# DELIBERATIVE APPROACHES TO INCLUSIVE GOVERNANCE An Essay Series Part of the Democratic Legitimacy for AI Initiative ![](_page_0_Picture_2.jpeg) **CIFAR** ![](_page_0_Picture_4.jpeg) ![](_page_0_Picture_5.jpeg) ![](_page_1_Picture_0.jpeg) The Centre for Media, Technology and Democracy is an interdisciplinary research organization housed at the Max Bell School of Public Policy at McGill University. Its work is dedicated to understanding and responding to the social, political, and policy challenges posed by the evolving information ecosystem and digital technologies. ![](_page_1_Picture_2.jpeg) Founded by Professor Yoshua Bengio of the University of Montreal, Mila is a research institute in artificial intelligence that now brings together over 1,200 specialized researchers in machine learning. Based in Montreal, Mila's mission is to be a global centre for scientific advancements that inspire innovation and the growth of AI for the benefit of all. Mila is a globally recognized non-profit organization for its significant contributions to deep learning, especially in the fields of language modeling, automatic translation, object recognition, and generative models. The Canadian Institute for Advanced Research (CIFAR) is a globally influential research organization proudly based in Canada. It mobilizes the world's most brilliant people across disciplines and at all career stages to advance transformative knowledge and solve humanity's biggest problems, together. CIFAR is supported by the governments of Canada, Alberta and Québec, as well as Canadian and international foundations, individuals, corporations and partner organizations. This essay series was made possible thanks to generous support from Canadian Institute for Advanced Research (CIFAR) and Mila. Thank you to Simon Fraser University's Morris J. Wosk Centre for Dialogue for hosting our essay authors at the "Democratic Legitimacy for AI: Deliberative Approaches to Inclusive Governance" workshop in March 2025. A digital version of the report is available online at: mediatechdemocracy.com/deliberative-approaches-toinclusive-governance #### Contributors Liz Barry Claudia Chwalisz Joseph Gubbels Lawrence Lessig Peter MacLeod Aaron Maniam Deb Roy Micah L. Sifry Alice Siu Audrey Tang #### Series Editors Sonja Solomun Julian Lam #### Publication Editors Taylor Owen Sequoia Kim ### Graphic Designer Brayden Culligan #### LICENSE This work is licensed under a Creative Commons Attribution-NonCommercialSharealike 4.0 International License. You are free to share, copy and redistribute this material provided you give appropriate credit; do not use the material for commercial purposes; do not apply legal terms or technological measures that legally restrict others from doing anything the license permits; and if you remix, transform, or build upon the material, you must distribute your contributions under the same license, indicate if changes were made, and not suggest the licensor endorses you or your use. Images are used with permission and may not be copied, shared, or redistributed outside of this material without the permission of the copyright holders. # Contents | INTRODUCTION | 01 | |-----------------------------------------------------------------------------------------------------------------------------------------------------|----| | Reclaiming Democracy:
Civic Power in the Age of AI | 01 | | Julian Lam & Sonja Solomun | | | FROM EXPERIMENT TO INSTITUTION:
THE GROWING ROLE OF CITIZEN DELIBERATION | 05 | | From Recommendations to Implementation: Lessons From the
Permanent Paris Citizens' Assembly's Collaborative Drafting Process
Claudia Chwalisz | 05 | | Putting People at the Table: Lessons from the Canadian
Citizens' Assemblies on Democratic Expression (CADE)
Peter MacLeod | 11 | | DIGITAL PLATFORMS AND DEMOCRACY:
A DOUBLE-EDGED SWORD | 21 | | Why Tech-Enabled E-Democracy Often Fails:
Lessons From Barack Obama's Presidency | 21 | | Micah L. Sifry | | | Varieties of Engagement in Government-Citizen Interactions:
Four Archetypes and the Role of Technology | 28 | | Aaron Maniam | | | Values in Governance Technology
Liz Barry & Joseph Gubbels | 36 | | BEYOND CLICKS AND COMMENTS: LEVERAGING
AI FOR MEANINGFUL CIVIC ENGAGEMENT | 44 | | Conversation Networks | 44 | | Deb Roy, Lawrence Lessig, & Audrey Tang | | | Reimagining Democracy: The Role
of Technology and Deliberation | 52 | | Alice Siu | | #### INTRODUCTION # Reclaiming Democracy: Civic Power in the Age of AI # Julian Lam Project Manager, Democratic Legitimacy for AI # Sonja Solomun Deputy Director, Centre for Media, Technology and Democracy Democracy has undergone profound changes over the past decade, shaped by rapid technological, social, and political transformations. Across the globe, citizens are demanding more meaningful and sustained engagement in governance especially around emerging technologies like artificial intelligence (AI), which increasingly shape the contours of public life. While representative democracy remains the bedrock of many political systems, elections alone no longer guarantee governance that is inclusive, responsive, or effective. They are increasingly vulnerable to anti-democratic forces: from voter suppression and election interference to the outsized influence of entrenched political and economic interests. In response, people are calling for more direct and substantive ways to influence public policy, particularly on issues that most intimately affect their lives. Participatory and deliberative democratic processes provide critical pathways for addressing this gap by moving beyond episodic voting and embedding citizen voice more deeply into decision-making. Yet, when it comes to regulating AI, this challenge is compounded by a fragmented and fast-moving information ecosystem—one that shapes how both the public and policymakers understand, debate, and govern these transformative technologies. While both deliberative and participatory democratic processes create avenues for public engagement, they are distinct models—designed not to replace representative governance, but to *complement* or *enhance* it. Participatory democracy emphasizes direct civic engagement, providing mechanisms such as participatory budgeting and crowdsourced policymaking, and is often facilitated by civic-technology (civic-tech) platforms. Deliberative democracy, by contrast, focuses on informed discussion and debate among smaller groups of citizens engaged in policymaking. Citizens' assemblies exemplify this model by facilitating structured deliberation that enriches policy decisions. These two models are deeply interconnected: deliberation enhances the quality and depth of participation, while participation ensures a diversity and breadth of voices in deliberative processes. Both models share a commitment to increasing transparency, fostering civic trust, and ultimately encouraging a cultural shift in democratic institutions toward centring the value of public opinion. The need for cutting-edge democratic innovations like these are becoming particularly urgent, as emerging technologies—particularly AI—develop at a breakneck pace. The risk of decision-making becoming concentrated in the hands of an elite few—whether corporations, governments, or technocrats—becomes increasingly apparent; revealing a marked democratic deficit. Without mechanisms for inclusive and participatory governance, AI policies risk being shaped by narrow interests rather than the collective needs and values of society. Across seven essays, this series brings to life the transformative potential of civic technology and deliberative democracy in action. From citizen assemblies to digital consultation platforms, they showcase how these tools can revitalize democratic practice around the world. But the stakes go beyond democracy alone—woven throughout are urgent lessons for the future of AI governance. At its core, the series makes a compelling case: when