority, or a special quorum is needed for the bill in question, based on its content and implications. Furthermore, it cross-references the bill with existing jurisprudence, drawing on past judicial decisions that may impact the interpretation and implementation of the proposed law. This feature ensures that Members of Congress are fully informed about the constitutional requirements and precedents relevant to their proposals. For cross-referencing proposals with existing jurisprudence, the system uses the XML versions of Chilean legal texts. These versions include unique identifiers for each legal document, allowing for precise and unequivocal references within a vectorised document.

CAMINAR-L6 is extensively used to ensure that legislative proposals comply with constitutional standards. By reducing the risk of passing unconstitutional laws, it helps maintain the legal integrity of the legislative framework. Members of Congress are equipped with the necessary constitutional insights to craft proposals that not only address the needs of the populace but also adhere to the foundational legal principles of the nation. This leads to a more robust and legally compliant legislative process, fostering public trust and confidence in the legislative body.

CAMINAR-L7: Parliamentary Debate History Assistant

CAMINAR-L7, the Parliamentary Debate History Assistant, is a crucial feature within Project CAMINAR, designed to provide historical context for legislative debates. This assistant plays an important role in enriching the legislative process by offering comprehensive access to the historical records of parliamentary debates.

By providing access to a wealth of past debates, it allows Members of Congress to understand the evolution of discussions on specific topics. This historical perspective is invaluable for crafting informed and contextually relevant legislative proposals. The historical data, including all Diarios de Sesiones from 1990 onward, are indexed and made accessible in digital format, specifically structured in XML. The XML versions of the Diarios de Sesiones allowed for the vectorisation of these documents, with the 'Speech' of each speaker being considered the minimum granular unit of information. This structuring and metadata tagging process, which has been carried out by the Office of Drafting for over a decade, ensures that these documents are properly organised for efficient retrieval and analysis.

Members of Congress can trace the development of issues over time, identify key turning points, and understand the rationale behind previous legislative decisions.

One of the standout features of CAMINAR-L7 is its ability to grant access to past debates through a sophisticated keyword search functionality. This feature allows users to quickly locate relevant debates by entering specific keywords or phrases. The assistant then retrieves all pertinent records, making it easy for Members of Congress to find and review discussions on the topics they are currently addressing. This streamlined access to historical data ensures that Members of Congress have the information they need at their fingertips, enabling them to make well-informed decisions. One of the main challenges in maintaining an updated historical record is the continuous process of structuring the Diarios de Sesiones. It is not enough to have digital documents; they must be properly structured with sufficient granularity to enable semantic searches at different levels of detail. This ongoing effort by the Office of Drafting ensures that the records remain usable for advanced AI language models and semantic search functions.

By reviewing past debates, Members of Congress can gain insights into how similar issues were handled in the past, what arguments were made, and what outcomes were achieved. This historical context can help them avoid repeating past mistakes and build on successful strategies. Additionally, understanding the evolution of discussions on specific topics can provide a clearer picture of public sentiment and the political landscape over time.

CAMINAR-L8: Legal Doctrine Assistant

CAMINAR-L8, the Legal Doctrine Assistant, is designed to bolster the legislative process by offering comprehensive access to existing legal doctrines. This assistant significantly enriches the legislative framework by providing Members of Congress with in-depth legal insights, ensuring that legislative proposals are well-founded and robust.

The primary purpose of CAMINAR-L8 is to offer Members of Congress access to existing legal doctrines on specific topics. Searches are primarily tailored to relevant Constitutional Court rulings by adjusting specific prompts and applying filters to the associated metadata. By providing a wealth of legal knowledge, this assistant aids in the development of legislative proposals that are deeply rooted in established legal principles and precedents.

One of the standout features of CAMINAR-L8 is its comprehensive database of legal doctrines, starting with constitutional law. This database is meticulously curated to include a wide array of legal doctrines, ensuring that Members of Congress have access to the most relevant and up-to-date legal information. The focus on constitutional law ensures that legislative proposals are developed with a solid understanding of constitutional principles, helping to safeguard the legal integrity of new laws.

In practical usage, CAMINAR-L8 serves as an indispensable tool for Members of Congress during the drafting of legislative proposals. By consulting the extensive database of legal doctrines, Members of Congress can ensure that their proposals are not only legally sound but also align with established legal principles. This alignment is crucial for the coherence and stability of the legal system, as it helps prevent conflicts and inconsistencies between new and existing laws.

CAMINAR-L9: Management of Budgetary Amendments Assistant

The Management of Budgetary Amendments assistant, designated as CAMINAR-L9, is a crucial module within the CAMINAR-L suite, specifically developed to address the intricate process of handling budgetary amendments within the Chilean Chamber of Deputies. This tool is designed to streamline the amendment process, ensuring that proposed modifications to the national budget are managed efficiently and in compliance with the legal framework governing public finance.

Image 7: CAMINAR-L9

Image 8: CAMINAR-L9

The primary purpose of CAMINAR-L9 is to provide a comprehensive overview and real-time tracking of budgetary amendments. Members of Congress frequently propose changes to the budget during legislative sessions, and these amendments can range from minor reallocations of funds to significant policy shifts. The challenge lies in managing these proposals in a way that ensures transparency, adherence to legal norms, and coherence with the overall budgetary goals. CAMINAR-L9 automates the process of reviewing, categorising, and tracking these amendments, thus reducing the administrative burden on legislative staff and providing Members of Congress with instant feedback on the feasibility and implications of their proposals.

One of the standout features of CAMINAR-L9 is its ability to integrate proposed amendments directly into the existing budget proposal framework, highlighting areas of conflict or redundancy. This functionality helps Members of Congress and their aides quickly identify where amendments may overlap with other proposals or existing budgetary allocations, ensuring that resources are not double-counted or misallocated. CAMINAR-L9 also provides real-time feedback on the impact of amendments, alerting users to potential violations of constitutional or regulatory norms governing budgetary practices.

In practical usage, CAMINAR-L9 has reduced the time required to process amendments and has improved the accuracy of budgetary decisions.

IV. CAMINAR-A

CAMINAR-A1: The Parliamentary Assignments Regulation Assistant

The Parliamentary Assignments Regulation Assistant is a module within the CAMINAR-A system. It plays an important role in the administration of parliamentary expenditures within the Chamber of Deputies of Chile. This advanced tool addresses the complex regulatory framework governing how parliamentary resources are allocated and spent, ensuring that expenditures adhere to the stringent rules set forth by the institution.

The Chamber of Deputies of Chile, like many legislative bodies, must navigate a labyrinth of rules and regulations that dictate how public funds can be utilised to support parliamentary functions. These expenditures, which range from salaries for support staff to operational costs such as office rentals and travel, are subject to strict oversight to maintain public trust and ensure that resources are used efficiently and appropriately. The system limits its search domain to the specific regulations governing parliamentary expenses, ensuring responses are based solely on concrete data. Structured versions of documents and regulations were created to represent the hierarchy and relationships that were not adequately captured in the original documents. The Parliamentary Assignments Regulation Assistant was developed to assist the Chamber in this challenging task by providing accurate, timely, and contextually relevant guidance on the permissibility of various expenses.

This assistant operates through the integration of advanced artificial intelligence technologies, including natural language processing (NLP) from ChatGPT4o. These technologies allow the assistant to comprehend and respond to queries posed in natural language, reflecting the diverse ways in which users may frame their questions. The system primarily uses ChatGPT to ensure accurate and contextually relevant responses. When a member of the Chamber of Deputies or their staff submits a query regarding the appropriateness of a particular expense, the assistant quickly references the relevant sections of the Chamber's regulations and official resolutions to provide a definitive answer.

The traditional process of determining whether an expense complies with the Chamber's regulations often required extensive manual effort, involving the consultation of numerous legal texts and, at times, multiple levels of bureaucratic approval. This not only slowed down the decision-making process but also introduced the potential for errors and inconsistencies in the application of the rules. The Parliamentary Assignments Regulation Assistant effectively mitigates these issues by automating the interpretation of the Chamber’s regulations, ensuring that responses are both swift and accurate.

The introduction of this module is part of a broader effort within the Chilean Chamber of Deputies to leverage digital tools to enhance its governance. The assistant is not merely a reactive tool but also serves a proactive role in promoting adherence to the institution’s regulations. By providing instant clarification on what constitutes an allowable expense, it helps to prevent unintentional violations of the rules, thereby safeguarding the Chamber against potential misuse of funds and enhancing its overall operational integrity.

The Parliamentary Assignments Regulation Assistant also reflects the Chamber of Deputies' commitment to transparency. By ensuring that all expenditures are rigorously checked against established regulations, the assistant contributes to the institution’s broader goals of accountability and public trust. This transparency is further supported by the assistant’s integration with other modules within the CAMINAR-A system, creating a comprehensive digital ecosystem that supports the Chamber’s administrative and legislative functions.

CAMINAR - A2: The Online Data Assistant for Parliamentary Assignments

The Online Data Assistant for Parliamentary Assignments, a module within CAMINAR-A, is essential to the real-time management of parliamentary expenditures in the Chilean Chamber of Deputies. This tool provides instant access to specific data regarding parliamentary spending, facilitating a more dynamic and accurate approach to overseeing financial resources. The system manages real-time access to parliamentary expenditure data by using a natural language model that classifies the type of question asked. Based on the classification, it internally calls different WebAPIs to retrieve the specific data in real-time.

The assistant is designed to respond to queries about the current status of various parliamentary expenses, such as fuel consumption or office rentals. When a Member of Congress or staff member inputs a request for information, the assistant immediately retrieves the relevant data from the Chamber’s database. This allows users to obtain precise figures and detailed reports on their expenditures, without the need for manual data retrieval.

The assistant's capacity to deliver real-time data has transformed the way financial information is accessed and utilised within the Chamber. By automating the process of data retrieval, it ensures that users have the most current information available, enabling them to make informed decisions based on up-to-date figures. This has reduced the time required to gather and analyse financial data, making the management of parliamentary assignments more efficient. The assistant integrates with the same methods used by CAMINAR A1, ensuring that the data provided is both accurate and up-to-date. The main challenge in maintaining accuracy is ensuring that the language model correctly interprets and classifies the questions, aligning them with predefined methods that can return the appropriate response.

Furthermore, the Online Data Assistant is capable of handling a range of specific inquiries, offering detailed insights into particular aspects of parliamentary spending. For example, it can provide exact amounts related to specific expenses or offer a summary of financial activities over a given period. This functionality is crucial for maintaining accurate records and ensuring that parliamentary resources are being utilised effectively.

The assistant is also a part of a broader initiative within the Chamber to improve the transparency and accountability of parliamentary operations. By making financial data readily accessible, it contributes to a more open and accountable system of governance, where expenditures can be monitored and reviewed with greater ease.

CAMINAR - A3: The Receipt Information Extraction Assistant

The Receipt Information Extraction Assistant, another critical module within the CAMINAR-A system, is designed to streamline the process of managing and verifying parliamentary expenditures in the Chilean Chamber of Deputies. This assistant focuses specifically on the extraction and validation of information from receipts and other financial documents submitted by parliamentarians as part of their expense reporting.

In the implementation of CAMINAR-A3, various models such as Anthropic's Claude 3.5, OpenAI's GPT-4, and Google's Gemini have been tested for their ability to extract and validate information from the diverse formats of receipts.

The assistant automates the extraction of essential details from various types of receipts, such as the document type, date, amount, and other relevant financial data. By doing so, it alleviates the need for manual entry and reduces the potential for human error in the processing of these documents. When a receipt is submitted, the assistant analyses the document, identifies the key information required for the expense report, and accurately extracts this data.

Currently, the system is undergoing validation, where comprehensive tests are being conducted to assess the accuracy of these models, taking into account the specific types of documents and unique characteristics of the receipts presented by Members of Parliament.

One of the distinguishing features of the Receipt Information Extraction Assistant is its ability to handle different formats of financial documents. Whether the receipt is a standard invoice or a more complex document, the assistant is trained to recognise and process the necessary information efficiently. This capability ensures that even when documents vary in their layout or presentation, the assistant can reliably extract the required data.

The extracted information is then validated against the Chamber’s existing financial records to ensure compliance with the established regulations. This validation process is crucial in maintaining the integrity of parliamentary expenditures, as it helps to identify discrepancies or errors that may arise during the submission of expense reports. By automatically cross-referencing the extracted data with the Chamber’s records, the assistant plays a key role in ensuring that all reported expenses are accurate and justified.

In addition to its extraction and validation functions, the assistant also supports the broader goal of enhancing the transparency and accountability of parliamentary spending. By providing a systematic and automated approach to managing receipts, the assistant contributes to a more efficient and transparent financial reporting process within the Chamber.

V. Principles Guiding CAMINAR-L

The CAMINAR-L project is driven by key principles that underpin its functionality and ensure its effectiveness in supporting the legislative processes of the Chilean Congress. These principles are crucial for maintaining the integrity, accuracy, and fairness of the legislative assistants, thereby enhancing the overall quality of legislative work.

The foundation of CAMINAR-L lies in its use of vectorised legal databases. This approach transforms complex legal texts into a structured format that allows for efficient processing and retrieval by AI algorithms. The transformation ensures that the system works with comprehensive and contextually relevant data, capturing the full scope of Chile’s legal framework, from the Constitution to specific regulatory norms. This meticulous structuring is vital for providing precise and tailored information to Members of Congress, facilitating well-informed decision-making and enhancing the legislative process.

Human validation is another cornerstone of CAMINAR-L. Despite the sophistication of AI, the role of human oversight cannot be overstated. Human validators play a critical role in reviewing and verifying the information generated by the AI, ensuring that the outputs are accurate and contextually appropriate. This process not only reduces the risk of errors but also ensures that the recommendations and insights provided to Members of Congress are aligned with the legislative context. The blend of AI efficiency with human expertise helps maintain high standards of reliability and trust in the system.

The regulation of bias within CAMINAR-L is a fundamental principle aimed at ensuring fairness and balance in the system’s outputs. Some applications within CAMINAR-L are designed to operate without bias, while others incorporate a controlled level of bias to reflect the specific needs and perspectives of users. This careful modulation of bias prevents the reinforcement of existing prejudices and promotes a more equitable legislative process. By managing bias effectively, CAMINAR-L supports objective and impartial decision-making, ensuring that the legislative recommendations provided are fair and balanced.

The principle of normative hierarchy is essential for prioritising the most important legal norms in the responses generated by CAMINAR-L. The system is designed to consult the highest-ranking legal norms first, beginning with the Constitution, followed by statutory laws, regulations, and other relevant legal texts. This approach ensures that the legislative advice provided is grounded in the most authoritative and relevant legal sources, supporting the creation of coherent and legally sound legislative proposals.

Complementing these principles is a dedicated focus on refining the prompts used to interact with the AI assistants. This role, often referred to as the 'prompt engineer' or 'prompt manager,' involves continuous analysis and improvement of the questions posed to the AI, ensuring that the system provides the most accurate and relevant information possible.

These guiding principles of vectorized legal databases, human validation, bias regulation, and normative hierarchy are integral to the functioning of CAMINAR-L. They ensure that the system provides accurate, reliable, and fair support to the legislative processes of the Chilean Congress, thereby enhancing the quality and efficiency of legislative work. Through these principles, CAMINAR-L embodies a commitment to innovation, integrity, and excellence in legislative assistance.

VI. The Impact and Transformation of CAMINAR

The CAMINAR project represents a watershed in the evolution of legislative and administrative functions within the Chamber of Deputies of Chile. This initiative has fundamentally redefined how the Chamber operates by integrating advanced AI-driven tools and digital technologies into its core processes. CAMINAR's implementation signals a decisive shift towards a more dynamic, data-driven legislative operations where decisions are informed by real-time insights, and administrative tasks are streamlined through automation.

The CAMINAR-L system offers significant benefits to the legislative process within the Chilean Chamber of Deputies, primarily through enhancing efficiency and demonstrating a commitment to modernisation and technological innovation. Each module within CAMINAR-L plays a crucial role in achieving these objectives, contributing to a more streamlined, accurate, and responsive legislative environment.

These principles ensure that the system provides accurate, reliable, and fair support to Members of Congress, enhancing the overall quality and efficiency of legislative work. By combining AI efficiency with human oversight, CAMINAR-L maintains high standards of reliability and trust, ensuring that legislative processes are both innovative and grounded in robust legal principles.

Innovation is another critical benefit of CAMINAR-L, demonstrating the Chamber of Deputies' commitment to leveraging modern technology within the traditional legislative process. The integration of advanced digital technologies and artificial intelligence in CAMINAR-L showcases the Chamber's forward-thinking approach to legislative work. By adopting these cutting-edge tools, the Chamber not only modernises its processes but also sets a precedent for other Parliaments.

Complementing these advancements in legislative processes, CAMINAR-A extends the same commitment to modernisation and efficiency to the administrative functions of the Chilean Chamber of Deputies. CAMINAR-A enhances the management of parliamentary assignments and expenditures through the use of artificial intelligence, ensuring that these processes are carried out with precision, transparency, and accountability. By automating key administrative tasks, CAMINAR-A not only reduces the potential for human error but also improves the speed and accuracy with which financial data is processed and validated.

The integration of CAMINAR-A into the Chamber’s operations underscores the institution's dedication to comprehensive digital transformation. Just as CAMINAR-L sets new standards for legislative efficiency, CAMINAR-A represents a significant leap forward in the administrative domain, ensuring that the Chamber’s financial management is both modern and robust. This holistic approach to digital innovation positions the Chilean Chamber of Deputies as a leader in the adoption of technology across all facets of parliamentary work.

The innovative nature of CAMINAR lies not just in its current capabilities but also in its adaptability and potential for future growth. The Chamber's commitment to continually refining and expanding the system ensures that it will remain a cutting-edge tool for legislative support for years to come.