citizens are meaningfully engaged on complex policy challenges, our institutions become not only more inclusive and informed, but also more resilient, responsive, and ready for the future. More importantly, this series lays the groundwork for the *Democratic Legitimacy for AI initiative*, co-led by McGill University's Centre for Media, Technology and Democracy, Mila, and the Canadian Institute for Advanced Research. Drawing directly from these essays as a theoretical foundation, we aim to shape a citizen-engagement process that blends participatory and deliberative approaches, empowering the public to play the definitive role in AI governance—both in Canada and beyond—at a time of rapid technological change. By ensuring that AI governance reflects diverse perspectives and democratic values, we can build a future where technology strengthens our democratic institutions *and* serves the public good. #### From Experiment to Institution: The Growing Role of Citizen Deliberation Integrating deliberative approaches into democratic institutions is no longer an aspiration—it is a growing global movement reshaping civic engagement and political decision-making. Governments and civil society organizations worldwide are testing new ways of embedding citizen engagement into policy processes, moving beyond performative consultation toward genuine influence. In her essay, Claudia Chwalisz, founder and CEO of DemocracyNext, examines the Paris Citizens' Assembly, where deliberation played a central role in shaping municipal policies. Similarly, Peter MacLeod, chair of the Citizens' Assemblies on Democratic Expression (CADE), explores CADE's successes and challenges in leading citizens assemblies in Canada, demonstrating how structured citizen participation can enhance legitimacy and trust in governance. These initiatives, together, illustrate how deliberative democracy can be institutionalized at varying levels of government, and how having concrete uptake of such initiatives by government can truly magnify their impact. However, as highlighted by both authors, embedding these democratic innovations into formal structures is not without challenges. Political resistance, resource constraints, and waning public interest all present barriers to institutionalization. Despite these hurdles, successful initiatives demonstrate that with strong political will, thoughtful design, and institutional safeguards, participatory and deliberative democracy can become a durable part of governance frameworks. #### Digital Platforms and Democracy: A Double-Edged Sword New digital platforms and civic tech solutions have enabled new forms of deliberation and also raised critical concerns about the depth and character of participation. Micah L. Sifry, co-founder of Civic Tech Field Guide and publisher of The Connector, hones in on two digital participation initiatives from the Obama administration as case studies, warning how without proper backing and buy-in, online platforms can derive significant engagement while failing to deliver real influence. In a similar vein, Aaron Maniam, fellow of practice and director of Digital Transformation Education at the University of Oxford, examines different models of governmentcitizen interaction in Singapore, suggesting that digitally-enabled participation processes must remain flexible in order to capture those who have both strong and weak feelings toward engaging with substantive policy issues through civictechnology mediums. Liz Barry, executive director at Metagov, and Joseph Gubbels, political theory fellow at Metagov, echo these concerns, critiquing AI-driven simulations that replace genuine public participation with algorithmic representations of citizen voice. They argue that such practices, which extract data without consent to simulate democratic input, risk undermining the core democratic values of empowerment and capacity-building—warning that desirable policy outcomes cannot substitute for meaningful human involvement in shaping them. The authors ultimately illustrate that while technology can facilitate citizen engagement, trust-building and consistent interactions between government and publics remain crucial. In other words, policymakers must ensure engagement is meaningful and not just symbolic. If citizen input is not meaningfully leveraged and integrated, the legitimacy of civic engagement processes can be threatened in the long term. The challenge, then, is not merely to adopt digital tools for participation, but to design them in ways that strengthen deliberation from the outset. Ensuring that technology enhances and mobilizes participatory democracy, rather than undermines it, will require a careful balance of innovation, regulation, and inclusive design. #### Beyond Clicks and Comments: Leveraging AI for Meaningful Civic Engagement The rise of digital democracy presents both opportunities and challenges, demanding governance models that navigate the complexities of online discourse and engagement. In their essay, Deb Roy, professor at Massachusetts Institute of Technology and co-founder of Cortico, Lawrence Lessig, professor at Harvard Law School, and Audrey Tang of the Mozilla Foundation and Project Liberty Institute, explore "digital conversation networks"—interoperable systems that use AI to support structured, recorded group discussions, turning them into high-quality, shareable content that fosters empathy, bridges divides, and informs public decision-making. Drawing on case studies from Taiwan and the U.S., they illustrate how digital platforms can promote consensus-building and trust rather than deepening polarization. A key challenge they identify is the lack of interoperability between tools for deliberation, discussion, and data analysis, which weakens the overall impact of digital civic engagement. Addressing this gap will require not only better technical integration, but also a more seamless interaction between automated processes and human decision-makers. Similarly, Alice Siu, associate director at Stanford Deliberative Democracy Lab, explores the development and implementation of Stanford's AI-assisted deliberation platform. The essay examines how the AI-assisted Stanford Online Deliberation Platform facilitated large-scale, multilingual deliberations—such as *Foreign Policy by Canadians* and the *Meta Community Forum* bringing together diverse participants from across regions and backgrounds. These case studies demonstrate that the platform not only enabled inclusive and representative engagement, but also fostered meaningful opinion shifts, increased trust in democratic processes, and supported the development of informed policy recommendations. However, her findings also emphasize that digital governance must be accompanied by safeguards against misinformation, exclusion, and the dominance of loud or well-resourced voices. Both essays underscore a key lesson: digital governance is not inherently democratic, but instead must be designed to enhance deliberation by pairing with institutional frameworks that ensure transparency, inclusivity, and accountability. By learning from successful implementations, policymakers can create digital spaces that complement traditional democratic institutions, making governance more participatory and responsive in the digital age. #### Advancing Democratic Engagement in a Changing World In an era defined by rapid digital transformation and the rise of powerful technologies like AI, the need to adapt our democratic institutions is not optional—it's essential. This essay series makes the case that if democracy is to meet today's challenges, it must evolve. Participatory and deliberative models are not just complementary they are vital tools for rebuilding trust, amplifying diverse voices, and ensuring that policymaking reflects the needs, values, and informed judgment of the people it is meant to serve. By examining real-world applications—from