Together, CAMINAR-L and CAMINAR-A demonstrate the Chamber of Deputies' comprehensive strategy to enhance both legislative and administrative operations through the adoption of advanced technologies. This dual approach not only improves the efficiency and reliability of parliamentary functions but also reinforces the Chamber’s commitment to transparency, accountability, and innovation in public governance.

VII. System Architecture and Design

The architecture and design of the CAMINAR system, encompassing both CAMINAR-L (Legislative) and CAMINAR-A (Administrative), represent a sophisticated technological framework, designed to support the complex operational needs of the Chamber of Deputies.

CAMINAR is built upon a tiered architecture that consists of distinct layers, each responsible for a specific set of functions within the broader system:

Application Layer: This layer comprises the servers hosting the various tools and interfaces used by the Chamber. This includes internally hosted servers using technologies such as Python and Microsoft Semantic Kernel to support the operation of various tools and applications.
Data Layer: At this level, the Oracle database servers, which are internally housed, manage the vast array of legislative and administrative data.
AI Service Layer: CAMINAR leverages a combination of cloud-based AI services hosted in the cloud, such as the Pinecone vector database and OpenAI language models, while others, including Whisper for transcription, are hosted on dedicated internal servers.

Central to CAMINAR’s operation is its robust data infrastructure, which forms the second layer of the system. Data management within CAMINAR is underpinned by Oracle database servers, housed internally to ensure the integrity and security of the Chamber’s legislative and administrative records. The databases utilised by CAMINAR are not merely repositories of information but have undergone extensive validation, cleansing, and vectorisation processes. This pre-processing of data ensures that the AI models deployed within CAMINAR can retrieve, analyse, and present relevant information with high precision.

The third layer of CAMINAR’s architecture is dedicated to AI services. This layer is distinctive in that it incorporates both cloud-based and locally hosted AI models. For the generative AI models that underpin much of CAMINAR’s functionality, such as real-time transcription and natural language processing, the system utilises ChatGPT Enterprise. For general tasks, OpenAI's models have been primarily used, including GPT-3.5 turbo, GPT-4, GPT-4o, and Whisper large-v3. These models were chosen after rigorous testing due to their superior performance, speed, and response quality in solving real-world problems. Additionally, Anthropic's Claude 3.5 has been utilised for certain specific tasks.

However, not all AI processes within CAMINAR are cloud-dependent. In cases where data sensitivity or regulatory compliance dictates the need for local processing, dedicated internal servers host the necessary AI services. These internal servers ensure that critical data never leaves the Chamber’s control, maintaining compliance with Chilean legal frameworks governing data privacy and security. The strategic decision to use both cloud and internal servers reflects a balance between leveraging the expansive computational resources of cloud-based AI and the need for stringent data governance in sensitive parliamentary environments.

The multi-layered structure of CAMINAR, combining AI services, data management, and application interfaces, allows the Chamber of Deputies to manage vast quantities of legislative and administrative data with precision and agility.

VIII. Development Cycle

The development cycle of the CAMINAR project was marked by a robust focus on research, innovation, and iterative refinement. The team adopted a methodical approach that emphasised experimentation and exploration with various AI models, tools, and services, all while addressing the specific challenges faced by the Chamber of Deputies. This research-intensive process allowed for a comprehensive evaluation of the available technologies, with each phase designed to test, validate, and refine the AI functionalities before their integration into the existing parliamentary systems.

The development process began with an extensive research and experimentation phase, wherein the technical team conducted a series of Proof of Concept (PoC) trials. These trials were essential in identifying viable solutions while discarding options that did not meet the project’s requirements. The PoC phase was intensive, requiring rigorous testing of each model and tool, with end users actively involved through the use of demonstration versions (demos). These demos were invaluable in ensuring that the outputs generated by the AI models were accurate, reliable, and aligned with the Chamber’s operational needs.

User participation, particularly from the Chamber’s staff, was essential during this phase. The active engagement of final users in the demo trials ensured that their feedback directly informed the refinement of the AI functionalities. This early-stage validation process provided crucial insights, allowing for the adjustments and fine-tuning of the AI outputs. Once the demos were successfully validated, the functionalities were integrated into the Chamber's existing systems using a classic software development process. This method ensured a smooth transition from experimental phases to full-scale deployment within the legislative and administrative frameworks of the Chamber.

An iterative approach defined the broader development cycle, incorporating continuous loops of research, development, testing, and validation. These short iterations allowed the team to rapidly discard unviable options while identifying the potential strengths of the tested solutions. Each iteration was geared towards incremental improvements, with constant feedback loops from the demos, ensuring that the AI functionalities were optimised for the specific needs of the Chamber.

Collaboration between the technical team and legal experts, Members of Congress, and other key stakeholders within the Chamber was a cornerstone of the development process. It was imperative that the AI tools being developed accurately reflected the legal and procedural frameworks of the Chamber. To this end, the technical team worked closely with domain experts from the outset, ensuring that the AI functionalities were properly aligned with the unique requirements of legislative work.

Early involvement of key figures, such as the Secretary General and legislative lawyers, was instrumental in validating the PoC trials and demos. These experts played a crucial role in the selection and refinement of the data sets that powered the AI models. Their contributions were essential in adjusting the prompts used by the AI and ensuring that the results generated during each stage of development were legally sound and contextually appropriate. This collaboration extended through all phases of the development cycle, ensuring that the outputs of the AI systems were continuously validated against the Chamber’s legal and operational standards.

IX. Data Management

The data management process within CAMINAR, particularly for CAMINAR-L (Legislative) and CAMINAR-A (Administrative), revolves around the efficient handling and vectorisation of legal databases. A key factor that facilitated the integration of AI into the system was the pre-existing availability of Chile’s complete legal database within the Chamber of Deputies. This database, which includes enacted laws, regulations, legal decrees, and bills, was stored in XML format prior to the development of CAMINAR.

The fact that all legislative documentation was already in XML format proved crucial to the process of vectorisation. This structured format provided the necessary granularity for the AI models employed in CAMINAR to interpret the legal texts with a high degree of accuracy. The vectorisation process involved converting these detailed XML documents into a format that could be efficiently processed by the language models, ensuring that they could retrieve, analyse, and respond to complex legal queries in a precise manner.

X. Navigating the Political Landscape: Securing Buy-In for CAMINAR

The development of CAMINAR, with its innovative integration of AI into the core functions of the Chilean Chamber of Deputies, represents a significant technological achievement. However, the true measure of its success lies not just in its technical prowess, but also in its ability to seamlessly integrate into the political reality of the legislative process. This necessitates navigating the complex landscape of political will, ensuring buy-in from key stakeholders, and ultimately, achieving institutional acceptance of these transformative tools.

According to the Secretary-General of the Chamber of Deputies of Chile, Mr. Miguel Landeros, "The technical work is the fastest. The political decision is the most difficult, although in our case, both processes were carried out expeditiously."

This highlights a fundamental truth: technological innovation, while crucial, must be accompanied by the political will to embrace change and adapt existing structures. In the context of CAMINAR, securing this political buy-in was essential for different reasons.

Firstly, the implementation of CAMINAR necessitates modifications to the Chamber's internal regulations and operating procedures. Addressing these concerns requires open dialogue, transparency about the intended benefits of CAMINAR, and a willingness to address concerns raised by Members of Congress.

Secondly, the adoption of AI-powered tools in a traditionally human-driven domain like Parliaments can raise questions about the potential displacement of human expertise and the impact on the roles of legislative staff. It is crucial to communicate that CAMINAR is designed to augment, not replace, human judgement. By framing CAMINAR as a tool that empowers rather than displaces, the Chamber can foster a more receptive environment for its integration.

Thirdly, the success of CAMINAR relies heavily on the active participation and engagement of Members of Congress. This necessitates addressing potential apprehensions about the use of AI, ensuring that the tools are user-friendly and intuitive, and providing adequate training and support to familiarise Members with the functionalities of CAMINAR. Building trust in the system's reliability, accuracy, and ability to enhance their legislative work is paramount.

Securing political will also involves addressing potential concerns about transparency and accountability. These concerns are not hypothetical, as demonstrated by questions raised about the potential for CAMINAR to reveal sensitive information, such as the salaries of public officials, requiring a careful balance between transparency goals and data privacy considerations. Demonstrating how CAMINAR strengthens these principles, for example, by providing clear audit trails of its recommendations and ensuring human oversight at critical stages, can help alleviate any reservations.

Ultimately, achieving institutional acceptance of CAMINAR required a multi-pronged approach that combined clear communication, collaborative implementation, and demonstration of value. Transparent and consistent communication with Members of Congress about the goals, benefits, and potential challenges of CAMINAR was essential. Involving Members and staff in the implementation process, soliciting their feedback, and incorporating their suggestions can foster a sense of ownership and reduce resistance to change. Providing concrete examples of how CAMINAR enhances legislative efficiency, improves the quality of legislative proposals, and strengthens transparency can build confidence and encourage adoption.

The Chilean Chamber of Deputies' success in navigating these political considerations will ultimately determine the long-term impact of CAMINAR. By securing buy-in from key stakeholders and fostering a culture of openness to technological innovation, the Chamber can solidify CAMINAR's position as a transformative force within the legislative process.

XI. User Feedback and Human Validation

The implementation of CAMINAR was driven by the continuous feedback and involvement of users. This process was marked by distinct challenges, particularly in the context of research, evaluation, and human validation. The process of identifying the best available options to address each specific problem required continuous investigation and iteration. As various cloud-based models underwent frequent updates within short periods, the team found it necessary to constantly re-evaluate and test the effectiveness of these models for each use case. This dynamic environment of evolving AI technologies demanded a sustained commitment to research and refinement, ensuring that the tools remained adaptable and effective.

One of the principal challenges encountered during the implementation phase was the engagement of everyday users to validate the tools and provide constructive feedback. Ensuring that non-expert users could interact meaningfully with the AI-driven assistants, and then contribute valuable insights to refine and enhance the solutions, proved to be a complex task. The process of gathering feedback from this user base, while essential to improving the system, involved overcoming initial hesitations and facilitating a thorough understanding of how the tools could be most effectively utilised. Nonetheless, this engagement was crucial for ensuring that CAMINAR remained user-centric and responsive to real-world legislative needs.

Future improvements to CAMINAR services are closely informed by this ongoing feedback process. Looking forward, several enhancements are planned. One of the key areas under development is the automatic recognition and identification of Members of Parliament in video recordings. This feature is intended to complement the existing automated transcription of plenary and committee debates by adding the names of speakers to the transcripts without manual intervention. The integration of this functionality would streamline the documentation of legislative proceedings, ensuring that each contribution is accurately attributed in real-time.

Additionally, there are concerted efforts to expand the range of outputs generated by AI tools, particularly in the production of simplified summaries of legislative texts. These summaries are being designed with the general public in mind, using accessible language that makes complex legislative information easier for non-expert citizens to understand. By improving the accessibility of legislative documents, the Chamber aims to enhance public engagement with the legislative process, reinforcing transparency and accountability.

In parallel, ongoing tests are being conducted on local servers that have been specifically designed for the implementation of models that do not depend on external services. This shift to locally hosted models is intended to address certain specific challenges, providing a more secure and adaptable solution that can be tailored to the unique requirements of the Chamber. This approach also reflects a broader strategy to ensure that sensitive data remains under the Chamber's control, in line with the stringent regulations governing parliamentary operations.

A key aspect of the CAMINAR system is the robust process of human validation that underpins each AI-generated output. Every interaction with the AI assistants, whether in the form of a query or a response, is subject to mandatory human oversight. This validation process is systematically recorded, with each interaction logged and the outcome of the human review meticulously documented. The system ensures that every piece of information generated by the AI is verified by a human expert before being finalised. Moreover, the identity of the individual responsible for the validation is recorded, ensuring accountability and providing an additional layer of reliability. This structured validation process guarantees that while AI serves to enhance the speed and efficiency of the legislative process, it does not replace the critical judgement and oversight provided by human expertise.

XII. Conclusion

The transformation of the Chamber of Deputies of Chile through the CAMINAR project highlights the profound impact that artificial intelligence can have on legislative and administrative processes. More than just a shift from paper to digital, CAMINAR represents a rethinking of how parliaments can function in the future. By leveraging AI for tasks like legal research, transcription, and financial oversight, the Chamber has demonstrated that efficiency, transparency, and accountability are not only achievable but can be significantly enhanced through strategic technological investment.

However, the success of CAMINAR lies not only in its technological capabilities but also in the Chamber’s ability to navigate institutional and political complexities. The project's effectiveness was contingent on securing buy-in from key stakeholders, managing concerns about AI’s role in legislative work, and maintaining the balance between automation and human oversight. This speaks to a broader lesson: the success of technological innovation in parliaments is as much about leadership and institutional will as it is about the tools themselves.

The Chamber’s experience with CAMINAR highlights that digital transformation is not a one-time shift but an ongoing process. The system’s adaptability to evolving legislative needs and emerging AI technologies will determine its long-term value.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

The Existing and Potential Uses of Natural Language Processing in Parliaments

Grant Vergottini — Mon, 09 Jun 2025 11:01:52 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction

The use of natural language processing in parliaments is often cited as one of the most promising applications of artificial intelligence related technologies. However, it is something that has been in use for many years. Legislation lends itself to natural language processing. In fact, it requires it. There are many good reasons for this. Amongst these are:

First, the corpus of legislation is very constrained, well written, and well structured.
Second, the vocabulary of parliamentary procedures is generally quite small, well controlled, and consistently applied.
Third, and perhaps most importantly, the parliamentary “machine” for legislating is well defined and tends to enforce a level of rigour that lends itself to good computability.

At the same time, there are obstacles to the adoption of some advanced technologies that must be addressed.

First, while parliamentary procedures are very well defined, they have been developed over many centuries. Today, many of these procedures are bound in place by legislative traditions ill-suited to modern automation. Changing deeply rooted traditions is always difficult, if not impossible. Instead, modern automation must adapt to the parliamentary procedures that exist, often limiting what can be achieved.
Second, while the language of legislation is well controlled, it has not been static. With time, it has evolved, leaving behind a corpus of documents defined by vocabularies that, while consistent at any point in time, are not consistent across time. Coupled with the difficulty that comes with documents bound as law and thus requiring legislative action to change, the tools must again adapt to the existing corpus rather than expecting the corpus to be revised.
Third, parliaments are generally risk adverse. With history going back much further than anyone’s memory, there is a concern that there are just too many unknowns in the procedures and documents a parliament manages to risk ever making any drastic changes – so change is always very measured.
Fourth, there is a high bar for accuracy. The language of legislation, as enacted, becomes the law of the land. Correcting errors usually requires further legislative action and is not well tolerated. While natural language processing has made leaps and bounds over the past few years, working with the precise standards of parliamentary procedures is a particularly difficult challenge.

Existing Uses of Natural Language Processing

Natural Language Processing has been used for several decades now to modernise how parliaments work. While some of these technologies are somewhat crude when compared to modern technologies, they have had to evolve around the challenges parliaments impose and have developed their own unique strengths that many of today’s general-purpose technologies do not have. Let us look at some of these existing applications of natural language processing.

Converting Existing Text

Legislative texts can date back centuries. Before any modernisation project can be undertaken, having a strategy to bring these texts into modern times is essential. These documents exist in all sorts of forms, from ancients scrolls of vellum to outdated digital files from obsolete computing systems. Even digital files can pose a challenge given the plethora of file format, character encodings, and even document models that might have been used.

Many parliaments have multiple legacy formats to contend with, from different eras of the past. When it is desired to bring the entire corpus of documents a parliament has into the modern world, the challenge can be vast.

The first challenge, after identifying the extent of the document set, is to figure out a way to normalise all the historical texts in a common digital medium. This itself is a challenge when the medium has decayed, the language is arcane, and even the writing, often handwriting, is difficult to read. While OCR technologies can sometimes be useful, this stage usually requires extensive manual check and establishing the digital representation as authentic is challenging. This process can take years to complete.

However, extracting the text into a digital representation is only the first step in the process of modernising the data. The next step is where natural language processing comes into play. Tools must be devised to tease out the legal semantics of words into a format fit for computer processing. Document structures, interdependencies, references, temporal relationships, definitions, terms, and other metadata must all be teased out of the text.

Existing tools can accomplish this sort of data conversion to a very high degree of accuracy, but at a considerable cost. Without well-defined specifications for how old texts were written, and yet a high degree of required accuracy, there is a lot of trial and error followed by extensive verification. Newer natural processing techniques promise to reduce the cost of data conversion considerably, but only if the accuracy bar can be met.

When it comes to the information representation of the legal semantics that are teased out of the document, it is necessary to have a document model specially designed for legislation. While there are several proprietary document models in use today, there is now also a standardised information model. It is the OASIS LegalDocML standard, also known as Akoma Ntoso. It provides the facilities necessary to record the structure, connectivity, and underlying metadata necessary as the basis for an industry of well-vetted (and trusted) tools based on AI-based technologies to emerge.

Amendment Impact

Another existing use of natural language processing is in amendment processing. Amendments, whether to statutes, codes, or even to bills can be very mechanical, proposing change in very terse and arcane ways. What cannot be readily seen is the impact that these amendments will have.

Amendment impact programs attempt to prospectively execute amendatory language to show the language of a law or a bill as it will look if the amendments are adopted. This can bring a degree of clarity to an amendment proposal that would otherwise be difficult to ascertain.

Standardising Language

While some parliaments have already adopted consistent language for specifying amendments, some jurisdictions wish to adopt this practice to enable future automation capabilities. Natural language processing techniques are employed to analyse the entire historical record of amendments and to identify usage patterns and all the variations. These techniques can then be used to define a complete vocabulary to be used by computer automations when automatically generating amendment language in future systems.