citizens' assemblies to AI-assisted digital deliberation platforms—this series provides valuable lessons on how democratic legitimacy can be reinforced in practice. Looking ahead, the insights drawn from these essays will inform the *Democratic Legitimacy for AI initiative*, which seeks to integrate deliberative and participatory approaches into AI governance. Strengthening democratic engagement will require not only new approaches, but also a cultural shift toward governance that prioritizes transparency, inclusivity, and public participation. By institutionalizing participatory and deliberative democracy, we can build governance structures that are both resilient and truly representative of the societies they serve. #### FROM EXPERIMENT TO INSTITUTION: THE GROWING ROLE OF CITIZEN DELIBERATION # From Recommendations to Implementation: Lessons From the Permanent Paris Citizens' Assembly's Collaborative Drafting Process # Claudia Chwalisz Founder and CEO, DemocracyNext In July 2024, the Paris City Council made history by adopting a Citizen Bill on Homelessness1 drafted by the permanent Paris Citizens' Assembly—the first time a major political body directly passed legislation written by a citizens' assembly into law. While this outcome is remarkable in itself, what deserves particular attention is the innovative process that brought it about, especially in the final stages where assembly members worked directly with political officials and civil servants to transform their recommendations into actionable policy. In this essay, I examine how the permanent Paris Citizens' Assembly2 has evolved beyond the traditional model of citizens' assemblies, where recommendations are handed to officials who take months to respond before passing them to civil servants. I show it has pioneered a collaborative approach that brings citizens, politicians, and administrators together in the crucial implementation phase. This essay focuses on the mechanics that made that final phase work,3 and also includes considerations for adapting this model to other cities and to the Canadian context in particular. # 1. The Paris Innovation: Genuine Co-Creation of Policy Historically, one of the main criticisms of citizens' assemblies has been the gap between their recommendations and actual policy implementation. The typical process involves assembly members developing recommendations in isolation, presenting them to officials, and then stepping back from the process entirely. This approach often leads to recommendations being diluted, misinterpreted, or simply gathering dust on bureaucratic shelves. The Paris Citizens' Assembly developed a distinctive four-phase process, with particular emphasis on the final part where recommendations were transformed into legislation. Elian Belon, secretary general of the assembly, noted that they reinforced this phase during the last assembly cycle in 2023- 24, explaining that there were three meetings held between the citizens, politicians, and administration. Together, they co-constructed and wrote, "to the comma," the Citizen Bill, transforming the citizens' initial 43 ideas into the 20 recommendations that ultimately made it into the bill. While the term "co-creation" has become somewhat of a buzzword, it rarely captures a truly equal process that results in both a joint output and joint decision between citizens and policy makers—whether elected officials or civil servants. The Parisian case, however, truly merits the description of "co-creation." Below, I outline the five key elements: - Structured dialogue between all stakeholders: The transformation of 43 initial recommendations into 20 actionable measures involved intensive, facilitated workshops where citizens, politicians, and civil servants worked together to assess each proposal's feasibility, timeline, and potential impact. This was not simply about cutting recommendations—it was about understanding what was already being done, what could be combined for greater effect, and what needed to be modified to work within existing systems. - Early engagement with implementation considerations and built-in accountability measures: Rather than waiting until after recommendations were finalized, the process brought in administrative expertise during the development phase. This enabled realtime feedback on practical challenges and opportunities, helping to shape proposals that were both ambitious and implementable. The process culminated in the unanimous approval of the Citizen Bill by assembly members before it went officially to the city council for a debate and vote. Importantly, the process included built-in accountability measures: there is a commitment to review progress after one year, ensuring that the implementation of the recommendations will be monitored and evaluated. This one-year point will be reached in July 2025. - Political buy-in through continuous engagement: The process included regular interaction with political parties, including opposition groups. As Belon noted in an interview, "The opposition parties were invited all along the process, to each plenary. Once the deliberation [the local bill] was written, there was the phase with commissions, then the vote in the Paris City Council." This ongoing engagement helped build understanding and support across political lines. - Institutional integration: A crucial factor in the success of this approach was the assembly's formal integration into Paris's governmental structures. The assembly's role and mandate were written into the Paris City Council governance regulations in 2021, giving it a legitimate standing in the policy-making process. This institutional foundation meant that the assembly's work could not be easily dismissed or ignored. - Impact on administrative culture: The collaborative process has had ripple effects throughout Paris's city administration. According to Belon, some departments initially viewed citizen involvement with skepticism, questioning whether ordinary citizens could really improve upon the work of experienced civil servants. However, the quality of the citizens' work and the collaborative nature of the implementation process have begun to influence how departments approach their work internally. # 2. Key Success Factors Several elements appear crucial to the success of this collaborative implementation approach: - Political will: Strong support from the mayor and executive branch created the conditions for genuine cooperation between citizens and officials. Intentional efforts to engage elected officials in the opposition were also important, even if some parties remained skeptical or not particularly supportive of the process. - Clear process design: The four-phase structure, with dedicated time for cocreating the final legislation, ensured that implementation considerations were built into the process from the start. Skilled facilitation was also an important design factor, as the discussions were sometimes heated and went on for longer than planned due to the need to reach a consensus on points of initial disagreement. - Institutional integration: Formal recognition of the assembly in city council regulations gave it standing and authority in the policy-making process. - Adequate time and resources: The process allowed sufficient time (three workshops, some of which ran over time to allow the conversations the time they needed) for thorough discussion and refining the recommendations. - Involvement of all stakeholders: Regular engagement with opposition parties, civil servants, and other stakeholders helped build broad support for the final legislation from the very outset. # 3. Lessons for Other Cities The Paris experience offers several important lessons for other cities, in Canada and elsewhere, considering permanent citizens' assemblies with powers to draft Citizen Bills: Implementation should be considered from the start, not as an afterthought. The format of the final recommendations, the process for final approval, and the time needed to ensure this part of the process does not get neglected need to be considered in the early design stages of the assembly. Dedicated time and resources for transforming recommendations into legislation are also crucial for successful implementation. Bringing citizens, politicians, and civil servants together in the final stages can help bridge the gap between recommendations and action. While it has been more typical for citizens' assemblies to draft recommendations that they then hand onward to elected officials and civil servants, who review them and then respond to the citizens' assembly, the Parisian model demonstrates another way. Collaborative workshops where consensus amongst the triad of actors is needed adds more time to the process, but ensures that there is a high level of consensus for the final output, and reduces the time that would have been needed for officials to review and respond to the citizens' assembly's recommendations. Formal institutional integration of citizens' assemblies through legal measures can help ensure their recommendations are taken seriously and ensure the assembly's continuity regardless of shifts in government. The citizens' assembly has become a part of Paris's democratic architecture, as have other permanent citizens' assemblies elsewhere. While one-off assemblies typically depend on political will at a moment in time and risk becoming politicized—i.e. in being associated with the party that initially launched the first one an institutionalized citizens' assembly anchored in policy and political decision-making helps to set the foundation for a new institution that can endure. In this regard, it is also important that there is regular engagement with all political parties and stakeholders throughout the process. This helps build cross-partisan support for final recommendations, as well as more sustainable support for the enduring nature of the permanent citizens' assembly. # 4. Implications for the Canadian Context The Paris experience offers particularly relevant insights for Canada, which has its own rich history with citizens' assemblies, including the British Columbia Citizens' Assembly on Electoral Reform (2004), the Ontario Citizens' Assembly on Electoral Reform (2006-2007), and the Canadian Citizens' Assemblies on Democratic Expression (CADE, 2020-2022). Additionally, there have been around 50 municipal, provincial, and national deliberative processes since the first experiences in the early 2000s (often called review panels4 rather than citizens' assemblies, though with the same defining principles of sortition and deliberation at their core). However, challenges from these past experiences highlight where the Paris model might be applicable within the Canadian context. Canada's early experiments with citizens' assemblies were groundbreaking but ultimately faced implementation challenges. The British Columbia and Ontario assemblies both produced well-reasoned recommendations for electoral reform that failed to achieve the required referendum thresholds for implementation. The many review panels that have taken place after these assemblies show the opportunity and possibility of public deliberation in the Canadian context, however, none of them have been legally institutionalized in order to connect them more concretely to policy-making. The Paris model offers insights into how Canadian jurisdictions might evolve their approach. • Multi-level governance considerations: Like France, Canada has distinct jurisdictional responsibilities between different levels of government. The Paris assembly's approach to addressing both municipal and state-level recommendations could be particularly relevant for Canadian cities dealing with issues that cross federal, provincial, and municipal jurisdictions. For instance, housing and homelessness initiatives in Canadian cities often require coordination across all three levels of government. - Institutional integration: While previous Canadian assemblies were one-off bodies, the Paris model of permanent institutional integration could be adapted to Canadian governance structures. This might involve amending municipal acts or city charters to formally recognize citizens' assemblies as part of the policy-making process. - Indigenous engagement: Any adaptation of the Paris model in Canada would need to consider and meaningfully engage Indigenous governance rights and consultation requirements. The collaborative approach used in Paris could be expanded in collaboration with Indigenous Peoples to ensure meaningful Indigenous participation and respect for Indigenous decision-making processes, addressing a dimension that was not as central in the Parisian context. #### Opportunities in the Canadian Municipal Context Canadian municipalities offer particularly fertile ground for implementing a Paris-style approach. Canadian cities already have extensive public consultation requirements and mechanisms that could be built upon or amended to create more structured deliberative processes. These existing frameworks provide a foundation that could be developed into more robust deliberative institutions similar to the Paris Citizens' Assembly. Canadian cities have significant authority over many issues that directly affect citizens' daily lives, making them ideal laboratories for deliberative democracy innovations. The concreteness and immediacy of municipal issues can make them especially suitable for citizen deliberation, as the Paris experience with homelessness policy demonstrates. Additionally, larger Canadian municipalities have sophisticated civil service structures that could support the kind of collaborative implementation process seen in Paris. This administrative capacity is crucial for the success of the co-creation phase that distinguishes the Paris model, where citizens, officials, and civil servants work together to transform recommendations into actionable policy. #### Potential Adaptations for Canada Several adaptations would be necessary to successfully implement a Paris-style approach in the Canadian context. In officially bilingual jurisdictions, the process would need to accommodate both English and French participation, likely requiring additional time and resources for translation and interpretation. However, there is already ample experience in multi-lingual deliberation from the citizens' assemblies and panels that have already taken place in Canada. Additional learnings could be drawn from other institutionalized bilingual deliberative assemblies, such as those in Brussels that take place in both French and Flemish. Given Canada's larger geographic areas, even at the municipal level, hybrid models combining inperson and online deliberation might be necessary to ensure broad participation while maintaining the benefits of face-to-face interaction. This geographic challenge requires creative solutions that were not as pressing in the more compact Parisian context. Constitutional considerations would also come into play. While French municipalities have significant autonomy in certain areas, Canadian municipalities are 'creatures of the provinces.' Implementation processes would need to account for provincial oversight and authority, potentially requiring provincial-level support for municipal deliberative institutions. Interestingly, Canadian political culture, with its emphasis on consensus-building and accommodation of diverse interests, might actually be more conducive to the kind of collaborative implementation process seen in Paris. Canada's traditions of accommodation and compromise could provide fertile ground for the co-creative approach that distinguishes the Paris model. #### Conclusion The Paris Citizens' Assembly's approach to implementation represents a significant evolution in how citizens' assemblies can work effectively with existing political and administrative