Identifying Conflicts

Natural language processing is also used to identify conflicts that may exist between bills. By prospectively executing amendments and then using language comparison techniques, conflicts become clearer earlier in the legislative cycle, obviating the need for complex and costly conflict resolution measures to be employed once legislation has been adopted.

Codification

One of the most complex activities a parliament can undertake is codification or re-codification. This is an essentially refactoring of existing law to make it more consistent, coherent, and easier to work with. Codification involves restating existing law in a new arrangement – sometimes after decades of chaos have settled into existing law. Sometimes this involves organising laws, found in yearly statute books, into codes arranged by topic. Other times, it involves reworking the codes to simplify or otherwise restate the law.

The complexity of this process means it can take many decades to complete. For example, the State of California began the process of codification in the late 1920s, and this process continued for the next 30 years. At the US Federal level recodification began in 1926 and continues to this day with no end in sight.

The process of codification is extraordinarily complex, ensuring that no substantive changes are made to the law in the process of restating the law. The paper trail that must be recorded is immense and must work against a dataset of law that does not stand still which this process is underway.

Natural language processing is increasingly being used to help expedite this difficult task, by improving search, tracking changes automatically, and helping to restructure or restate existing provisions.

Emerging Uses for Natural Language Processing

With the rapid advances in Artificial Language technologies over the past few years, including the field of natural language processing, several new opportunities to apply the technologies are opening.

However, caution is required when applying off-the-shelf technologies to the process of creating legislation. While the legislation produced by a parliament provides a vast and well-defined corpus of text to apply to natural language processing, the information model, behind the scenes is very complex.

Legislative documents exist in many states of being. Bills begin as drafts and, if enacted, end up as laws. Along the way, they are subject to a lot of debate which can result in numerous amendments which propose changes, sometime in conflict with one another. Understanding the complex lifecycle of a bill and the associated documentation is key to being able to provide effective tools to automate the process.

One of the most important aspects to consider is privacy. The attorneys that work on legislation often work with staffers and politicians in ways that are subject to attorney-client privilege. Whether it is a draft that is being prepared or an amendment to be proposed, there are times when the privacy veil must be protected. Often, the underlying policy document that can tie all the disparate documents that define a piece of legislation in totality is the very document that is subject to attorney-client privilege. When training a natural language processor, the dataset must always be provided in such a way that privacy does not inadvertently get pierced.

Many of today’s laws were enacted long before computing tools were available to automate the process. These laws were written in the paper era and still have many of the limitations that come as a result. Sometimes, information is missing or specified in complex ways that require a human with an extensive understanding of the context to follow. For example, a citation or reference to a provision of a law always has a temporal aspect, even if not explicitly stated.

Also, although parliaments are reluctant to admit it, there can be numerous errors that have never been caught and fixed. Sometimes there is ambiguity – such as when numbering gets inadvertently duplicated or reused. Often, manual procedures are in place to work around these complexities. It will fall upon any natural language processing tools to identify these situations and apply the same intricate procedures a human would to work around them.

The vocabulary of legislation is relatively small and quite precise. Unfortunately, this is not always a benefit. While most jurisdictions derive their procedures from a common set of core procedures defined by the early Westminster system, often these procedures have been warped and reshaped over time by individual parliaments to suit their own needs. Sometimes, the result has been to redefine terminology in ways that are both familiar and confusing. Configuring natural language processing technologies to be aware of these subtleties can be quite a challenge.

One of the key considerations of any tools introduced into parliamentary procedures must be the degree to which it improves the functioning of the organisation. Today, laws are crafted by attorneys with special qualifications to make the judgments necessary to enact valid laws. These practitioners of the law have very specialised knowledge of areas of the law. Tools must be provided that improve their effectiveness and efficiency rather than striving to replace them.

The result should be tools that act as a co-pilot rather than auto-pilot when it comes to drafting legislation. There is a risk here. Much as the calculator can take away an engineer’s intuition for numbers, too much of a reliance on a drafting co-pilot can lead to complacency in drafting. For an engineer or a software developer, errors that come from too eagerly accepting the computer’s suggestion are often quickly uncovered. For the drafting attorney, where the check and balances may be more subtle, these sorts of error might easily go unnoticed.

This leads to the next consideration – algorithmic transparency. For an application to be working alongside a drafting attorney in the role of a co-pilot, full transparency is a must to understand any inherent bias in its algorithms, whether intentional or not.

Policy Making

Much of the hype surrounding the application of Natural Language Processing and other AI-based technologies to legislation tends to be superficial, expressing the task of making law in generalities. This often derives from a superficial understanding of how legislation works. When considering the actual process of legislation, the application of new technologies becomes less clear. That is not to say there is no application for new AI-based natural language processing. Rather, it is an acknowledgement of the complexity of the problem.

Writing legislation is a complex series of steps. It starts with a decision on a policy change and ends with specific changes to the law to implement those policies. Along the way is a complex series of steps known as the legislative process, involving debate and commentary along the way. Let us look at the steps:

The first step is to identify, in plain language, the policy effect that is desired. There may be many stakeholders involved with differing opinions that will shape the overall policy. While this stage usually kicks off the process of creating legislation, it may continue throughout the legislative process up until the final version is put to the final vote. Much of this aspect is subject to privacy rules and cannot be used as source data for training an AI-based system. However, it is quite possible that an AI-based assistant can be used to formulate the policy, suggesting changes based on its understanding of existing law, the current political climate, and other factors available to it.
The next step is to translate the desired policy effect into the mechanical language of a bill. Depending on the legislative drafting practices adopted by a parliament, there might be several steps to take plain language policy effects and turn them into concrete changes in the law. This is where natural language processing could have a significant impact on the process. The process of translating plain language into specific amendments to existing law is particularly complex. However, as the original policy change may remain fluid, the process of tracking the relationship of plain language policy changes to concrete amendments to the law becomes quite complicated. Specialised natural language processing technologies are necessary to maintain these relationships, adapting them as new changes are proposed, or existing ones altered.
The next step opens even more complexity. Once a bill draft is introduced, it becomes subject to the amending cycle where many amendments, both supportive and not, might be introduced to the language of the bill. These amendments, which will be described in more detail later, need to be reversed through the process to understand not only the impact on the law, but also how they modify the policy changes being proposed.

Augmenting Drafting

Drafting attorneys are a key asset of a drafting office. Their knowledge of the law is invaluable. Many times, the experience they have amassed over the years can help guide them to draft effective legislation. It should be the goal of any AI-based system to augment the skills of drafters to allow them to be more effective as well as distributing the know-how of the drafting office to provide continuity in drafting capability as drafters retire or move on to other roles.

Specialised tools skilled in understanding the historical context of legislation, both past and present can provide valuable insights into what may or may not be politically acceptable, what the consequences of proposals might be, how likely a policy change might be invalidated or preempted due to constitutional or other higher-level laws.

Summarisation

Often, the mechanical language of legislation is accompanied by a summary or digest that explains the policy intent of the legislation. This is usually a sanitised summarisation of the policy documents and discussion that led to the legislation being proposed. The summarisation capabilities of natural language processing can naturally be used to produce this description, with appropriate safeguards to ensure that all privacy considerations are considered.

Analysing Comments

In some areas of legislating law, public commentary can play an important role. There are several ways in which natural language processing can play a role:

If the number of comments received is very large, natural language processing tools can sift through the comments to make sense of them before a human ever looks at them, arranging and grading them for processing.
Natural language tools can also be used to sift out comments introduced by troublemakers intending to gum up the process with noise or by laying traps to derail legislation.

Policy Simulation

There is often discussion about how the law might be simulated. This would help to do what-if analysis and understand the impact of various proposals and amendments. As a general capability, this remains a far-off dream. However, in specific areas such as tax policy, simulation models already exist. The real question is whether these models can be integrated with natural language processing to allow policy change proposals to be tested against these models in near real time by stakeholders lacking specialised know-how to drive these models

A related capability would be to apply experience data from similar legislation to predict the outcome of legislation – allowing various scenarios to be tested in the current political climate and improving the likelihood of a positive outcome.

Improving Search

One of the most obvious applications for natural language processing is to improve the search experience. Of course, allowing some to request a query using natural language is desirable, but there is so much more that can be offered.

Today, legislative information is distributed across a vast set of documents and databases. While tools are available to show the relationship between all these disparate data sources, it is quite primitive, lacking much of the semantic information that would allow an AI-based system to apply truly intelligent queries.

Data sources that can and should be integrated include statutes, codes, prior drafts (usually with privacy considerations), pending legislation, enacted information, analysis, votes, proposed and adopted amendments, and so on. The connections quickly become very complex.

Inter-Jurisdictional Harmonisation

Building these tools is beyond the expertise and budget of any single jurisdiction. For these tools to become a reality, there is a need for some level of cooperative industry of tool vendors to emerge. This is where having a more robust information model, such as that provided by LegalDocML, can be invaluable – paving the way for tools that integrate information across a parliament and between jurisdictions in truly smart ways.

Building these tools and technologies creates new challenges. Some legal scholars refer to a future they call the “legal singularity” – when all laws have been harmonised across jurisdictions, made complete, and made computer processable – almost as a computer program.

This is a long-term dream that is unlikely to ever be realised. Jurisdictions simply have little motivation to work on such a monumental undertaking given the immense cost and disruption it would cause.

Instead, new tools and technologies must be adapted to the local vocabulary and processes that already exist. This means two levels of training are necessary for a natural language processing system. The first is meta-level of understanding about how a parliament or jurisdiction specifies, creates, and disseminates their legislation. The second is an understanding of the laws that result.

Inter-Jurisdictional Cooperation

There are narrow subjects where some level of cooperation is essential because of the capabilities it would enable. Examples include the enforcement of treaties and other international agreements. Global challenges such as climate change, cryptocurrencies, social media, and financial and other regulations demand increasing cooperation among parliaments. Natural language processing tools offer capabilities that promise to reduce the friction of complying with these agreements.

Conclusion

Natural language processing is already a well-established technology employed by parliaments today to automate much of their modernisation initiatives and existing procedures. However, with the rapid advancement of Artificial Intelligence technologies in recent years, new opportunities to provide innovative solutions for parliaments are being enabled. The challenge is going to be to adapt off-the-shelf technologies to the specialised needs of parliaments where complex information models are the norm, where privacy considerations are mandatory, and where extreme accuracy is demanded. The goal must be to augment the drafting attorney, making them more effective, reducing bottlenecks, and providing drafting offices with continuity as drafters retire or move on after amassing decades of experience that they take with them.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Building The Case For AI-Powered Parliaments & The Necessary Guardrails: Houses Of The Oireachtas

Fri, 30 May 2025 17:59:44 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction (Building The Case)

This paper provides a comprehensive summary of the ICT teams work to date on assessing the possibilities of AI in Parliaments, specifically within the Houses of the Oireachtas. It outlines their journey from initial research and familiarisation with Generative AI to the development of use cases and the formulation of a responsible AI framework and overarching AI strategy. This document aims to build the case for AI-powered Parliaments while emphasising the need to develop and establish the necessary guardrails to ensure responsible and effective implementation.

Since the last quarter of 2022 to present day (Q3, 2024) the Houses of the Oireachtas has been engaged in a multi-step process of assessing and evaluating the potential use of Generative AI. From Q4 2022 to end Q1 2023 that process involved early familiarisation and self-driven research in the area, as it became a prominent topic globally. Dedicated members of the Houses of the Oireachtas ICT team engaged with Generative AI content across all communication mediums such as research publications, news articles, lectures and talks on content streaming platforms and Government publications. This early emersion in the topic allowed team members to be well informed on the topic and key aspects within and across it to be able to assess it and its potential relative to our data and processes.

Informed by this self-driven research the Houses of the Oireachtas ICT team, across late Q1 and Q2 2023, then began to ideate use cases themselves while also engaging directly with ICT partners such as Microsoft and Accenture. With Microsoft the team looked at possible early use cases for natural language processing search and content summarisation on public data sets such as Parliamentary Questions. They also spent time understanding the defining the differences between services such as Azure OpenAI and the emerging Copilot which would become generally available in Q4 2023. With Accenture the team took the internal ideation that had been progressed around 30 possible use cases to develop them out more fully with assistance from the Accenture AI team and broadly categorise them.

Just as the self-driven research stage had developed the team’s ability to engage at a technical and use case level with ICT partners, the partner engagement then led the team to be able to begin to develop a briefing and engagement plan with key stakeholders. Across Q3 2023 meetings and workshops were held off-site with key staff working in wider digital transformation areas and the management board of the Houses of the Oireachtas. The purpose of the workshops was to distil all the information gathered and potential applicability of Generative AI into understandable concepts and relatable use cases for the organisation. To do that the team prepared a briefing and prototype use case demonstration session covering a wide range of topics.

Stakeholders were introduced to foundational knowledge such as understanding artificial intelligence’s spectrum from Narrow/Weak commonly understood ubiquitous forms to newer emergent Generative forms. The focus then turned to Ireland’s national landscape by examining how various Government departments and agencies are examining and taking early steps towards potentially adopting this general-purpose technology. From there the briefing then moved towards peers and what international counterparts in other Parliaments were doing or planning to do, to gain a broader perspective. Key national publications such as “AI Here For Good: National Artificial Intelligence Strategy For Ireland” were reviewed alongside critical cybersecurity guidance “Cyber Security Guidance on Generative AI for Public Sector Bodies” tailored for public sector bodies from Ireland’s National Cyber Security Centre.

Furthermore, the group were then briefed on the key framework components of the coming EU AIA (Artificial Intelligence Act), which would later be published in Q1 2024, and its implications for organisations looking to adopt and integrate Generative AI into work processes and service delivery. These were important discussions to prompt and engage in questions and answers around with the group before moving onto demonstrations of and discussions on use cases or proofs of concept. This attention to detail on guidance and proving of a firm grasp on coming regulation clearly illustrated that the team progressing the case for Generative AI knew that balance between embracing innovation and meticulously considering governance & risk management were both crucial for advancement.

With foundational concepts and the importance of regulation and governance covered it was then possible to brief stakeholders on practical Generative AI examples. The team started in the area of unintended consequences and the risks posed to content and information (video) generated by Parliaments which could be manipulated by increasingly real deepfakes that have emerged from Generative AI technologies. From there however the practical side of Generative AI was equally emphasised, through demonstrations from the private sector, showcasing Generative AI applications in the area of call centre assistant agents and content creation tools already in use by advertising teams of global consumer brands to exemplify real-world utility.

This set the stage to turn to early internal prototype demonstrations, which had been developed in consultation with Microsoft and Accenture to showcase how Generative AI could transform internal processes. Specifically, it was demonstrated how Generative AI could dramatically change user experience and engagement with Parliamentary Question data through contextual search. Additionally, the team demonstrated how the manual process of generating highly detailed summaries from Bills & Bill Memos in the form of Bills Digests could be recreated by training Generative AI.

The comprehensive nature of the briefing and workshop demonstrations enabled the team to then propose documenting the research and exploration to date into a next steps paper for the management board which would propose an actionable path forward. The paper was developed across Q4 2023 and presented and approved in Q1 2024. In it 6 of the original 30 use cases which had been developed were outlined in more detail with a proposal to advance 2 put forward. Estimates and a timeline to end Q2 2024 were agreed to. In July 2024 an update and further steps paper was then presented to management board. This paper outlined the work of the team involved in developing the use cases, the architecture that was designed and a recommendation of how the work should continue. Before outlining the next steps, the Houses of the Oireachtas will take across the remainder of 2024 a summary of these 2 use cases follows:

Use Case 1 – Contextual Search & Summarisation

Use Case 1 was built for the Library and Research Services (L&RS) to assist them in carrying out their research activities. The application has two separate uses:

Contextual Search

The user uploads any number of documents, which are generally research papers in PDF format relating to a particular topic of research. The user can then use a chat interface (similar to ChatGPT) to ask questions about the research topic. The application provides a response to the query based solely on the information within the uploaded documents, along with in-text citations referencing the source(s) of the response. The application is a method for the researcher to search across multiple documents with one natural language query, receiving a comprehensive response which synthesises information from multiple sources to give an answer. This saves the user the time that would otherwise be required to manually search through the documents for content relating to their query and compile an answer.

Summarisation

The user can use this tool to generate a summary of an uploaded document, such as a research paper that the individual has written. The summary generated can be in one of three style options: Blog Post, Executive Summary, or Other (user specifies their requirements using text). The application leverages GenAI to provide a comprehensive summary of the document. The user can also input additional instructions to the application if they want to further refine the summary to their preferences.

Iterative feedback and development process for use case 1

For use case 1, user feedback was received in the form of a template. This allowed the SMEs to submit feedback on any response generated by the model and facilitated iterative improvements to the application over the proof-of-concept timeline. The users were asked to provide scores for correctness, completeness, conciseness, and an overall score for each response, along with a free text response to explain the reasoning for the scores provided. Once feedback was gathered and stored, a feedback session was coordinated to talk through the feedback. The feedback was then used to form actionable changes included in a change log. Changes were prioritised and the developers implemented a feasible number of changes before generating new outputs and starting the feedback process again. Any changes that could not be implemented given the time constraints were included in a backlog.

The application delivered to the testers also included the option to provide feedback directly for each response. This has meant that ongoing feedback can continue to be gathered, which in future can be used to make further improvements to the application. In addition, this use case application was shared with the Office of the Parliamentary Legal Advisor to determine whether it may be useful to researchers in other units and the response to date has been positive.

The application has been welcomed by the business team who have been working on this project. In particular, they see a benefit in being able to summarise research papers that they have written themselves. Suggestions have been made as to how the application could be improved further and this can be scheduled in parallel with the other steps to be taken on Generative AI.

Use Case 2 – Automated Report Generation

Use Case 2 was built for the Digital Team to assist them in proving the concept of automatically creating the first draft of a Parliamentary journal. It should be noted that the scope of this use case did not include business items which are not taken in the House and coming from the Clerk Sheet.