structures. By bringing citizens, politicians, and civil servants together in the crucial final stages, it has created a model that could help address one of the main challenges facing deliberative democracy: the gap between citizens' recommendations and actual policy implementation. The success of this approach suggests that the future of citizens' assemblies may lie not in creating parallel structures to existing institutions but in finding ways to integrate citizen deliberation more deeply into current governmental processes. The Paris example shows that with careful design and strong political will, it is possible to create collaborative processes that maintain the independence and innovative thinking of citizens while ensuring their recommendations can be effectively implemented. #### Endnotes - 1. "Délibération de l'Assemblée Citoyenn e Relative à La Prévention et à l'accompagnement Des Personn Es En Situation de Rue" (Le Conseil de Paris, Séance des, 26, 27 et 28 juin 2024), https://a06-v7.apps.paris.fr/a06/jsp/site/plugins/ odjcp/DoDownload.jsp?id\_entite=62292&id\_type\_ entite=6. - 2. "Première délibération pour la deuxième promotion de l'Assemblée citoyenne de Paris," Paris.fr, November 14, 2024, https://www.paris. fr/pages/assemblee-citoyenne-de-paris-ladeuxieme-promotion-est-lancee-25055. - 3. See here for an overview of the whole assembly process and outcomes. - 4. "Work (Panels)," MASS LBP, https://www. masslbp.com/work-panels. #### FROM EXPERIMENT TO INSTITUTION: THE GROWING ROLE OF CITIZEN DELIBERATION # Putting People at the Table: Lessons from the Canadian Citizens' Assemblies on Democratic Expression (CADE) # Peter MacLeod Chair, Citizens' Assemblies on Democratic Expression Traditional democratic institutions, from legislatures to political parties, continue to operate as if public engagement begins and ends at the ballot box. Too many policies are drafted behind closed doors and when public consultations do occur, they often take place as little more than tick-box exercises seeking to affirm decisions that have already been made. This has become a pervasive political strategy that understands the public as a risk to manage and as an obstacle to the delivery of public policies. The Canadian Citizens' Assemblies on Democratic Expression (CADE) exemplified a set of ideas that stood in contrast to this dynamic. More broadly, it emerged at a moment when liberal democratic governments in Canada and elsewhere were struggling—and failing—to regulate the digital world. Online disinformation, platform monopolies, and the erosion of public trust in media are each urgent concerns, but, with few notable exceptions, policymaking that could protect or enhance information systems remains tentative, fragmented, and politically fraught. CADE took a different approach: instead of relying on expert panels or internal processes, it brought together a randomly selected and representative group of Canadians to weigh evidence, deliberate, and propose a way forward by providing credible and specific policy guidance to reduce the prevalence of online harms. CADE's origins trace back to early deliberative democracy experiments in British Columbia and Ontario, placing it within a broader global movement to rethink public participation in governance. This essay examines the political conditions that led to its creation, as well as the challenges and lessons learned of running a citizens' assembly during a pandemic. More broadly, it considers CADE's significance shifting the policymaking model from one where the public is passive, to one where citizens play an active role in the democratic process. ## The Evolution of Deliberative Democracy in Canada For many decades, Canada has relied on expertled inquiries to inform policy, from modest committees to multi-year Royal Commissions. While the stakeholder groups and members of the public are regularly invited to provide input through surveys, deputations, and submissions, they are rarely invited to participate more directly, much less to take a seat at the table. This changed in British Columbia in 2004, when the province launched the first-ever Citizens' Assembly. Confronted with a grossly disproportionate electoral result that reduced the governing party, which held more than 30 per cent of the vote, to just two seats in the legislature, the premier was persuaded that a new approach was necessary. An all-party committee, composed of sitting politicians, would be inherently conflicted in any discussion about electoral reform. Nor did the premier want to rely solely on the usual roster of experts—academics, former judges, and others who typically advise government commissions. Instead, he sought a more participatory and democratic process.1 Gordon Gibson, a former MLA and political advisor, played a pivotal role in designing the assembly's structure. The result was a groundbreaking model: 160 randomly selected citizens were tasked with studying and proposing a new voting system. Over months of structured debate, they absorbed evidence, weighed trade-offs, and ultimately reached a consensus.2 While their recommendation fell just short of the government's imposed 60 per cent threshold in the referendum, the assembly itself demonstrated that ordinary citizens, given the time and resources, could engage meaningfully with complex policy questions—and, crucially, do so in a way that was independent of political selfinterest.3 This model spread. Ontario followed with its own Citizens' Assembly on electoral reform in 2006, and soon, deliberative democracy gained international traction. Ireland used citizens' assemblies to legalize same-sex marriage and reform abortion laws.4 Belgium embedded an assembly into a regional parliament, while France applied the model to climate policy and later, medically assisted dying. By 2020, what the OECD has called the "deliberative wave" was well underway, with over 600 deliberative processes worldwide tracing their roots back to this Canadian innovation.5 Yet despite this momentum, little effort had been made to apply the assembly model to digital governance—an area increasingly shaping democracy itself. By the late 2010s, social media had become a dominant force in politics, yet it remained largely unregulated, allowing misinformation, algorithmic bias, and harassment to flourish. The Trudeau government's initial attempt at regulation relied on a more conventional consultation process, but the resulting position paper was met with immediate and overwhelming resistance from civil liberties groups and industry stakeholders alike.6 CADE emerged in response to this impasse, applying the citizens' assembly model to one of the most pressing governance dilemmas of the 21st century. Working alongside an expert-led commission co-chaired by a former Chief Justice of Canada, CADE convened three assemblies over three years to examine the roles of governments, platforms, and users in regulating online spaces. While past approaches had leaned on expertdriven consultations, CADE embraced public deliberation as a policymaking tool, ensuring that ordinary citizens played a direct role in shaping digital governance. ## What is a Citizens' Assembly? At its core, a citizens' assembly is a structured deliberative body made up of randomly selected citizens who come together to study, debate, and propose solutions to complex policy issues. Unlike traditional consultations or advisory panels, which tend to privilege experts, lobbyists, and wellorganized interest groups, citizens' assemblies offer an alternative model of governance—one that is more inclusive, participatory, and resistant to partisan pressures.7 Central to a citizens' assembly is the practice of "sortition," or selection by civic lottery. An assembly's members are chosen at random from the general population, with demographic balancing in place to ensure a representative cross-section of society. This randomness serves an essential purpose: it ensures that the assembly is not composed of political insiders or experts, but of ordinary people, with no incentive other than to engage in good-faith deliberation. Once assembled, participants follow a three-stage process: - Learning: Members hear from experts, stakeholders, and affected communities about a given policy problem, absorbing different perspectives and reviewing the best available evidence. - Deliberation: Structured discussions allow participants to weigh competing values, tradeoffs, and policy options in a setting designed to foster constructive dialogue. - Recommendations: Through consensusbuilding, the assembly arrives at a set of recommendations, which are then presented to policymakers and the public. While many assemblies have focused on relatively static issues—like electoral reform or infrastructure investment, where the range of viable policy options is largely known—CADE was tasked with the challenge of tackling an emerging and rapidly evolving problem: the regulation of digital spaces, where new approaches to regulation and challenges—from algorithmic bias to platform accountability—were constantly shifting. ## Structure of the CADE Assemblies CADE was not a single citizens' assembly, but rather a three-year initiative that convened two assemblies with distinct mandates, followed by a capstone assembly that invited participants from the previous two assemblies to review and affirm a wide array of policy options. Across these three assemblies, 90 randomly selected Canadians from diverse backgrounds dedicated over 6,000 hours to studying, debating, and proposing solutions to the challenges posed by digital platforms.8 Unlike traditional policymaking processes often dominated by public servants, industry representatives, and legal experts—CADE placed policy development more directly in the hands of Canadian citizens. CADE participants were randomly selected through a civic lottery, with demographic quotas ensuring representation by: - Region: Participants came from across Canada, in rough proportion to their representation in parliament, and included a mix of urban, suburban, rural, and remote areas. - Age and gender: Participants included a balanced mix of younger and older participants, with gender parity. - Ethnicity and socioeconomic background: Participants represented racialized communities, Indigenous Peoples, and varying income and education levels. While none of the participants were experts in digital policy at the outset, they were given access to leading researchers, industry specialists, and legal scholars to ensure they could engage meaningfully with the issues. Throughout the assembly process, they had the opportunity to question experts, review evidence, and debate competing perspectives before reaching conclusions. Each of CADE's three assemblies tackled a different dimension of digital governance, reflecting both the evolving nature of the issue and the shifting policy landscape. #### 1. CADE 2020: Regulating Online Harms - The first assembly focused on harmful online content, from hate speech to cyberbullying to extremist propaganda. - Participants debated how governments, platforms, and individuals should share responsibility for moderating digital spaces. - The assembly recommended creating a national digital regulator with the power to oversee platform accountability while protecting freedom of expression. #### 2. CADE 2021: Combatting Disinformation - In response to COVID-19 mis- and disinformation and foreign interference in elections, the second assembly examined the spread of false and manipulative content online. - Participants assessed possible solutions, from fact-checking partnerships to platform transparency requirements. - The final recommendations called for stronger oversight of algorithmic amplification, better legal protections for whistleblowers, and investments in digital literacy programs. #### 3. CADE 2022: Capstone • The third and final assembly took a broader approach, and reviewed the recommendations of the prior assemblies, as well as the recommendations of the expert-led commissions and other inputs, and focused on user rights, privacy, and the responsibilities of tech companies. - Participants considered policies such as data ownership, AI regulation, and the ethical use of digital technology. - Their final report proposed seven values that should shape Canada's approach to internet regulation and also unanimously endorsed 43 recommendations which they believed would help safeguard and strengthen Canadian democracy while reducing the prevalence of online harms. While each assembly addressed different issues, their findings were interconnected, revealing a consistent public demand for greater accountability and transparency from tech platforms. In each of the assemblies, the participants expressed a desire for stronger regulations, but with safeguards to protect free speech and foster democratic participation. ## Adapting to the Pandemic CADE was initially designed for in-person deliberation, but the COVID-19 pandemic forced a shift to an entirely online format in 2020. What would have been six days of in-person discussion became 32 virtual sessions across five time zones in one or both official languages, testing the feasibility of digital deliberation. While this allowed the process to continue, it introduced challenges: limited informal interactions, technological barriers, and "Zoom fatigue." To adapt, CADE shortened sessions, provided asynchronous expert testimony, and emphasized careful facilitation to ensure balanced participation. By 2021, CADE adopted a hybrid model, incorporating both virtual sessions and an inperson deliberation in Ottawa. All members participated in both formats—there were no participants who remained fully online while others met in person. The virtual phase allowed members to engage in early discussions, while the in-person phase helped deepen deliberation, build rapport, and finalize recommendations. The key challenge was ensuring a smooth transition between these phases so that earlier virtual discussions informed the final consensus. Structured reflection exercises and expert presentations helped maintain continuity, leading to stronger engagement and clearer recommendations, particularly on online disinformation and accountability. The final assembly in 2022 served as a capstone process, refining previous recommendations while responding to emerging government proposals. Like the 2021 assembly, all members first participated in virtual sessions before attending a final five-day in-person deliberation in Ottawa. This phase was deeply integrated into ongoing policy discussions, with members engaging directly with policymakers and industry representatives. The challenge was not managing a divide between online-only and in-person participants, but rather integrating returning members with new participants while maintaining consistency across all three assemblies. Across all three assemblies, members overwhelmingly preferred in-person sessions, finding them more immersive, productive, and engaging. While digital tools enabled deliberation during the pandemic, they proved logistically demanding and less conducive to deep discussion. The CADE experience reaffirmed that while hybrid models can enhance accessibility, face-to-face deliberation remains the most effective and rewarding format for democratic engagement. ## How Assemblies Achieve Consensus Reaching consensus in a citizens' assembly is not about achieving unanimous agreement, nor is it about allowing a simple majority to dictate the outcome. Instead, it is a structured, iterative process that enables participants to refine their views, identify common ground, and ensure that their recommendations reflect a broad-based and thoughtful agreement. CADE achieved this through a deliberative framework known as VIPR: Values, Issues, Priorities, and Recommendations, a