Iterative feedback and development process for use case 2

A combination of Traditional AI (OCR - Optical Character Recognition) along with Generative AI was used to solve this business problem. The focus was building automated Parliamentary journals. These are very text-heavy reports that require data from multiple sources, including both digital and non-digital (handwritten) documents. The first input file was handwritten Clerk Sheet, in a mix of Irish and English. Multiple people write on the same Clerk Sheet throughout the day, so there is a high variance in the handwriting. The sheet contains information that is not available elsewhere, such as the start and end times for each topic that was discussed in the Dáil on a particular day.

The first step of the solution leveraged the OCR technology to create a digital, comma separated value (CSV) version of the handwritten sheet - which was the first output for the user. The second step utilised this OCR’d Clerk Sheet, along with blank templates of the journals and completed journals (to use as reference material) which were provided by the Digital Team. The AI then filled in the blank journal template using the data from the OCR’d Clerk Sheet by leveraging the completed journals as a reference.

The third step involved the AI application connecting to the internet (www.oireachtas.ie) to fetch the debate transcripts related to the relevant sitting day so that the information could be included in the Clerk sheet. As the verbatim transcripts can be very long, spanning many pages of text, the AI revises its initial draft report from the previous step, augmenting information into each of the sections where required (again, utilizing completed journals as reference). Essentially the AI takes relevant information from the Debate, where required, based on the reference journals. This augmented information includes resolutions on topics discussed on that date (which are crucial to these Parliamentary journals).

For the final step a layer of formatting is introduced into the process. After the AI application has processed and collated the corresponding data from the handwritten Clerk Sheet (via OCR) and official report (from Oireachtas.ie), using historical journals as reference guides to fill in the blank template, it then goes through a two-stage formatting process. This is because the first output of the draft journal is in plain text, without any formatting, as large language models (LLMs) only work in simple text outputs. To address this and deliver the formatted output required, HTML (Hypertext Markup Language) coded versions of the templates, which include HTML that define the bolding, text sizing, positioning etc., were developed. The HTML code could then be rendered to produce a formatted version of the report as a .html document, and finally converted to a Word document.

The digital team has engaged with the GenAI developers on this application as the Journal Office had flagged that they would not have sufficient availability to engage. The output shows good potential in terms of how the output looks and the ability to work from a scanned handwritten Clerk Sheet. This is a complex output in that different business items require different elements from the debate to be inserted into the Journal. Arising from this, a number of further issues to be tackled and addressed have been identified and work will continue on this use case to generate the best possible draft Journal for checking. When this work is completed, it is intended that the Journal Office and Seanad Office will be shown the output and further assessment can be conducted.

Next Steps (The Necessary Guardrails)

Having constructed the case for the potential use of AI in Parliament and then receiving approval to demonstrate that potential through early proof of concept use cases it is now vital to focus on developing the necessary guardrails in the form of a responsible AI framework and overall AI strategy that will be required before this technology can be used to transform any processes or services. This was the key next step recommendation made by the ICT team to the management board of the Houses of the Oireachtas in July 2024.

Responsible AI

The ICT team stressed that in the use cases to date, the applications had been assessed using Accenture’s Responsible AI framework until the Oireachtas could conduct the work to build out an internal Responsible AI Framework. It was recommended and approved that across Q’s 3 & 4 of 2024 the ICT team will hold a series of workshops to develop a Responsible AI Framework which can be utilised for all applications developed for live use within the Oireachtas. A number of key stakeholders across the Oireachtas will be involved in this process, including the Data Protection Officer, Chief Risk Officer and Office of the Parliamentary Legal Advisor. It will also be essential to ensure that the EU Artificial Intelligence Act and the associated guidance is a foundational element of the Responsible AI Framework for the Oireachtas.

Early planning work on the Responsible AI framework workshops has commenced and across the remaining 4 months of 2024 the ICT team and stakeholders will work to assess and develop a framework around these 8 key risk dimensions.

Soundness: Ensuring that AI systems are reliable and perform as intended, minimising errors and inaccuracies.
Transparency: Making AI processes and decisions understandable and accessible to stakeholders, fostering trust and accountability.
Privacy: Protecting personal data and ensuring that AI applications comply with data protection regulations.
Sustainability: Developing AI solutions that are environmentally sustainable and resource-efficient.
Fairness: Ensuring that AI systems are unbiased and equitable, preventing discrimination and promoting inclusivity.
Robustness: Building AI systems that are resilient to adversarial attacks and can maintain performance under varying conditions.
Accountability: Establishing clear lines of responsibility for AI decisions and actions, ensuring that there is oversight and governance.
Liability & Compliance: Ensuring that AI applications adhere to legal standards and that there are mechanisms in place for addressing any legal issues that arise.

Overarching Strategy

Sitting above the Responsible AI Framework will be the overall AI Strategy. The AI Strategy work will be coupled along with the engagements around the Responsible AI Framework. The full set of actions across both (research, cross-directorate workshops, skills gap analysis, AI maturity assessment, development of value assessments and metrics)are designed to generate the required outputs (vision statement, principles, values, roadmap, resource plan, training and change management plan, governance operating model and budget estimate) which will then be brought to the Management Board for review, feedback, hopefully approval, and then long term implementation from Q1 2025 onwards.

To conclude, for now, it is the ICT teams hope that upon the successful completion of the Responsible AI Framework and the overarching AI Strategy, approval will be granted for the development of a comprehensive three-year programme of AI development work. This programme’s aim will be to further integrate AI technologies into our Parliaments processes, while ensuring that innovation is balanced with robust governance and risk management.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Multilingual Parliaments powered by AI

Brahma Mishra — Tue, 04 Mar 2025 17:03:38 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction

While AI has made significant strides across many sectors, some of the most visible and closely observed organisations are taking a more measured approach in adopting this transformative technology. Understandably, their hesitation and scepticism stem from two concerns – maturity of the AI products and putting the adequate guardrails in place to avoid unwanted disruptions in parliamentary operations. Change management and resistance from the existing workforce fearing retrenchment is another issue to contend with.

Challenges Facing Multilingual Parliaments

Parliaments around the world generally fall into three categories: monolingual, bilingual, and multilingual. While multilingual parliaments are fewer in number, the challenges they face are unique and often more complex.

Over time, most parliaments have developed reliable systems to record speeches, debates, and discussions, often relying on highly skilled stenographers, or Hansard staff. Their work is crucial in ensuring the accuracy of parliamentary records. However, as technology continues to advance, new opportunities arise for Hansard offices to enhance their processes, as explored in another chapter in this book.

Key Communication Barriers in Multilingual Parliaments

In parliamentary and committee settings, both live and recorded communications must be easily understood by all members. However, in bilingual and multilingual parliamentary environments, this isn’t always feasible. The Parliament of India, with 22 recognised regional languages, and the European Parliament, which operates in 24 languages, face significant challenges in ensuring smooth communication. In such scenarios, language barriers can sometimes make it difficult, if not impossible, for all members to fully engage and participate in discussions.

The other challenge is, disseminating the information captured by the Hansard staff. First, they need to be reviewed and vetted (which is currently limited by recording processes). Then, depending on the language requirements, the transcript needs to be translated before getting disseminated.

AI-Driven Solutions for Multilingual Parliaments

Addressing the challenges of multilingual parliaments requires embracing advanced technologies that support inclusive and comprehensive digital transformations, while ensuring that the human element remains central to parliamentary processes.

Utilising the self-learning and generative capabilities of AI, a branch of Natural Language Processing is maturing fast in terms of recognising speeches, understanding nuances such as punctuation, sentiment etc. and cutting out occasional profanity before reproducing the communication in the target languages. Platforms with capabilities of recognising speech without prior training have eliminated the awkward hours spent training the model on each user’s voice. The technology also has the ability to do gender recognition and produce the translated output matching the input gender.

Simultaneous Interpretation

Voice recognition hinges on understanding of the phraseology, grammar, tone, and sentiment of a segment of human voice. Advanced algorithms also have to track the speed of delivery that can be a crucial component in syncing translated audio with transcript or close captions, which differs in number of words and sentence lengths from the original language.

Advanced NLP platforms bring together several pieces of the puzzle in trying to reproduce the most accurate, structurally correct, and sentimentally resonating translated voice output while ignoring ambience noises.

One notable example from the private sector is MeetMonk, an AI startup working on the deep-learning NLP algorithms, that have been working with parliaments in this domain.

The image below represents an example of such a platform, with simultaneous interpretation for 85 languages with very high accuracy, low latency, and with human voice output.

Image 1: Voice to Voice Interpretation interface from MeetMonk

Live Stream Translation

The problem is more accentuated when live stream audio/video needs to be translated. Many parliaments broadcast the proceedings through live streaming with significant challenges for multiple languages in real-time mode. The challenge with the live stream translation is to translate audio first and then match it up with the input video frame by frame so that the video looks natural. With advanced audio syncing algorithms, parliaments can ensure the time-sliced video is tagged correctly and then the translated audio is superimposed accurately.

The video is accompanied with live close captions in the target language it is broadcasted. This technology aims to modernise the way parliaments’ TV transmissions happen today.

Document Translation

This is one area that every parliament needs to focus its attention on to increase efficiency. Parliaments currently translate millions of pages into multiple languages through a largely manual process. While this approach is thorough and time-tested, it is resource-intensive and time-consuming.

With the AI models maturing quickly, and OCR (Optical Character Recognition) becoming more accurate, the new emerging trend is to employ AI to read through and translate documents much faster. In the final step, human reviewers vet the translated documents to ensure its accuracy. This is both fast, efficient, and cost effective. Below is an example of document OCR and translation into multiple languages.

Document Translation platforms can translate documents of different sizes into multiple target languages very quickly. The translation is simultaneous, it displays a word count for tracking, and has the ability to download and review all in one place. That makes the process much faster and efficient.

Image 2: Text Interpretation Interface from MeetMonk

Image 3: Text Interpretation Interface from MeetMonk

Overcoming Barriers to AI Implementation in Parliaments

Despite its significant power in making operations smoother in many parliaments, AI still faces challenges in getting widespread implementation. It is because of various human and technical challenges.

AI for multilingual parliaments is a set of tools that can improve based on the feedback it receives. It is evolving fast and is capable of scaling new technological challenges every day. However, there are many technical challenges that are specific to the needs of parliaments. Below is a list of some of the challenges.

Deploying advanced recognition algorithm for languages without need of prior training
Using matured LLM (large language models) for the target languages
Vocalising with human-like voice
Optimising for low latency and high fidelity
Utilising profanity filter
Recognising gender and reproducing corresponding output voice
Diarisation and speaker separation

The platform must be intricately designed to adapt and efficiently support a range of legislative activities, and adequately address the needs of different legislative offices. Meticulous planning is crucial for an implementation that marks a significant step in modernising parliamentary operations, harnessing the power of advanced technology for improved efficiency, transparency, and citizen engagement. It must also be very mindful of keeping humans in the loop, deploying thoughtfully designed guardrails, and ensuring that the gaps in the processes are caught and swiftly corrected. Only then, will it have the means to serve the purpose it promises and set a new benchmark in the use of AI in governance.

Conclusion

The adoption of AI in multilingual parliaments presents a transformative opportunity to overcome long-standing communication barriers, improve operational efficiency, and ensure that all members are able to participate fully in parliamentary processes. By leveraging cutting-edge AI technologies, parliaments can modernise their translation, documentation, and broadcasting systems, making them more inclusive and accessible to diverse linguistic groups.

However, successful implementation requires careful consideration of both human and technical challenges. The importance of maintaining a human element in decision-making, deploying AI solutions thoughtfully, and ensuring thorough vetting processes for translated content are crucial to achieving the desired outcomes. With AI continuing to evolve rapidly, it offers unprecedented potential for innovation in parliamentary functions, provided that the right balance is struck between technology and tradition.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Practical Experiences that Can Help Practitioners Modernize Legislative Bodies and Leverage AI Effectively

Wade Ballou — Wed, 12 Feb 2025 12:02:28 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction

The legislative process is at an interesting inflection point for a number of reasons. First, one of the silver linings from the COVID Pandemic has been the worldwide connection of many who are interested in modernizing their parliaments and its various processes. International webinars have connected parliamentarians, officers, and staff from across nationalities, highlighting many similarities at the institutional level.

Second, parliamentarians have come to recognize that by incorporating digital technology they are better able to understand their constituents’ desires and the impacts of their legislative decisions as well as to respond in a more timely manner to perceived needs.

Third, digital technology affords the possibility of improving the efficiency, accuracy, and accessibility in the legislative process, especially through the application of artificial intelligence.

Fourth, legislative drafting, which has all too frequently been an neglected in the policymaking process and viewed as an obstacle and not a tool, is now receiving more attention.

Fifth, these developments are challenging existing processes and traditions, many of which are centuries old, paper-based, and slow to change. Finally, parliaments and their institutions are frequently under-resourced.

Drafting Office

Central to these developments is the drafting office, whosecore function is to express parliamentarians’ solutions in legislative language to the constituent problems If enacted into law, these solutions carry out the intended purpose.

Drafting is central to governing. The Code of Hammurabi, for example, expresses a policy in writing on stone. Whether the medium is stone, paper, or digital, the policy of the governing authority must be expressed. Policy making and policy expression are intertwined, yet separate. In today’s world, a well-crafted draft requires knowledge in four realms: subject matter, politics, law, and law writing (or drafting). Typically this expertise comes from three sources:policymaker (comprised of the parliamentarian and their staff), subject matter experts, and drafting experts. The interaction of these three roles produces the draft. An able drafter with less subject matter knowledge requires more input of subject matter expertise. A generalist drafter needs a subject matter specialist while a drafter with subject matter specialty can function well with a more generalist policymaker.

The drafting function consists of receiving a request to express a policy in legislative language, understanding the problem to be addressed in both factual and legal terms, and crafting a legal solution using appropriate legislative language. Drafting occurs in a dynamic, legislative environment where efficiency and draft quality are often in tension. Efficiency is typically of great interest in the political process while quality is of great interest in the legal implementation of the resultant law. Efforts to improve the efficiency of the drafting process have often accrued to the benefit of the political process. Yet, meeting speed and timing requirements of the political process may result in sacrificing legal quality in order to achieve political success in the short term. The digital revolution with its large datasets and the application of artificial intelligence has the potential to assist the drafter in efficiently producing quality drafts.

Introduction of the Computer and Increased Risk

The introduction of the computer for the drafting process greatly reduced the time required to move from thought to the written word. Yet there are risks in introducing more technology into the drafter's environment. Drafts and related legal advice are "credence goods", which means that their quality is difficult for legislators to assess before using them in the legislative process. As a proxy for this assessment, the parliamentarians and their staff rely on the word of the drafter. The drafters must be able to rely on the technology for assistance. If unexpected errors or results occur, this trust can be undermined. Increasing the speed of drafting also increases the likelihood of errors, which reduces the credence good value attached to the draft. Small word changes become easier to make and carelessness in process (not necessarily the individual actors) can become unavoidable, despite the emphasis on quality drafting. The push for speed can lead to an over-reliance on quick turnarounds, resulting in bills being scheduled for consideration before the language is ready. This increases the pressure on all involved to produce a draft and increases the risk that it will be insufficient to implement the intended policy. Components of this risk include the risk that the draft will:

lack internal consistency.
fail to execute technically or fit appropriately into current law.
conflict with existing law.
be unintentionally vague or ambiguous.
be over- or under-inclusive.
will encounter parliamentary procedural issues during the course of its consideration, including unwanted committee referral or points of order.
if enacted, will be legally unenforceable.

Where might digital tools increase both efficiency and quality in the drafting environment? What are the pain points in the drafting process that can be addressed?

Drafting Process

Drafting begins long before the drafter is engaged. First, the policymaker must identify a problem to be addressed and the desired solution, or policy. Broadly speaking, this is the constituent and stakeholder engagement stage. There can be valuable conversations during this stage with a drafter as the specifics of the problem are identified and the policy refined. Then, sufficiently specific drafting instructions must be provided to the drafter. A truism expresses a critical aspect of drafting: a drafter cannot produce a draft for a policy unless they understand the problem, the policy, the solution, and the law. When drafting instructions are not sufficiently clear, more clarifying conversations before and during the drafting process are needed. Failure to engage in these conversations results in false drafting starts and drafts that do not reflect the desired policy. Better guides, such as well-designed dynamic forms and chatbots, made for the use of the requester, can lead to clearer, more complete instructions as well as identify information that will be needed. The information collected by these guides and chatbots could be passed automatically through the intake stage of the drafting office to the appropriate team.

Request Intake

Drafting offices have two general methods for receiving requests. Most requests come through a central intake process. Those policymakers (committees and leadership) with established relationships with drafting teams will make their requests directly to the appropriate drafters. In the case of central intake, the intake team triages requests and refers them to the appropriate team. This central point can become a bottleneck at times of high request volume. Stress on the intake team increases as the speed of routing of requests increases. Inevitably, backlogs develop and errors in the referral of these requests increases. Valuable drafting time can be lost. There is increased friction within the office and between the office and policymakers. Digital technology allows routine requests directly to the appropriate drafting team, while incomplete or complex requests can be routed to the intake team for personal handling. However, improvements to the intake system without additional improvements to the drafting system simply increases the backlog with the drafting teams.

Heart of Drafting

The heart of drafting requires a knowledgeable drafter writing with the conventions appropriate to the legislative practice of the jurisdiction concerned. It also requires access to current law. The conventions used in the United States Federal system are well described in the Legislative Drafter’s Desk Reference, 3rd Edition: Best Practices in Drafting Federal and State Laws and Regulations, by Sandra Strokoff and Lawrence Filson (The Sunwater Institute, 2024).