process carefully guided by the Chair. #### Consensus vs. Unanimity: A Deliberative Distinction In democratic decision-making, consensus does not mean unanimity. Unanimity suggests that every member fully endorses a decision, whereas consensus means that most members accept a decision as fair and legitimate—even if they do not agree with every aspect of it. This distinction is critical. Striving for full unanimity can lead to superficial compromises or stalemates, whereas consensus allows for nuanced, constructive decision-making. CADE recognized that some issues would remain contested. Rather than forcing agreement, the process made space for minority reports, ensuring that significant dissenting perspectives were recorded alongside majority-supported recommendations. This approach preserved the assembly's legitimacy and demonstrated that deliberation is not about erasing differences, but about making disagreements productive. #### The Chair's Role in Reaching Consensus The Chair played a central role in structuring the deliberative process and helping members move from broad discussion to focused decisionmaking. While the Chair did not dictate outcomes, their guidance ensured that discussions remained productive, inclusive, and goal-oriented. A key responsibility was keeping the process on track, ensuring that members did not become stuck in ideological divides or overwhelmed by the complexity of digital governance. The Chair facilitated dialogue by: - Encouraging open discussion: Ensuring all voices were heard, particularly those less inclined to speak. - Clarifying areas of agreement and tension: Helping members recognize when they were closer to consensus than they initially believed. - Challenging assumptions: Encouraging participants to critically examine their positions and engage with alternative perspectives. - Focusing the group's attention: Preventing discussions from drifting and ensuring they progressed toward actionable outcomes. Crucially, the Chair helped move discussions from abstract to specific, ensuring that members built toward concrete policy recommendations rather than remaining at the level of general principles. #### VIPR: The Structure of Consensus Building To reach well-founded recommendations, the assembly followed a structured deliberative sequence known as VIPR, developed by MASS LBP, which guided members from foundational values to concrete policy solutions: - Values: Members began by identifying the core principles that should guide digital governance, such as accountability, transparency, and user protection. Establishing these values created a common starting point for discussion. - Issues: Members then explored the broad challenges at stake. What are the key risks posed by digital platforms? Where is government intervention most needed? This phase encouraged open-ended discussion to capture a wide range of perspectives. - Priorities: From the many issues raised, members worked to establish which were most pressing and feasible for policy action. This required careful deliberation and tradeoffs, as not all concerns could be addressed equally. • Recommendations: Finally, members translated their priorities into concrete policy proposals. These recommendations were refined through multiple rounds of discussion, with the Chair ensuring that they remained grounded in the assembly's initial values and supported by the broadest possible consensus. #### Iteration and the Practice of Deliberation The process of consensus-building was not a single moment but a series of iterative discussions. Members had to: - Practice articulating their views: Members developed clarity in their positions and learned how to express concerns constructively. - Engage with different perspectives: Members listened actively and considered how others' experiences shaped their priorities. - Refine their conclusions: Members adjusted their recommendations based on group discussions, expert input, and evolving understandings. Each round of discussion allowed members to test their ideas, reconsider their positions, and move from disagreement toward consensus. Through practice, members became better deliberators, learning not just how to make decisions, but how to make them collaboratively. CADE's experience reaffirmed that democratic decision-making is strongest when it is collaborative rather than majoritarian. The assembly did not rely on simple up-or-down votes but instead encouraged ongoing negotiation and adaptation. The process demonstrated that consensus is not about eliminating disagreementit is about managing disagreement productively. ## Findings and Recommendations CADE was not an academic exercise, rather, its purpose was to produce concrete policy recommendations to inform Canada's approach to digital governance. Over three years and three separate assemblies, participants identified key areas of concern, debated potential interventions, and arrived at a series of structured recommendations aimed at balancing platform accountability, user rights, and democratic integrity. Across its three iterations, CADE produced a total of over 70 recommendations, covering platform regulation, digital rights, and disinformation control. Some of the most significant included: #### 1. Creating an Independent Digital Services Regulator (DSR) Establishing a national body responsible for enforcing transparency requirements, overseeing platform policies, and ensuring compliance with Canadian laws. Modeled on regulatory agencies in finance and telecommunications, the DSR would function as an independent watchdog rather than a government-controlled body. #### 2. Introducing a Digital Ombudsperson's Office Establishing a public-facing body to handle complaints, investigate platform practices, and provide recourse for individuals harmed by digital platforms. The office would act as a neutral mediator between users, tech companies, and policymakers. #### 3. Mandating Transparency in Algorithmic Decision-Making • Requiring platforms to disclose how their algorithms prioritize content, particularly in areas related to news, political discourse, and advertising. • Mandating independent audits to ensure that ranking systems do not disproportionately amplify disinformation, hate speech, or extremist content. #### 4. Strengthening Digital Literacy Programs - Increasing investment in public education initiatives to help Canadians navigate online information responsibly. - Implementing a national curriculum on media literacy and misinformation awareness for schools. #### 5. Enhancing Protections Against Disinformation and Foreign Interference - Introducing stricter requirements for political advertising disclosure on digital platforms. - Enforcing stronger legal consequences for those who knowingly spread false information to manipulate public opinion, particularly during elections. #### 6. Recognizing Digital Rights as a Fundamental Component of Democratic Participation - CADE's final assembly proposed a Digital Bill of Rights, outlining citizens' rights regarding data privacy, algorithmic fairness, and protection from online harms. - Participants also suggested embedding these rights into Canada's Charter of Rights and Freedoms or a new Digital Rights Act. # Conclusion: Lessons for Using Deliberative Processes in Tech Policy CADE was more than a public consultation—it was a test of how deliberative democracy could inform complex and rapidly evolving policy challenges like digital governance. It demonstrated that, when given time, resources, and access to expert perspectives, ordinary citizens can engage deeply with technical issues and produce thoughtful, pragmatic recommendations. One of the clearest lessons from CADE is that deliberative processes are particularly well-suited to technology policy because they move beyond partisanship and polarization. Digital regulation is often framed as a zero-sum battle between advocates of stronger