The Comparative Print Suite, developed by the Legislative Counsel and the Clerk of the House illustrates the development of the component parts of drafting and their development over time. There are several insights and lessons to be learned from this project which can be applied to other modernization efforts in the legislative drafting area.

Comparative Print Suite

Although the Comparative Print project formally began in 2016, its genesis is in the early days of the Office of the Legislative Counsel (HOLC). HOLC was founded in 1919, and the first drafters brought a professional approach to drafting that began to introduce uniformity and consistency to Federal law. This approach could be highly technical, making cut and bite amendments to law difficult to understand to the non-specialist reader.

Therefore, in 1929 the House of Representatives adopted a rule that required committee reports accompanying bills reported from committee to contain a comparative print showing how the bill would change law. This requirement, known as the Ramseyer Rule after its author, Rep. Ramseyer, has been in effect in the House in some form since then. Over time, HOLC reduced the laborious task of researching the law for each draft by developing and maintaining notebooks of the up-to-date version of nonpositive laws that were commonly amended. These notebooks consisted of the original statute with penciled repeals and taped inserts. Positive laws were available from the Office of the Law Revision Counsel and required much less effort to research. From time to time these compilations would be typeset by the Government Printing Office and the process would continue. These compilations were maintained by HOLC to expedite both drafting and the Ramseyer process. They were also used by House committees and agencies. (See Office of the Law Revision Counsel discussion of positive and nonpositive law: positive law codification (house.gov)) Updating these compilations could be slow and tedious, as the cut & bite amendments contained in an amendatory statute were reversed to incorporate the enacted changes into the law. These Ramseyers served as quality assurance for the drafter, both for the technical execution of the amendments into law and to read the law in its “as amended” form to check its meaning. Downstream from the drafters, Members, staff, and stakeholders used the Ramseyers as a way to quickly read the proposal in context.

This manual process for producing Ramseyers served the House for 75 years until 2004, when bills, resolutions, and amendments began to be created and maintained in a XML format, the common data format chosen pursuant to a Congressional mandate. As part of the XML data revolution, HOLC developed and maintained a software tool for internal use to assist in the creation of Ramseyers and to create and maintain its growing set of compilations of nonpositive Federal law. This tool could only be used by highly trained specialists and required manual intervention and judgment to achieve the desired quality.

In 2010 HOLC integrated a variation of the Ramseyer tool into its XML authoring tool. This advancement gave HOLC drafters the ability to create on-demand mini-Ramseyers from within the authoring tool for the purpose of their own use to quality check their work before providing drafts to Members and staff. However, these tools did not achieve an acceptable level of accuracy for distribution outside of HOLC. Yet, Members and staff were working with greater legislative complexity, a faster process and larger documents, increasing the need for Ramseyers in more stages of the process.

The situation changed in 2017 at the beginning of the 116th Congress with an amendment to the Rules of the House. The House required the Clerk and the Legislative Counsel of the House, by the end of the calendar year, to begin providing documents that were the approximate equivalent to the Ramseyer, but now at the stage of the process when changes were made at the Rules Committee and before the document could be considered on the House floor. This is the beginning for the development of the Comparative Print Suite (CPS). Each project should have an identified beginning point.

First, the Clerk and HOLC identified and categorized the challenges, which were of three kinds: political, technological, and resources. Politically, we began by identifying the stakeholders, which included the Clerk, the Legislative Counsel, the Parliamentarian, House leadership, the Rules Committee, the Committee on House Administration, and the Member who succeeded in making the change to the House rules. We committed to working in a nonpartisan way, including majority and minority, and to communicate with all of them throughout the process. This group early on identified the defining value: accuracy of product. Members would be using these products to engage in the legislative process. They had to trust the product before they could rely on it.

There were several collateral issues to face before building the software tool itself. The United States does not have a unified code of laws. Those that had been digitized existed in disparate formats. In addition, the digitized XML compilations, developed for the purpose of producing a nice print was not clean at a data level. Relevant data sources had to be identified, existing documents conformed to the appropriate XML standard, data processes standardized, and databases built specifically for the CPS tool. Much of this already existed in part due to the development of the aforementioned HOLC internal process.

An important lesson is to develop in tandem when possible and to collect small wins along the way. An initial set of 40 Statute Compilations were published in PDF form as a collection on the GPO website in late 2018. As of August 2024, there are over 2,400 documents in this collection, available in both PDF and the USLM version of XML. These documents are available to the public at no charge.

Perhaps the key challenge was improving the accuracy. To do this, the bills on congress.gov were analyzed for amendatory instructions and a natural language developed to express this structure as well as the citation patterns used in drafting. Subject matter experts from HOLC assisted in identifying patterns and in verifying output as the CPS tools were developed. In addition, using software to interpret language is not easy. The early XML-based tools used various forms of pattern matching, rudimentary at best, given the complexity of legislative documents, to identify where in a document an amendment was striking or inserting language. However, the structured nature of XML documents allowed for disparate XML formats to be conformed to a common standard. Additionally, citation work from a previous project proved invaluable for identifying precise locations for amendments. Another key lesson: build upon prior work and learn from it.

As discussed earlier, a successful draft requires drafting expertise (legal and writing), subject matter expertise, and detailed knowledge of the policy to be drafted. These characteristics are rarely found in one person. The very thrust of the project was to give the policymaker and the subject matter expert the ability to easily read the law in its proposed form. This required extensive engagement with non-drafters and a user design team to assist with the user experience. The project also identified the need for educational opportunities and notifications from within the tool itself of those situations where changes were not executed and why. Once these were sufficiently developed, a pilot group was formed and trained. Their feedback was solicited, and suggestions were incorporated into later releases. Training was required for user access to those portions of the software deemed to require additional knowledge to understand the results given by the software.

The design and development of the tool, which integrated the user interface with the natural language processing and the laws database, took place before releasing a minimum viable product. At the time of release, the product had been used by HOLC personnel, experienced committee counsels, and few more general legislative assistants. Additional work included increased server capacity, user security to maintain confidentiality, and cybersecurity checks. The project continues to solicit user feedback and maintains a log of possible future enhancements.

The Comparative Print Suite was built upon existing tools and knowledge, successfully making an important legislative tool available to a larger user group. There are other components within the drafting office that can be modernized to address current and future needs of the legislative process. In addition to speed and quality, other values to be protected are confidentiality of requests, data integrity, and system security. These values together are key ingredients in the foundation of trust, which is a key element of the legislative process. Parliamentarians must be able to trust that their requests for assistance, policy goals, and legislative strategy will be held in confidence. Drafters must understand these goals and the strategy relating to a proposal in order to produce a draft appropriate for use in the process. Previously, the draft has been described as a credence good. The drafter is providing a credence service. When considering areas to modernize, these values that build and maintain trust must always be paramount.

Additional Aspects of Drafting Offices and Project Considerations

Legislative data and the processes that rely on it must always be built and maintained with the highest standards. The United States Congress uses the USLM XML format, which is fully described here: Legislative Documents in XML at the United States House of Representatives. Drafting in this format and maintaining data in it allows for seamless use throughout the process without the problem of being translated between formats. Policymakers provide drafting instructions and drafts to the drafting office in a variety of formats. These formats must be translated into the standard legislative data format. These translations of data risk incorrect formatting and loss of data. Therefore, entering policy data in the appropriate format from the start is a modernization project that has the potential to improve collaboration between the policymaker and the drafter, with a result of increased quality and efficiency.

One can imagine a legislative draft as having a reading side, typically in PDF or Microsoft Word form, and a data side. The United States Congress uses USLM XML as its data format. Because it is in XML, its structure and tagging are important for downstream use. The data itself can be reused. However, a document management system is required. This data is typically of high quality and the database is controlled by the drafting office. The data should be maintained in a centralized repository with storage, management, and retrieval aspects. The system should be capable of versioning and showing track changes for improved collaboration with clients. This data can be used for searches and as a source of data for AI tools, so long as confidentiality is appropriately maintained.

The authoring tool for drafting directly in the XML format should be integrated with quality assurance features, such as spelling, style, and grammar checking. It should also be able to integrate with legal research databases to allow for the transfer of data easily into the draft with appropriate citations. The authoring tool and its related database should contain templates to improve both quality and efficiency.

Teamwork and the ability to work on legislative drafts both synchronously and asynchronously has become an important collaboration need. The main challenge here lies in the mismatch of expertise—legal and drafting expertise, policy expertise, and subject matter expertise. While this is a technology issue in part, it is mostly a people challenge, requiring participants to be willing to receive and accept the various perspectives. And, although not directly part of this chapter, the legislative process itself may need to be adjusted to take into account review from these various expertise positions to ensure that quality is maintained at the highest level.

Finally, data is essential when deciding where to focus modernization efforts and measuring their impact. The critical question is: what should be measured? Collecting data on various drafting processes, turnaround times, and other document metrics is one approach. To begin this process, one might ask about the pain points in the process, bottlenecks, performance, and resource allocation.

A Word on AI

As part of the review of the current systems and planning for the future, the role and capabilities of AI to improve the process will likely come up. When doing so, it's essential to follow the principle of "do no harm".. Be wary of the new shiny toy. Always follow the institution’s technology governance policies. Introducing a new technology requires the hard work of technological development and change management that goes with it. AI marketing suggests an implementation approach first and then figure out what went wrong. Be wary of proceeding this way. Development for the legislative process needs the opposite approach. There are no second chances. Project failure risks consequences for all who are associated with the project. The primary focus on using AI as part of a legislative tool kit is to assist human judgment, not replace it. AI is not a substitute for wisdom. However, there is a place for natural language processes (see the Comparative Print Suite above), machine learning, predictive analytics, and carefully chosen large language models. Some or all of these could be harnessed to increase efficiency by automating routine tasks, assisting with legal research, and improving draft quality.

Addressing Risk

While there is inherent risk in any project, there is a way to proceed. One can begin by analyzing the current system and identifying those pain points, bottlenecks, or areas to improve. In doing so, always keep in mind how the improved system fits into the overall framework. Simply moving a pain point or bottleneck to another part of the system may not be desirable. One should at least be working with the impacted part of the system as part of the project to minimize its negative impact.

Select a project and identify the goals to be achieved. Plan by beginning small, set benchmarks, incorporate iterative development methodologies with continuous feedback and testing and the big picture steps needed to reach the goal. Map all risks and work to minimize them throughout the project. Also. identify all stakeholders and their respective roles in the project, including users, clients and beneficiaries, and those providing resources for the development. Understand the resources that are available, including people, funds, and technical resources. Communicate with them early in the process and throughout using appropriate change management to keep all informed. Solicit their input and feedback. Address all concerns that arise. Celebrate small wins along the way, including rolling out a viable product as early as possible. Make liberal use of pilot groups and grow them over time. Don’t wait for direction - take initiative. Once the project begins, commit to it.

There are other steps to be taken to mitigate the risks of projects to modernize parliaments that impact legislative drafting. These are steps that apply to the people involved in the legislative process. Ensure that drafters remain relevant by being fully trained and remain up to date in legal developments and drafting best practices. Continuous professional development opportunities for drafters should be encouraged. Drafters should strive to develop and maintain relationships with their clients to promote collaboration throughout the drafting process and to address issues and concerns while they are small. The drafting office should work with parliamentarians and key policy staff to encourage legal and policy review of legislative drafts at each stage of the process. These reviews can be informal and ideally would simply become “this is the way we do things”.

Even the best planned modernization projects will face obstacles and these should be expected and planned for. Resistance by some to change, including refusing to participate in the project and actively undermining the effort. If attempts to clearly articulate the advantages of the project and work through the concerns do not resolve the obstacle, a person may need to be replaced, especially where that person is actively undermining the effort.

Other Constraints

All projects are subject to financial limitations. Accurate budgeting and tight project management are of primary importance to keeping the project on time and within budget. Pivoting to the next best alternative may be necessary. For this reason, prioritizing the most important components of the project to develop the viable product is important. After the project successfully delivers this product, then future enhancements can be delivered in regularly planned updates, subject to the availability of funds. Funding may become more accessible as users begin to see the product’s benefits and future improvements.

One should always expect to encounter technical difficulties during any modernization project. The hardware and software environment are complex. Developing, implementing, and maintaining additional complexity into an already complex environment will produce its own complexities. Particular attention should be paid to protecting sensitive legislative information from unauthorized access.

Conclusion

This chapter has provided a high-level view of change in the legislative drafting process through intentional modernization of information technology systems. In drawing this to a close, there are several matters to keep in mind. First, the legislative environment is people-based and highly relational. While technological developments can assist people in doing their jobs, it cannot replace them. Second, throughout any project, it is essential to keep the problem and its solution at the forefront.

With these two thoughts, creating a tool that effectively serves its intended users requires the participation of a user design team throughout the development process. Successful change in the legislative environment requires collaboration among many diverse entities in the House, subject matter experts, technologists, and contractors. All must adopt a shared understanding of the problem and solution and be committed to working together. Where possible, build on the foundation of prior projects and incorporate lessons learned from them.

It's also important to recognize that technology is constantly evolving, and processes are subject to ongoing change. Use software tools that are appropriate for the project and resist the temptation to use the latest "shiny toy". Always keep the users in mind, not just in the design of any tools developed but by providing comprehensive training and support.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Committee Guiding AI's Legislative Journey in the U.S. House of Representatives

Luís Kimaid — Fri, 07 Feb 2025 11:01:17 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

Introduction

The U.S. House of Representatives has embarked on a significant journey to integrate artificial intelligence (AI) into its legislative processes. Central to this journey has been the leadership and strategic guidance provided by key committees, particularly the Committee on House Administration (CHA) and its Subcommittee on Modernization. These bodies have played a pivotal role in ensuring that AI is not only incorporated effectively but also responsibly, setting a standard for legislative governance in the age of digital transformation.

Strategic Leadership and Vision

The journey of AI integration within the U.S. House of Representatives was marked by the strategic foresight of the Committee on House Administration. Recognising the potential impact of AI on legislative activities, the committee took a proactive approach to steer the direction of AI policy. Rather than merely reacting to technological advancements, they sought to anticipate the challenges and opportunities that AI presented.

This proactive stance was evident in the establishment of clear objectives for AI integration. The committee identified key areas where AI could enhance efficiency and accuracy in legislative work, such as in bill drafting, policy analysis, and constituent services. By setting these objectives early on, they provided a roadmap for AI implementation that aligned with the broader goals of the House.

Member Engagement and Political Support

The integration of AI into the House's legislative activities was significantly bolstered by the active involvement and political support of key members of the Committee on House Administration. Early in the process, the committee recognised the importance of involving its members, particularly those who were already engaged with modernisation efforts and AI policy, to ensure that the initiative would gain the necessary traction.

This strategic involvement was crucial. By engaging members from the Subcommittee on Modernisation, the committee was able to build a strong foundation of support. These members were not merely passive observers, they were strongly supportive of the AI governance framework, reviewing and approving plans before they were implemented.

Their active participation ensured that the AI strategies were aligned with broader legislative priorities and that the initiatives reflected the interests and concerns of the House. This alignment was essential for securing the ongoing support necessary for the successful implementation of AI policies. Furthermore, the engagement of these members helped to address potential resistance within the House by legitimising the AI initiatives through their endorsement and active involvement.

The committee’s approach also involved continuous engagement with these key members throughout the AI integration process. This ongoing dialogue allowed the committee to adapt its strategies as needed, ensuring that the policies remained relevant and effective in the face of evolving challenges and opportunities.

Building on a Legacy of IT Modernisation

Over the past two decades, the CHA has overseen significant technological shifts, beginning with the advent of the internet and the subsequent rise of mobile devices. Each of these milestones required a strategic approach to ensure that the House could effectively harness new technologies while maintaining its core functions and responsibilities. The committee's role in these earlier transformations provided invaluable experience in managing the complexities of technology adoption within a legislative context.

For instance, during the rise of mobile technology, the CHA was instrumental in developing policies that balanced the need for modern communication tools with the security concerns inherent in a legislative environment. Likewise, the CHA played a similar role in supporting the House's adoption of cloud-based technologies, gaining valuable experience in developing policies that balanced the need for cutting-edge digital services with the security concerns inherent in cloud-based technologies. These experiences proved critical when the committee began exploring AI integration, as they allowed the CHA to anticipate the kinds of governance structures and oversight mechanisms necessary to guide the responsible use of AI.

As the committee embarked on the AI journey, it drew directly from these past experiences, applying lessons learned from previous IT initiatives to ensure that AI was implemented in a way that enhanced, rather than disrupted, the legislative process. This historical perspective was key in shaping the CHA’s approach to AI, allowing the committee to navigate the technological landscape with a steady hand and a clear vision.

Policy Development and Governance

One of the most critical roles played by the committee was the development of comprehensive governance frameworks. Understanding that AI’s potential must be balanced with robust safeguards, they prioritised the creation of policies that would govern AI’s use within the House. This involved extensive collaboration with various stakeholders, including external experts, legislative staff, and Members of Congress.

The committee was instrumental in drafting AI-specific policies that addressed crucial issues such as data security, ethical considerations, and transparency. These policies served as the foundation for AI use in the House, ensuring that any deployment of AI technologies was conducted within a well-defined and secure framework. Their focus on governance highlighted their commitment to maintaining the integrity of the legislative process while embracing innovation in a complicated and large ecosystem that relies on in-house IT systems and many external software providers.

Coordination and Oversight

Effective coordination across different legislative branch agencies and House offices was another hallmark of the committee's approach to guiding AI’s legislative journey. They facilitated regular communication and collaboration among various House offices and legislative agencies, ensuring that AI initiatives were aligned with the institution’s overall strategic goals.

In addition to coordination, the committee exercised rigorous oversight to monitor the implementation of AI policies. This oversight was not a one-time effort but a continuous process, with regular updates and reports requested from relevant agencies. By maintaining this level of scrutiny, they ensured that AI tools and systems were functioning as intended and that any potential risks were swiftly identified and mitigated.

Strategic Use of Flash Reports

An essential aspect of the CHA strategy in guiding AI’s legislative journey was the strategic use of flash reports. These reports were more than mere updates; they were instrumental in shaping the trajectory of AI integration within the U.S. House of Representatives. Beginning in 2023, the CHA initiated a series of flash reports that meticulously documented the progress, challenges, and evolving strategies related to AI deployment in the legislative process.

The flash reports served multiple purposes. Internally, they functioned as a robust tool for continuous oversight, allowing the CHA to track the implementation of AI across various House offices and agencies. By requiring regular updates and detailed accounts from these entities, the CHA ensured that AI initiatives were not only advancing but also adhering to the established governance frameworks.

Externally, these reports provided a transparent account of the House’s approach to AI, offering insights into how AI was being harnessed to enhance legislative functions while maintaining rigorous standards of security, ethics, and accountability. The publication of these reports reinforced the House’s commitment to transparency and accountability. By continuously reviewing the information and insights gathered in these reports, the committee was able to refine policies, address emerging challenges, and ensure that AI implementation remained aligned with the broader goals of the House.

Adaptation and Continuous Improvement

The committee was not content with static policies for AI integration in the House. Understanding that the fast-evolving nature of AI technology required a dynamic approach to governance, they established mechanisms for the continuous review and refinement of AI policies and practices.

This adaptive approach allowed them to respond to new developments in AI and to adjust the House’s strategies accordingly. For example, as new AI use cases emerged, the committee ensured that these were evaluated against the existing governance frameworks and that adjustments were made where necessary. This commitment to continuous improvement was crucial in keeping the House’s AI initiatives relevant and effective.

Agile Oversight as a Strategic Tool

Traditional oversight methods, while effective for many purposes, were not well-suited to the fast-evolving nature of AI. To address this challenge, the CHA employed a new approach known as "agile oversight."

Agile oversight was more than just a monitoring tool, it was a strategic framework that allowed the CHA to maintain continuous engagement with the various House entities involved in AI implementation. By requiring these entities to provide monthly updates, the committee created a structure that encouraged regular, incremental progress rather than one-time assessments. This approach ensured that issues could be identified and addressed in real-time, allowing the CHA to adapt its strategies quickly as new challenges or opportunities emerged.

This method of oversight was particularly effective in fostering a culture of accountability and collaboration. The monthly reporting requirement meant that House offices and agencies could not simply provide a one-time update and then proceed without further scrutiny. Instead, they were drawn into an ongoing dialogue with the committee, enabling the CHA to offer guidance and support. This continuous feedback loop was essential for ensuring that AI initiatives remained aligned with the House’s strategic goals and governance frameworks.

Moreover, agile oversight allowed the CHA to tailor its interventions based on the specific needs of each entity. By maintaining a flexible and responsive oversight structure, the committee could allocate its resources and attention where they were most needed, thereby maximising the effectiveness of its guidance.

Guidance Through Strategic Events and Workshops

A key aspect of the CHA's approach to AI integration was the strategic use of events and workshops. These gatherings were instrumental in fostering collaboration, sharing expertise, and developing a cohesive governance framework for AI within the U.S. House of Representatives.

The CHA organised events, such as the AI guardrails roundtable, which brought together House officials, external experts, and civil society representatives. These sessions provided a forum for addressing critical issues like human oversight and cybersecurity, ensuring that all stakeholders were aligned on the principles guiding AI use.

By structuring these events to include both broad discussions and focused, scenario-based exercises, the CHA helped demystify AI and build confidence among House staff and members. The collaborative nature of these workshops, which required multiple internal offices to work together to address disruptive AI scenarios, not only fostered a deeper understanding of AI's impact but also facilitated the establishment of AI guardrails that became foundational to the House’s AI policies.

Setting an Agile AI Oversight

Through their leadership, the Committee on House Administration within the U.S. House of Representatives has set an agile AI oversight. Their approach, characterised by strategic planning, comprehensive governance, effective coordination, and adaptive oversight, demonstrates how Parliaments can navigate the complexities of AI adoption. By guiding AI’s legislative journey with foresight and diligence, they have not only enhanced the legislative process but also ensured that it remains secure, transparent, and aligned with democratic values.

Conclusion

The Committee on House Administration's approach to integrating artificial intelligence into the legislative framework of the U.S. House of Representatives exemplifies a meticulous balance between innovation and institutional integrity. By systematically addressing the challenges brought by AI through robust policy development, continuous oversight, and strategic coordination, the committee has effectively embedded AI into the legislative process while upholding the procedural rigour that defines legislative governance.

Central to this effort was the development of comprehensive governance frameworks that ensured AI’s application within the House was both secure and aligned with established legislative norms. The committee’s use of agile oversight techniques allowed for real-time adjustments and refinements, ensuring that AI tools were not only operationally effective but also ethically sound and legally compliant.

Moreover, the deployment of flash reports served as an interesting mechanism for both internal and external accountability, enabling the committee to maintain transparency in its AI initiatives while continuously refining its approach based on empirical data and feedback. This iterative process reflects a deep understanding of the legislative need for adaptability in governance, particularly as it relates to rapidly evolving technologies.

The Committee on House Administration's work in this domain sets a legislative precedent, offering a model for how parliaments can systematically approach the integration of disruptive technologies while maintaining the procedural safeguards necessary for parliamentary governance. As AI continues to evolve, the frameworks and strategies developed by the CHA will likely serve as a foundational reference for future legislative efforts, ensuring that technological innovation enhances, rather than undermines, the legislative process.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Unlocking Legal Knowledge with Multi-Layered Embedding-Based Retrieval

João Lima — Wed, 05 Feb 2025 11:01:33 GMT

This publication is part of the book “Artificial Intelligence in Legislative Services: Principles for Effective Implementation”. To download the entire book, use the button below:

This work addresses the challenge of capturing the complexities of legal knowledge by proposing a multi-layered embedding-based retrieval method for legal and legislative texts. Creating embeddings not only for individual articles but also for their components (paragraphs, clauses) and structural groupings (books, titles, chapters, etc), we seek to capture the subtleties of legal information through the use of dense vectors of embeddings, representing it at varying levels of granularity. Our method meets various information needs by allowing the Retrieval Augmented Generation system to provide accurate responses, whether for specific segments or entire sections, tailored to the user’s query. We explore the concepts of aboutness, semantic chunking, and inherent hierarchy within legal texts, arguing that this method enhances the legal information retrieval. Despite the focus being on Brazil’s legislative methods and the Brazilian Constitution, which follow a civil law tradition, our findings should in principle be applicable across different legal systems, including those adhering to common law traditions. Furthermore, the principles of the proposed method extend beyond the legal domain, offering valuable insights for organizing and retrieving information in any field characterized by information encoded in hierarchical text.

Introduction

The increasing volume and complexity of legal corpora pose significant challenges for legal professionals, including those in legislative consultancy, where the efficient access and analysis of legal texts are critical. Traditional keyword-based search methods often fall short in capturing the nuances of legal language and the intricate relationships within legal documents (Saravanan et al., 2009; Mimouni, et al. 2014).

Recent advancements in Generative Artificial Intelligence (GenAI) and Retrieval Augmented Generation (RAG) systems offer promising avenues for more efficient and accurate legal information retrieval. Embeddings, which are dense and compact vector representations of text, effectively capture the meanings of words, phrases, or documents (Yu & Dredze, 2015; Chersoni et al, 2021). Legislative documents, including bills and legal normative statutes, inherently have a hierarchical structure. This intrinsic hierarchy calls for an approach with variable granularity, capable of representing both smaller segments and broader groupings of legal texts through embeddings.

This chapter proposes a multi-layered embedding-based retrieval method that captures the semantic content of legal texts. By creating embeddings for articles, their components, and their structural groupings, we aim to provide a more nuanced and comprehensive representation of legal knowledge. Our proposed approach enables RAG models to respond to user queries with varying levels of detail, ranging from small portions to comprehensive sections.

Fundamental Concepts

Before moving forward with an exploration of our method and its applications, it’s essential to define the core concepts pivotal to this study: embeddings, aboutness, retrieval augmented generation, and chunking. These foundational elements form the base of our methodology. If you are already familiar with these terms, feel free to skip to the next section. Our goal is to ensure clarity in our discourse for IT analysts and stakeholders within the legal sector alike, who may not be well-versed in this specialized terminology. By clarifying these key concepts, we aim to eliminate any potential knowledge gaps.

Embeddings: At the heart of our approach lie embeddings, dense vector representations of text that encapsulate semantic meanings within documents. These vectors position words, sentences, or entire documents in a multidimensional space, where the spatial proximity between points signifies semantic similarity. This allows us to obtain a measure of the semantic distance between texts (words, sentences or documents). Distant vectors imply the texts that they stand for are dissimilar, while close vectors imply they are similar.

In the legal domain, the intricate subtleties among various terminologies are captured through embeddings, which can be visualized in a two-dimensional or three-dimensional space via a process that reduces the dimensionalities of the original embedding. Let’s examine an example using the fundamental concepts of Wesley Hohfeld’s framework.

Figure 1 provides the source text for each embedding, alongside the resulting vector representations. The textual descriptions define the legal terms ‘Right’, ‘Duty’, ‘Power’, and so on, outlining their correlatives and opposites. These definitions were transformed into dense numerical vectors, as shown in the last column, capturing the semantic essence of each term. The vectors, which are the outcomes of the embedding process, allow for the spatial mapping and analysis of these concepts in a multidimensional vector space.

Figure 1. Hohfeld’s Legal Positions Definitions and Corresponding Embedding Vectors.

Figure 2 represents these embeddings in three dimensions and shows the graph projection using the Plotly library. It was necessary to reduce the original 3,072 dimensions to 3 dimensions using the PACMAP algorithm. The first image shows the projection of legal positions, while the second image details the cosine similarity between these positions and between the positions and two specific questions. In the second image, green outlines show the semantic proximity between opposing and correlated legal positions. Similarly, outlines in blue indicate the close relationship between the two questions in the model and the legal positions they query. For example, for the question ‘What is Legal Immunity?’, the closest vector is ‘Immunity’.

Figure 2. 3D Map of Hohfeld’s Legal Positions and Similarity Matrix

Beyond capturing the contextual meaning of text, embeddings also reflect some syntactic features of documents. Texts that are semantically identical but presented with slight formal variations—such as differences in capitalization or line breaks—result in distinct yet close vectors. This attention to semantic and syntactic features shows how embeddings comprehensively capture a wide range of linguistic nuances.

Moreover, it’s important to challenge the conventional reliance on keyword-based retrieval methods such as TF/IDF (Ayed & Biskri, 2020) or BM25 (Huang & Huang, 2024), which primarily focus on the frequency of term occurrences. Unlike these traditional techniques, embeddings provide a more precise approach by concentrating on the semantic content of texts. This allows discerning the underlying meanings and themes, rather than merely counting words. This semantic focus makes embeddings particularly effective in complex fields like law, where exact terms may vary but the overarching legal concepts and relationships remain consistent.

Before exploring further into the other concepts, one might wonder about the intrinsic nature of an embedding. How does it capture and reflect the essence of the text it represents? This question leads us to a pivotal realization: embeddings can be seen as the digital “aboutness” of the content they encode.

Aboutness: The concept of aboutness refers to the main theme or subject matter addressed by a text. It encompasses entities, concepts, and relationships, and provides a comprehensive overview of its content. Identifying a document’s aboutness is necessary for pinpointing relevant information in response to specific queries, thereby facilitating the precise retrieval of legal information.

In examining the foundational concept of ‘aboutness’, it is valuable to reference Gilbert Ryle’s early analysis of the term in his article “About,” from Analysis in 1933. Ryle differentiates between interpretations of ‘aboutness’, particularly emphasizing the ‘about conversational’ aspect. He articulates that an expression captures the central theme of a conversation if a considerable number of sentences within that conversation either directly use that expression, employ a synonym, or make an indirect reference to the concept it denotes. This occurrence should dominate without a competing expression or referent being more prevalent across the dialogue. Ryle’s viewpoint suggests that ‘aboutness’ serves as a marker for the primary concern within a discourse, indicated through frequent references or mentions, rather than through the mere presence of specific keywords (Ryle, 1933).

This consideration of Ryle’s perspective aids in refining our approach to embedding legal documents. Recognizing ‘aboutness’ as more than a function of explicit text, but rather as a reflection of the thematic consistency across a document, informs our strategy. It acknowledges that the core theme or subject matter of a discourse is not necessarily tied to the explicit repetition of certain terms but is instead often conveyed through varied linguistic mechanisms that signal ongoing engagement with the theme.

Retrieval Augmented Generation (RAG) is a method that enhances response formulation by leveraging data as a foundational source. This technique capitalizes on both traditional access methods and the semantic prowess of embeddings, thus ensuring that the generated responses are not only relevant but also contextually rich. The framework is a structured approach divided into three main phases: Indexing, Retrieval, and Generation.

In the indexing phase, an extensive index is created using a variety of external sources. If employing traditional approaches, several steps are necessary for textual indexing, including processes like text normalization, which encompasses tokenization and stemming, along with the removal of stopwords, to refine the data for enhanced searchability. Alternatively, advanced approaches involve using pretrained language models to produce semantic vectors of texts (embeddings). In both approaches, it is essential to decide on the indexing unit , whether it be the entire document, uniform parts of the document, or other criteria.

The retrieval phase utilizes the index to find information relevant to user queries. In traditional methods, this process looks at term frequency and the presence of specific terms to rank documents. Common techniques such as TF/IDF or BM25 prioritize documents based on how frequently the search terms appear. Conversely, when using embeddings within the RAG framework, this phase includes selecting indexed items based on their semantic proximity to user questions, employing vector similarity techniques such as cosine similarity or Euclidean distance. This advanced approach allows evaluating the semantic nuances of queries, enhancing ranking by closely aligning results with user intentions and the context of the query. The mathematical representation of user queries through embeddings creates a certain language independence in that the same query expressed by similar words (synonyms or near-synonyms), or even in different languages, can achieve similar results.

The generation phase involves creating responses by combining the retrieved information (concatenated chunks as context) with the original user query. This process typically uses a large language model to ensure the response is contextually relevant and reflects the contextual retrieved data accurately.

To move from the broader framework of RAG to the specifics of the indexing phase, we must first focus on the technique of chunking, which plays a key role in organizing and processing text for retrieval.

Chunking: This process involves dividing text into smaller, manageable segments, such as sentences, paragraphs, or specific legal clauses. Segmenting complex documents into parts allows for a granular examination of each segment’s content and its contribution to the document’s overarching meaning, thus facilitating the analysis and processing of intricate legal texts.

Semantic chunking, as practiced in semchunk, offers a method for dividing text into smaller, semantically coherent segments, utilizing a recursive division technique that initially splits text using contextually meaningful separators. This process is repeated until all segments reach a specific size, ensuring each chunk represents a unique thematic or conceptual element. However, this approach overlooks the intrinsic hierarchical organization of legislative texts, which are structured in a complex hierarchy where the systematization of provisions is explicitly defined by the author. While traditional semantic chunking effectively creates manageable segments using only one layer, it overlooks the multitude of layers derived from the hierarchical organization. In other words, it treats all segments on the same level, disregarding the structured organization that defines their legal significance and their interrelations.

A clear grasp of the aforementioned concepts is essential before exploring the specific characteristics of articulated normative texts. The ensuing section will elaborate on how these principles are applied to represent the complex, hierarchical nature of legislative texts, enabling more effective retrieval of legal knowledge.

The Specificity of Legal and Legislative Information

A defining characteristic of legislative texts is an inherent hierarchical structure. It reflects the systematization of legal dispositions, ranging from broad legal areas to specific articles, clauses, and sub-clauses.

In structured normative texts such as constitutions, every detail has substantial semantic importance. Unlike ordinary language, where redundancy and informality are common, constitutions are carefully crafted documents where each word, phrase, and clause is deliberately chosen to convey specific legal meaning and effect. This precision requires a granular approach to information retrieval. Analyzing and representing legal knowledge at varying levels of detail – from individual clauses and sub-clauses to broader articles, chapters and titles – allows us to capture the full spectrum of meaning embedded within these texts.

Figure 3 depicts the organizational structure of legislative documents in Brazil, conforming to the guidelines of Complementary Law No. 95/1998. The highest level is the “Legal Norm” (Norma Jurídica), under which the Main Text (Texto Principal) is the primary body of the document. Optionally, one or more Annex (Anexo) may be included, indicated by a dashed line, suggesting its non-mandatory nature. It is important to note that the number of Annex can vary, ranging from zero to several, indicating the flexibility for including them.

The foundational unit of this framework is the “Article,” which consists of a mandatory “Lead paragraph” (“caput”) and, potentially, one or more “Paragraphs” (“Parágrafo”), indicated by dashed lines to denote their optionality. These “Paragraphs” elucidate details or exceptions pertaining to the “Lead paragraph.” If enumeration is necessary, the lead paragraph or paragraphs can branch into “Sections” (incisos), which may be further divided into “Items” (alíneas), and then into “Subitems” (item) for detailed specificity.

Simple statutes may consist of only articles, but complex statutes and codes group articles into progressively larger units: “Sections [group of Articles]” and “Subsections” are possible components of “Chapters,” which are compiled under “Titles.” These titles are then organized into “Books,” and multiple books can be part of a larger “Part,” reflecting the elaborate nature of legal codes.

Figure 3. Internal hierarchy of a Brazilian statute, detailing its constituent elements.

Figure 4 provides an overview of the hierarchical structure of Title I and Chapter I of Title II of the Brazilian Constitution. It visually represents how the constitutional text is organized, starting from the broadest divisions and narrowing down to specific provisions. Each “Article” serves as a foundational element, with the “Lead paragraph” (caput) stating the principal point. “Sole Paragraphs” (parágrafo único) provide further explanations or stipulations when an article has only one. For detailed enumeration within an article, “Sections” (incisos) break down the content into subsections, each identified by Roman numerals. These sections can further subdivide into “Items” (alíneas), denoted by lowercase letters, and 'Subitems' for the most specific details, though not depicted in this portion of the structure. This systematic approach allows for a clear and organized reading of the constitutional provisions.

Figure 4. Segment of the Hierarchical Structure of the Brazilian Constitution

Figure 5 is an extract from Title I of the Brazilian Constitution, highlighting its hierarchical structure using color-coded annotations. The Document Component layer indicates the beginning of the main text of the Brazilian Constitution, while the Basic Unit Hierarchy layer contains all the Titles, beginning with Title I. The Basic Unit layer delineates Article 1, detailing the core values of the federation such as sovereignty, citizenship, human dignity, social values of labor and free enterprise, and political pluralism. The Basic Unit Component layer focuses on the “Lead paragraph” (caput) of Article 1, which declares Brazil as an indivisible federation and defines its democratic, legal foundation. Lastly, the Enumeration layer specifically identifies enumeration elements like Section III (inciso) of the lead paragraph of Article 1, which underscores ‘human dignity’ as a fundamental principle.

Figure 5. Layers of the beginning of the Brazilian Constitution

Considering their fundamental role in systematizing legal norms and principles, “Articles” are the most suitable unit for initial semantic chunking within legislative texts. As illustrated in Figure 4, applying this approach to Title I of the Brazilian Constitution results in four distinct chunks, each representing an individual article. While this provides a solid basis for analysis, our proposed methodology, detailed in subsequent sections, seeks to further enhance this process. Generating a broader set of chunks for embedding allows us to create a more detailed and complete representation of legal knowledge, ultimately leading to greater accuracy and effectiveness in legal information retrieval.

Building on our understanding of the hierarchical nature of legislative texts, we now present our proposal for a multi-layered embedding-based retrieval method. This approach aims to capture the semantic content of legal documents at various levels of granularity, from entire documents to individual clauses. The following section will outline the specifics of our chunking methodology in the indexing phase and the strategies for filtering chunks in the retrieval phase.

Proposal for Utilizing a Multi-Layered Embedding-Based Retrieval

This section explores our proposed approach, offering recommendations that cover two RAG phases: chunking methodology during indexing and filtering chunks at retrieval.

Chunking methodology during RAG’s indexing phase

Our proposed solution defines the following layers for segmenting the textual content of a legislative or legal document:

Document Level: At the highest level, we generate one embedding for each document. This captures the overarching theme, purpose, and scope of the legal text. This comprehensive vector can then be used to perform automatic classification of the document, aligning it with relevant legal categories.
Document Component Level: This level concentrates on capturing the essence of document components, which can be systematically presented through articles (such as the main text) or depicted using various structures (like tables or unstructured text). Examples encompass the main text and additional texts components such as justifications (in bills) and schedules (in annexes), with each being assigned its own embedding to represent its distinct contribution to the overarching legal context.
Basic Unit Hierarchy Level: Embeddings are also generated for books, titles, chapters, and sections. This captures the broader themes and relationships between groups of articles.
Basic Unit Level: Each article, as a fundamental unit of legal text, receives its own embedding. This captures the specific legal issue addressed by the article and its core provisions.
Basic Unit Component Level: Further granularity is achieved by creating embeddings for the obligatory head or the paragraphs that compose each article. This allows a detailed understanding of the semantics in each legal provision.
Enumeration Level: Further granularity is achieved by creating embeddings for item that are part of an article head or a paragraph. This allows a detailed understanding of the specific details in each legal provision.

By incorporating embeddings at these different levels, we create a multi-layered representation of legal knowledge, enabling RAG models to respond to user queries with varying levels of detail.

To illustrate the granularity achieved by our proposed method, let us consider the example presented in Figure 4. A standard approach using articles as the embedding unit would segment this text into 4 chunks, corresponding to the 4 articles. However, our hierarchical approach yields a much richer representation. At the Document-Level, we have 1 chunk encompassing the entire text. Similarly, the Text Component-Level also produces 1 chunk, as it encompasses the main body of the text. The Article-Level retains the 4 chunks corresponding to the articles, but the Article Component-Level breaks these down into another 25 chunks: 4 lead paragraphs (caputs), 2 sole paragraphs, and 19 sub-sections (incisos). Finally, the Grouping-Level provides 1 chunk representing the single title. In total, our method generates 32 chunks, thus providing a significantly more detailed representation compared to the 4 chunks of the traditional approach.

Figure 5 not only showcases the structure of Title I of the Brazilian Constitution but also reveals an insightful detail regarding Article Component-Level embedding within legal texts. Specifically, when it comes to enumerative elements such as ‘Sections’ (incisos), ‘Items’ (alíneas), and ‘Subitems’ (itens), it is not sufficient to consider their text in isolation. Since each element elaborates on a part of a larger context, it is essential to embed their text within the context of the superior elements up to the level of the ‘Lead paragraph’ (caput) of the article or the paragraph to which they belong. For instance, instead of isolating the phrase “a dignidade da pessoa humana” (the dignity of the human person), the embedded text should read as follows: “[A República Federativa do Brasil, formada pela união indissolúvel dos Estados e Municípios e do Distrito Federal, constitui-se em Estado democrático de direito e tem como fundamentos:] dignidade da pessoa humana”. This method of embedding ensures that the specific provisions are always understood within the full scope of their intended legal framework.

Filter strategy during RAG’s Retrieval phase

The strategy for filtering chunks in the retrieval system utilizes cosine similarity alongside additional parameters to ensure precise and contextually relevant responses. After computing the cosine similarities between the query embedding and each chunk's embedding within the dataset, the results are sorted based on similarity scores, prioritizing the highest values.

The selection process for the chunks includes several critical steps. Firstly, a baseline token count is set, which in this example is 2,500 tokens. This limit helps manage the total volume of text being considered, preventing overly lengthy or unmanageable responses. Additionally, a baseline percentage for similarity deviation, set here at 25%, is used to compare each chunk’s similarity against the highest found similarity. This comparison helps to filter out chunks that are significantly less similar to the query, maintaining a focus on those most closely related in semantic content.

Chunks are also filtered based on their textual boundaries to avoid overlap in the content presented. If a chunk’s textual range is already covered by another chunk with a higher similarity score, it is not included again, ensuring diversity and relevance in the returned results. This helps in maintaining a concise and focused response area. For instance, if “Art. 5, § 1” is retrieved, there is no need to retrieve “Art. 5, § 1, I” because the latter is encompassed by the former.

The final selection of chunks also considers the total number of tokens accumulated thus far against the baseline. After at least seven segments have been chosen, the process of selecting additional chunks continues until the total number of tokens exceeds the baseline threshold or the similarity falls below 25% of the highest similarity found, whichever comes first. This method balances detail and breadth, providing a comprehensive yet focused array of text for response generation and optimizing both the quality and efficiency of the information retrieval process.

To validate the effectiveness of our multi-layered embedding-based retrieval method, we conducted a comparative analysis against the traditional flat chunking approach. This analysis involved visualizing the embeddings and testing the retrieval results using a series of queries about the Brazilian Constitution. The following section presents the results of this comparative study, highlighting the advantages and potential limitations of our proposed method.

Comparative of Flat vs Multi-Layered Chunking Approaches

This section will present a two-dimensional projection visualization of embeddings and demonstrate a test implementation of RAG for the Brazilian Constitution. We will compare flat (divided by articles) and multi-layered chunking approaches. In the second part, we will test the retrieval results of both approaches using six queries about provisions of the Brazilian Constitution.

In this experiment, we employed the text-embedding-3-large model, configured to generate 256-dimensional vectors. To present the data in a comprehensible manner, we reduced these high-dimensional embeddings to a two-dimensional plane using the dimensionality reduction technique (PACMAP). The visualizations were then created using Plotly, the interactive graphing library, to facilitate a clearer understanding of the data structures and their semantic correlations.

Figure 6 showcases a visualization of article embeddings (traditional method of uniform fragments) of the provisions in Brazil’s Federal Constitution, displayed in a two-dimensional space for easier comprehension. Each point represents an article plotted according to its embedding in a simplified two-dimensional space. This offers a broad overview of how each article is positioned relative to others, potentially reflecting their semantic proximity and thematic linkages.

Figure 6. Article Embedding of the Brazilian Constitution (highlighting Article 2)

Figure 7 exemplifies our hierarchical embedding approach, using the same corpus of the Brazilian Constitution text. The visualization in Figure 7 illustrates the hierarchical embedding method's capacity to generate a multitude of proxies—2954 in total—that more precisely correspond to the semantic content of the legal text, unlike the traditional method which produces only 276. The visualization becomes significantly more intricate. It includes embeddings from various levels of granularity, as advocated in this paper. This comprehensive approach incorporates not only the articles but also their components such as paragraphs, clauses, and sub-clauses, as well as their groupings like titles and chapters.

Figure 7. The Hierarchical Embedding of the Brazilian Constitution (highlighting 4^th Paragraph of Article 14)

Figure 8 takes a close look at the Article 5 region, such as its numerous clauses and sub-clauses. Given that Article 5 contains over 70 such components, this figure demonstrates the increased density of points surrounding the Article 5 embedding. It illustrates the detailed semantic landscape that emerges when each subdivision is given a distinct representation, underscoring the sophistication of the proposed multi-layered embedding approach. In Figure 6, note that Article 5 is represented by just a single point, whereas in the hierarchical view, it resembles a cluster.

Figure 8. The Hierarchical Embedding of the Brazilian Constitution Article 5 region

When comparing Figures 6 and 8, it becomes evident that the latter, in addition to possessing the vector representing Article 5 in its entirety (purple), also contains dozens of other vectors representing more specific aspects of this article that deals with fundamental rights. This allows the RAG strategy to offer more relevant excerpts in a more concise and consequently, more economical manner, if the question should be answered based in fine grained provisions.

Comparative of Flat vs Multi-Layered Embedding-Based Retrieval

Table 1 presents the questions formulated for a simple comparative test of response retrieval and generation using Multi-Layer (ML) and Flat embedding approaches. In crafting these questions, we aimed to encompass both specific and comprehensive inquiries. Appendix A lists all the devised questions along with descriptive information about the selected chunks and an evaluation of each chunk's relevance to the response, graded from 'essential' through 'complementary' to 'unnecessary'. Additionally, we have included the corresponding responses from the generation phase, accompanied by a brief commentary on each response. In this comparative test, the "gpt-4-turbo-preview" model was used during the generation stage, limiting the output to 1000 tokens and setting the temperature at 0.3.

Table 1. Case Study Questions and Basic Statistics

Figure 9 presents a box plot with a comparative visualization of the maximum and minimum similarity scores achieved by two distinct techniques, namely Multi-Layer and Flat chunking approaches, across a series of analytical questions. Each box plot distinctly represents the interquartile range (IQR), medians, thus facilitating a clear overview of the distribution patterns and variance within the data.

Figure 9. Comparison of Max and Min Similarities for ML and Flat Approaches

Upon close examination of the box plot comparing maximum and minimum similarity scores for the Multi-Layer and Flat embedding approaches, it is evident that the Multi-Layer approach tends to exhibit a narrower range at the extremes. This observation is particularly significant as it suggests that the chunks selected using the Multi-Layer embedding are more semantically consistent and closely aligned with the user’s query embedding.

Figure 10 illustrates the classification results of chunks by relevance (color-coded) and also displays cosine similarity between the question and the chunks on the Y-axis. This visualization allows us to observe how quickly the similarity decreases in the sequence of returned chunks. Notably, only in two of the eight cases does the highest similarity chunk coincide between the two chunking methods.

The Multi-Layer technique shows a higher proportion of essential chunks (approximately 37.86%) compared to the Flat Embedding method, which accounts for only about 16.39% of essential chunks. This suggests that the Multi-Layer approach may be better at identifying critical content in the texts, which could be advantageous for applications requiring nuanced understanding of the document's content. Moreover, the Multi-Layer Embedding also has a lower proportion of complementary chunks (around 3.88%) compared to the Flat chunking method (approximately 8.20%). The Flat approach has a significantly higher proportion of chunks classified as unnecessary (about 75.41%) compared to the Multi-Layer Embedding, which has about 58.25% of chunks in this category.

Figure 10. Questions and Chunk Relevance by Chunking Method

A particular challenge noted in both methodologies, but more effectively managed in the multi-layered approach, is the semantic overload present in Article 5 of the Brazilian Constitution. Questions 6, 7, and 8 (see Appendix) highlight this issue, where the flat embedding struggles to navigate the complex semantic layers within this article. The Article, which covers a broad spectrum of fundamental rights, requires the retrieval system to discern and prioritize among numerous embedded concepts.

Conclusion

The proposed multi-layered embedding-based retrieval approach offers a promising avenue for advancing legal information retrieval. By capturing the inherent structure and diverse granularities of legal knowledge, this method facilitates information access, ultimately empowering legal professionals and researchers in their pursuit of legal knowledge. The method’s ability to represent legal texts at different levels of granularity ensures that whether a query seeks an overview or a detailed examination, the retrieval system can provide the LLM with the relevant chunks as database source to create an answer that matches the user’s specific request.

The flexibility of question embeddings significantly enhances legal searching by permitting queries in plain language or even in different languages, narrowing the divide between legal practitioners and the general public. A legal expert might pose a question in technical terms such as ‘Where does power emanate from?’ (‘De onde emana o poder?’) or in Italian, 'Da dove viene il potere?'. Alternatively, a layperson asking ‘De onde vem a autoridade desse povo que manda?’ (Where do those in power get their authority ?) will receive an equally accurate response. This capability illustrates the system's semantic depth—it prioritizes the collective meaning over individual words.

The comparative evaluation between flat and multi-layered embedding-based retrieval methodologies reveals significant differences in the capacity to recognize and utilize essential text chunks. The flat method often fails to identify critical segments that are crucial for precisely answering queries. This limitation stems partly from its inability to appreciate the full semantic relevance due to the absence of broader contexts and additional layers of information.

In contrast, the multi-layered approach demonstrates a superior ability in identifying and selecting relevant passages, owing to its text analyses at various granularity levels. This not only enhances the accuracy of the responses but also reflects a deeper comprehension of the legal content, facilitating the retrieval of information that is genuinely pertinent to the user's inquiry. Moreover, unlike the flat approach, the multi-layered methodology effectively circumvents redundant or irrelevant chunks, which is critical for maintaining the efficiency of legal information retrieval systems.

In conclusion, the adoption of a Retrieval Augmented Generation (RAG) system with a multi-layered embedding approach can enhance the efficiency and accuracy of legislative consultations. This technology empowers consultants by enabling quicker access to pertinent legal information and facilitating informed decision-making in the drafting and legal analysis of legislation. By streamlining the retrieval and utilization of legal texts, it improves the quality and precision of legislative outputs, fostering a more effective and well-grounded legislative process.

As we look toward future enhancements and expansions of this research, several areas have been identified for further exploration and development.

Inter-Article Relationships: Investigate methods to represent relationships between articles, such as cross-references and dependencies. This could involve creating embeddings for pairs or groups of articles, capturing how they interact and inform each other.
Temporal Dimension: Consider incorporating a temporal dimension into the embeddings, reflecting the evolution of legal texts and amendments over time. This could involve creating embeddings for different versions of the legal text.
Vector Dimensions: During our proof-of-concept tests involving the Constitution we utilized 256 dimensions. Given that the adopted language model is capable of supporting up to 3072 dimensions, future investigations should assess whether increasing the number of dimensions can enhance performance even further. This exploration will help determine if the additional computational cost is justified by potentially improved accuracy and contextual sensitivity in the retrieval of complex legal documents, thus optimizing the trade-off between resource expenditure and retrieval efficacy.

The exploration of multi-layered embedding-based retrieval in legal texts not only enhances the precision of information access but also bridges the gap between complex legal terminologies and the general public’s understanding of them. As we continue to refine and expand upon this methodology, the potential for scalable, linguistically diverse, and contextually rich legal information systems seems increasingly attainable. Future research will undoubtedly unlock further capabilities, making legal knowledge more accessible and navigable for professionals and laypersons alike.

References

Ben Ayed, A.; Biskri, I. (2020). On the Relevance of Query Expansion Using Parallel Corpora and Word Embeddings to Boost Text Document Retrieval Precision. Computer Science and Information Technology, 9, 1-7.

Chersoni, E.; Santus, E.; Huang, C-R.; Lenci, A. (2021). Decoding Word Embeddings with Brain-Based Semantic Features’. Computational Linguistics, 47, 1-36.

De Oliveira Lima, J. A., Griffo, C., Almeida, J. P. A., Guizzardi, G., & Aranha, M. I. (2021). Casting The Light of the Theory of Opposition onto Hohfeld’s Fundamental Legal Concepts. Legal Theory, 27(1), 2–35.

Huang, Y., & Huang, J. (2024). A Survey on Retrieval-Augmented Text Generation for Large Language Models. arXiv preprint arXiv:2404.10981.

Mimouni, N.; Nazarenko, A.; Paul, È.; Salotti, S. (2014). ‘Towards Graph-based and Semantic Search in Legal Information Access Systems’. International Conference on Legal Knowledge and Information Systems.

Ryle, G. (1933). 'About'. Analysis, 1(2), 10-12.

Saravanan, M.; Ravindran, B.; & Raman, S. (2009). ‘Improving legal information retrieval using an ontological framework’. Artificial Intelligence and Law, 17, 101-124.

Yu, M.; Dredze, M. (2015). ‘Learning Composition Models for Phrase Embeddings’. Transactions of the Association for Computational Linguistics, 3, 227-242.

Appendix – Questions

Question 1. What are the foundations of the republic?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 2. Talk about the social function of property.

Key: E – Essential; C – Complementary; U – Unnecessary

Question 3. What are the attributes of the vote?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 4. How is tax revenue distributed?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 5. What are the rights of children and teenagers?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 6. What is assured to the jury?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 7. How is the right to association given?

Key: E – Essential; C – Complementary; U – Unnecessary

Question 8. How is legal assistance characterized for those with insufficient funds?

Key: E – Essential; C – Complementary; U – Unnecessary

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Access the House Resolution

House Resolution Guiding the Integration of Artificial Intelligence in U.S. Legislative Operations

Luís Kimaid — Thu, 30 Jan 2025 12:03:35 GMT

Introduction

The U.S. House of Representatives has a storied legacy of debate, decision, and democratic action. However, as technology advances, even the most time-honoured institutions must evolve. House Resolution 5 of the 119th Congress underscores this imperative by embedding within its provisions a commitment to integrating artificial intelligence and advancing technological capabilities within the U.S. House of Representatives. The inclusion of these elements within such a foundational legislative document signifies their growing significance in the administration of legislative operations and their capacity to enhance institutional functionality.

Artificial Intelligence and Institutional Priorities

The resolution explicitly mandates the Committee on House Administration, the Clerk, the Chief Administrative Officer, and other relevant officers to continue efforts to integrate artificial intelligence technologies into the operations and functions of the House (H. Res. 5, Sec. 3(j)). This directive aligns with the institutional priorities set forth in the House Information Technology Policy 8 (HITPOL 8), particularly in relation to AI integration. The emphasis on artificial intelligence within this document is not merely aspirational; it reflects a strategic commitment to harnessing AI’s potential to streamline parliamentary operational processes.

Artificial intelligence, as referenced in the resolution, is expected to contribute to administrative efficiencies by automating routine tasks, thereby allowing legislative personnel to allocate greater resources to substantive policy work. The resolution directs the House to explore AI applications in streamlining administrative processes, enhancing decision-making capabilities, and improving oversight through legislative drafting and analysis (H. Res. 5, Sec. 3(j)(2) and (3)).

Digital Accessibility and the Expansion of Machine-Readable Documents

Beyond AI, the resolution highlights the necessity of broadening the availability and utility of legislative documents in machine-readable formats (H. Res. 5, Sec. 3(k)). This initiative aligns with the overarching objective of improving public accessibility to legislative information and equipping House staff with enhanced tools to compare legislative proposals more effectively. By expanding the scope of machine-readable documents, the resolution facilitates transparency and accountability, enabling Members, analysts, and the public to track legislative changes with greater ease.

The requirement to enable comparative prints that display differences between versions of legislation and amendments to existing laws underscores the House’s commitment to legislative clarity and accessibility. The mandate for continued efforts in this domain (H. Res. 5, Sec. 3(k)(2)) reflects a structured approach to technological advancement in governance.

Enhancing Legislative Operations Through Digital Infrastructure

The resolution extends its modernisation efforts to improving the electronic document repository operated by the Clerk for use by House committees (H. Res. 5, Sec. 3(l)). This initiative reflects a recognition of the increasing reliance on digital infrastructure to manage legislative records efficiently. By streamlining cross-posting processes, particularly for the Committee on Rules, the resolution promotes a more integrated approach to legislative documentation. This measure is crucial in ensuring that crucial legislative information is more readily accessible, thereby strengthening the procedural integrity of legislative operations.

Such provisions ensure that votes, amendments, and witness disclosure forms are systematically organised and made available in a manner that enhances both legislative efficiency and public engagement. The resolution mandates improvements to the repository with the aim of increasing transparency in the legislative process (H. Res. 5, Sec. 3(l)).

Conclusion

House Resolution 5 establishes a legislative framework that acknowledges the indispensable role of artificial intelligence and digital modernisation in shaping the future of parliamentary governance. By incorporating these elements into the standing rules of the House, the resolution signifies a legislative commitment to leveraging technological advancements to enhance efficiency and transparency.

The importance of these provisions within the resolution extends beyond administrative improvements; they reflect an acknowledgment of the broader societal shift toward digital reliance. The resolution’s provisions set a precedent for future legislative modernisation, as has happened in the past with other technologies, ensuring that the House remains up to date with technological integration while upholding its commitment to democratic and effective parliamentary governance.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Access the Guidance at the Official Site

Safeguarding Parliamentary Integrity: Reflections on the United Kingdom’s “Artificial Intelligence Guidance for Members”

Luís Kimaid — Wed, 29 Jan 2025 11:02:45 GMT

Introduction

Released in January 2025 by the Parliamentary Digital Service, the Artificial Intelligence Guidance for Members addresses the growing presence of AI in the legislative work of the United Kingdom. Acknowledging that AI has advanced well beyond early automated tools, this guidance focuses specifically on generative AI and its newfound capacity to produce realistic text, images, and code. It balances the promise of AI, chiefly its ability to streamline various tasks, with sober reminders about how easily inaccuracies, data leaks, and ethical pitfalls can arise. This document, endorsed in its Preface by the Speaker of the House of Commons and the Lord Speaker, affirms that proactive understanding and careful usage of AI are now essential for modern parliamentary duties.

Understanding AI

Artificial intelligence, as described in the guidance, is ultimately about machines presenting the illusion of intelligence by predicting the most likely sequence of words or images, based on vast pools of data. Although AI itself is not a new concept, the capabilities of generative AI have recently accelerated. Complex language models such as chatbots can produce confident-sounding responses even when these are mistaken or fabricated (“hallucinations”). The guidance makes clear that such outputs can lead to serious repercussions if adopted uncritically for drafting correspondence, summarizing policy documents, or even assisting in committee work.

Because of these risks, the document offers a series of definitions to illuminate the underlying mechanisms of AI. It provides a glossary, from “deep learning” and “large language models” to “disinformation” and “deepfakes”, helping clarify how an AI system can both expedite daily tasks and inadvertently spread false information. The publication’s overarching message is that Members must look beyond AI’s ability to create polished outputs and also weigh the credibility of its sources.

Key Provisions and Practical Guidance

A central point in the document is the need to maintain human oversight. Even if an AI tool can rapidly summarize lengthy documents, the ultimate responsibility for assessing the accuracy and suitability of the generated material lies with the individual Member. This directive stems from the reality that generative AI tools sometimes fabricate seemingly precise facts or attributions without verifying their correctness.

The document also urges caution about data-sharing practices. Because generative AI tools commonly store user inputs on external servers, any sensitive or personal data provided to these tools could become vulnerable. Members are reminded that personal information related to their staff or constituents must be guarded in accordance with data protection law, and should not be input into tools whose data-handling practices are unclear. The guidance highlights the potential for inadvertent copyright infringement, since material produced by AI often draws on extensive and not always identifiable sources. It is therefore incumbent on Members to ensure that text or imagery generated by AI does not breach copyright or intellectual property requirements.

Furthermore, the publication addresses the need for Members to develop policies for AI usage among staff. Such policies could cover everything from how AI-generated text should be flagged for review to rules governing the input of privileged information. This is particularly relevant, given that an unauthorized AI tool could potentially access restricted parliamentary material, something that could compromise parliamentary privilege or risk broader data breaches.

Safeguarding parliamentary privilege

The guidance highlights that parliamentary privilege may be engaged in two ways when Members use AI. First, when AI-generated text is used as part of formal proceedings, Members remain fully accountable for what is produced and presented. The guidance underscores that no matter how sophisticated a tool may appear, the primary responsibility for the accuracy and appropriateness of content lies with the Member. Second, the document cautions that AI tools could conceivably access or store privileged information, such as unpublished committee papers. To prevent breaches of confidentiality, any tool that fails to protect such material may be removed to safeguard parliamentary security.

Steps Taken by the Houses’ Administrations

According to the guidance, an AI Working Group has been established to create a policy framework and provide training for those who wish to adopt AI in their work. This effort involves identifying appropriate AI applications and exploring new ways to bolster parliamentary functions, always in line with the institution’s standards. Members are invited to share feedback and ideas with the Digital and Data Skills Centre of Excellence, helping to shape a consistent and transparent approach to AI’s growing presence in parliamentary life.

Conclusion

The Artificial Intelligence Guidance for Members reflects a measured attempt to reconcile rapid technological changes with the traditional pillars of legislative responsibility. On one level, it highlights the growing influence of generative AI in everyday parliamentary tasks, inviting Members to harness its abilities for summarizing documents or drafting speeches. Yet it simultaneously underscores the risks that come with automation - particularly those posed by “hallucinations,” data leaks, and the uncertain provenance of AI-generated content. These concerns point to a fundamental tension: while digital tools can expedite research and communication, they also introduce new vulnerabilities.

Crucially, the text defines a framework where the human element cannot be sidelined. Even as the guidance acknowledges the productivity gains of AI, it repeatedly insists on individual accountability for verifying output and protecting confidential information. This stance extends to the concept of parliamentary privilege, which the guidance highlights the need for attention. By delineating clear lines of responsibility, the document attempts to keep the legislative process shielded from the distortions and breaches that unrestrained AI deployment might bring.

The creation of an AI Working Group and the invitation for Members to collaborate with the Digital and Data Skills Centre of Excellence demonstrate a commitment to ongoing adaptation rather than a one-time instruction manual. These steps show recognition that AI’s capabilities and the related ethical, legal, and procedural issues will continue to evolve.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.

Book about "Artificial Intelligence in Legislative Services: Principles for Effective Implementation"

Bússola Tech — Mon, 11 Nov 2024 13:49:14 GMT

Partners

Coordinator: Luís Kimaid, Executive Director at Bússola Tech.

Authors:

André Freire - Câmara dos Deputados of Brazil
Brahma Mishra - MeetMonk
Ciarán Doyle - Oireachtas of Ireland
Fotios Fitsilis - Hellenic Parliament
German Tarasewiez - Cámara de Diputados de la Nación of Argentina
Grant Vergottini - Xcential Legislative Technologies
James Pinnell - Commonwealth Parliamentary Association (CPA)
João Lima - Senado Federal of Brazil
Juan de Dios Cincunegui - Universidad Austral
Luís Kimaid - Bússola Tech
Moises Benamor - Organization of American States (OAS)
Monica Palmirani - Università di Bologna
Márcio Tancredi and Adriano Torres Ribeiro de Castro - Senado Federal of Brazil
Prince Kenneh - Parliament of Sierra Leone
Rodolfo Vaz - Câmara dos Deputados of Brazil
Wade Ballou - U.S. House of Representatives

Partners:

Using Artificial Intelligence in Legislative Services - Considerations from Society of the Clerks of the Table Africa Region Panel in Arusha/ Tanzania in May, 2024.

Bússola Tech — Fri, 09 Aug 2024 15:11:00 GMT

Authors: Luís Kimaid from Bússola Tech, Prince Kenneh from the Parliament of Sierra Leone and David Mulenga from the National Assembly of Zambia.

Reviewers: João Lima from the Senate of Brazil and Ari Hershowitz from Govable.

Summary

Artificial Intelligence (AI) offers transformative potential for parliamentary operations, much like the historical innovations of the printing press and the internet. AI can significantly enhance legislative services by automating tasks, supporting legislative drafting, and improving public engagement. Key opportunities include real-time transcription of parliamentary debates, drafting assistance, document digitisation, automated document distribution, and real-time translation. However, integrating AI also raises challenges related to accuracy, transparency, and compliance with procedural rules. Ensuring human oversight and continuous evaluation are crucial for responsible AI implementation. By carefully balancing innovation with democratic principles, parliaments can effectively harness AI to improve efficiency, accuracy, and public trust in legislative processes.

Introduction

Artificial Intelligence (AI) has evolved significantly and is now a powerful tool that can enhance the efficiency and effectiveness of parliamentary operations. Much like the introduction of the printing press in the 15th century revolutionised the distribution and accessibility of information, and the internet transformed global communication and information sharing, AI presents a new frontier for parliamentary innovation. The implementation of AI in parliaments offers unique opportunities for automating tasks, supporting legislative drafting, and improving public engagement. However, it also raises important questions about transparency, accountability, and maintaining democratic integrity. This document, summarising discussions at the Society of the Clerks of the Table, Africa region panel, provides a detailed overview of AI's role in legislative services and outlines key considerations for its implementation.

Opportunities for AI in Legislative Services

Hansard:
- AI-Driven Transcription: AI can transcribe parliamentary debates in real-time, ensuring accurate and timely records. It can also tag and categorise speeches, making it easier to search and reference.
- Speaker Identification: AI can automatically identify speakers in debates, improving the accuracy and organisation of transcripts.
Legislative Drafting:
- Drafting Assistance: AI can assist in drafting legislation by analysing existing laws and generating new bill drafts, ensuring consistency and comprehensiveness.
- Amendment Management: AI tools can manage and analyse legislative amendments, highlighting similarities and potential conflicts, thus streamlining the amendment process.
- Comparative Analysis: AI can compare different versions of bills and amendments, providing detailed reports on changes and their implications.
Historical Archives:
- Document Digitisation: AI can enhance the digitisation of historical legislative documents through Optical Character Recognition (OCR), making them searchable and accessible.
- Content Categorisation: AI can categorise and summarise historical documents, facilitating easier navigation and research.
Legislative Committees:
- Document Distribution: AI can automate the distribution of legislative documents to appropriate committees, reducing manual errors and biases.
- Meeting Summarisation: AI can summarise committee meetings, providing concise reports and action points.
Citizen Participation:
- Public Comment Analysis: AI can analyse public comments on legislative proposals, identifying key themes and sentiments, and providing summaries for Members of Parliament.
- Constituent Communication: AI can draft personalised responses to constituent inquiries, ensuring timely and relevant communication.
Parliamentary Website:
- Content Aggregation: AI can organise and categorise content on the parliamentary website, improving user navigation and information retrieval.
- Semantic Search: AI can enhance search capabilities on the website, providing more accurate and relevant results.
Legislative Research:
- Research Assistance: AI can assist in legislative research by analysing large volumes of data, identifying relevant information, and providing summaries.
- Data Analysis: AI can perform complex data analysis, uncovering patterns and insights to inform legislative decision-making.
Procedural Advice and Research:
- Real-Time Procedural Advice: AI can provide real-time procedural advice during legislative sessions, assisting with complex procedural queries and research tasks.
Plain Language:
- Simplification of Legislative Language: AI can simplify complex legislative language, making it more accessible to the public and to Members of Parliament, ensuring better understanding without losing legal precision.
Simultaneous Interpretation:
- Real-Time Translation: AI can provide real-time translation of parliamentary debates and discussions, facilitating better communication in multilingual parliaments.

Challenges and Democratic Considerations

Accuracy and Transparency:
- Accuracy: AI models can reflect existing inaccuracies in their training data or produce confabulations (hallucinations), leading to imprecise or incomplete responses. Ensuring high-quality data and incorporating human verification are essential to mitigate this risk.
- Data Grounding: AI applications should ensure that data is secure and used responsibly. Parliaments must establish clear protocols for data access, storage, and processing, respecting privacy and confidentiality.
- Transparency and Accountability: AI systems must be transparent, and their decision-making processes explainable, to maintain public trust and uphold democratic principles.
Compliance with the Rules of Procedure:
- Regulatory Compliance: AI systems must adhere to parliamentary rules of procedure and operational guidelines to ensure responsible and effective use.
Human Oversight and Collaboration:
- Human Review: AI should augment human capabilities rather than replace them. Human oversight is essential to ensure AI outputs are accurate and contextually appropriate. There should always be someone accountable for the review process.
- Stakeholder Engagement: Engaging diverse stakeholders, including AI experts, ethicists, legal professionals, and the public, is crucial in developing ethical and oversight mechanisms.
Security and Data Privacy
- Security: Parliaments must have a security system to ensure authenticity of the legislative data and procedures and its protection about the legislative process.
- Guidelines for the usage of AI: Draft guidelines document for the usage of commercially available tools, and how to interact with them, include policies about sensitive and personal information.

Implementing AI in Legislative Services

Modular and Integrated Solutions:
- Technological Integration: AI solutions should have a modular design but remain integrated with other parliamentary systems. For example, ensuring data compatibility between the drafting office and the Office of the Clerk.
- Iterative Development: Start with small, well-contained projects and gradually expand, taking an iterative approach to AI implementation.
Training and Capacity Building:
- Staff Training: Investing in training for parliamentary staff to work with AI technologies is vital for a smooth transition and effective implementation.
- Collaboration and Sharing: Parliaments should collaborate and share use cases, problems, and solutions to mitigate risks and leverage AI's benefits.
AI in Digital Services:
- Focus on User Satisfaction: AI strategies should enhance legislative services for parliamentarians, staff, and citizens, ensuring that technology facilitates rather than hinders parliamentary activities.
- Proactive Evaluation: Continuous evaluation and adjustment of AI systems are necessary to keep applications relevant, effective, and aligned with ethical standards. Parliaments should be attentive to the integration of AI into simple day-to-day services like email and scheduling, which will soon become standard and impact everyday operations.

Key Considerations

AI technologies hold significant potential to transform legislative services in parliaments, enhancing efficiency, accuracy, and public engagement. However, integrating AI requires a careful approach that balances technological innovation with procedural compliance and democratic integrity. By adopting a proactive and collaborative strategy, parliaments can leverage the transformative power of AI while maintaining transparency, accountability, and the effectiveness of democratic institutions.

The views expressed in this article are derived from the analysis of the author and do not necessarily reflect the official policy or position of the represented institutions, nor should they be considered and should not be construed as an endorsement or recommendation of any kind. The information presented in this article is derived from multiple sources. We encourage readers to access official sources from the institution in question.