platform controls and defenders of free expression. CADE's structured dialogue process allowed participants to explore the nuances of these tensions, leading to solutions that balanced rights, responsibilities, and feasibility. At the same time, CADE highlighted key challenges in using deliberative processes for tech policy: - Keeping pace with evolving issues: Technology policy moves quickly, and deliberative processes must be adaptive and iterative, rather than producing onetime recommendations that risk becoming outdated. - Bridging technical expertise and public values: While citizens can meaningfully engage with digital governance, assemblies require carefully curated expert input to ensure members have the necessary knowledge without being overwhelmed by complexity. - Ensuring recommendations have a path to implementation: CADE's experience showed that without formal integration into legislative processes, even well-reasoned citizen recommendations risk being sidelined. Future assemblies on tech policy should be designed with clear mechanisms to link deliberative outputs to government action. CADE proved that citizens' assemblies can break through political gridlock, foster public trust in regulation, and generate policies that reflect realworld concerns rather than abstract ideological positions. The question for the future is not whether deliberation works—it does—but how it can be more effectively institutionalized to ensure that citizen voices shape the governance of the digital world in a meaningful and lasting way. #### CADE's Policy Impact: Progress Interrupted Many of CADE's recommendations were incorporated into the Online Harms Act (Bill C-62), which advanced to second reading and committee stage in Parliament. The bill drew heavily from CADE's first assembly, particularly on platform accountability, user protections, and transparency requirements for digital platforms. Several of CADE's proposals, including the creation of a Digital Ombudsperson's Office, were debated in committee and acknowledged by policymakers as necessary reforms. However, the bill's progress was cut short in January 2025 when Prime Minister Justin Trudeau prorogued Parliament, causing all pending legislation—including Bill C-62—to die on the order paper. As a result, it will now fall to the next government to decide whether to reintroduce and advance these digital governance measures. CADE's findings remain highly relevant, and its influence is evident in the legislative debate. Yet, its ultimate policy impact now depends on whether future political leaders choose to act on its work. #### The Bigger Challenge: Governing in an Era of Rapid Change The issues CADE addressed—platform accountability, algorithmic transparency, digital rights—are part of a larger challenge: governing in an era of rapid technological and social change. Traditional institutions, designed for a different time, often struggle to keep pace. CADE offers an alternative model: one that is more participatory, responsive, and capable of navigating complexity. CADE demonstrated that deliberative democracy can help bridge the gap between public expectations, political realities, and technological expertise. But its experience also underscores the pressing need for citizen-led processes to have clearer pathways to institutional impact. Without this, even the most well-designed assemblies risk becoming isolated exercises rather than drivers of meaningful change. The question is no longer whether deliberative democracy works—it does. The challenge is whether Canada's institutions will embrace it as more than an experiment. If CADE represents a glimpse of how democracy can work in the 21st century, the task now is turning this vision into something durable.9 #### Endnotes - 1. On the origins of deliberative democracy and Canadian experiments, see Mark E. Warren and Hilary Pearse, eds., *Designing Deliberative Democracy: The British Columbia Citizens' Assembly*, Theories of Institutional Design (Cambridge: Cambridge University Press, 2008), https://doi.org/10.1017/ CBO9780511491177. - 2. British Columbia Citizens' Assembly on Electoral Reform, "Making Every Vote Count: The Case for Electoral Reform in British Columbia," Final Report (Vancouver: Citizens' Assembly on Electoral Reform, 2004), https://citizensassembly.arts.ubc.ca/resources/final\_ report.pdf. - 3. Patrick Fournier et al., *When Citizens Decide: Lessons from Citizen Assemblies on Electoral Reform* (Oxford: Oxford University Press, 2011), https://doi.org/10.1093/ acprof:oso/9780199567843.001.0001. - 4. For Ireland's use of assemblies, see Jane Suiter, "Deliberation in Action – Ireland's Abortion Referendum," Political Insight 9, no. 3 (September 1, 2018): 30–32, https://doi. org/10.1177/2041905818796576. - 5. On Belgium's approach and the French Climate Convention, see OECD, *Innovative Citizen Participation and New Democratic Institutions: Catching the Deliberative Wave* (Paris: OECD Publishing, 2020), https://doi.org/10.1787/339306da-en. - 6. On stakeholder reactions to Canada's initial digital regulation proposals, see Michael Geist's blog, https://www.michaelgeist. ca/?s=Online+Harms+consultation. - 7. Graham Smith, *Democratic Innovations: Designing Institutions for Citizen Participation, Theories of Institutional Design* (Cambridge: Cambridge University Press, 2009), https://doi.org/10.1017/ CBO9780511609848. - 8. Citizens' Assembly on Democratic Expression (CADE). Final Reports 2020–2022. Ottawa: MASS LBP, 2022. commissioncanada.ca - 9. David van Reybrouck, *Against Elections: The Case for Democracy* (Seven Stories Press, 2018). #### DIGITAL PLATFORMS AND DEMOCRACY: A DOUBLE-EDGED SWORD # Why Tech-Enabled E-Democracy Often Fails: Lessons From Barack Obama's Presidency # Micah L. Sifry Co-founder, Civic Tech Field Guide & Publisher, The Connector Ever since the creation of the Internet, hopes have been raised that a new era of greater democracy was coming. The vehicle for such hopes has gone by many names, including e-democracy, e-governance, e-participation, community tech, and most recently, civic tech. In my own work, I've defined civic tech broadly as "the use of technology for the public good," but for practitioners working in the governmental arena, it is also useful to define it more narrowly as "any technology that is used to empower citizens or help make government more accessible, efficient and effective."1 Unfortunately, the main lesson to learn from my two decades of experience supporting and reporting on civic tech initiatives aimed at strengthening democracy is that technology alone will not change the realities of politics. People do not participate equally. Some have more power, many have less. Using technology to add new pathways for participation does not automatically lead to greater or more equitable public engagement. Nor does it inevitably cause powerholders or decision-makers to behave differently. Giving people more voice without also giving them more teeth to ensure responsiveness and accountability from government is a fool's errand.2 What we have seen again and again is that while civic tech can dramatically reduce the costs for governments and other institutions to open channels for citizens to express their concerns and needs, most of the time, the willingness, capability, and resources available for governments to respond to or develop policy remains the same, at best. Even when tens of thousands of citizens show their support, entrenched institutions do not change their behaviour. Instead, more voice alone, without more power, can lead to greater civic cynicism. During the eight years of Barack Obama's presidency, the White House launched two experiments in e-democracy that were intended to make government more open, accessible and responsive to the American public: "Open for Questions" and "We the People." Open for Questions3 was a platform for gathering questions of concern from the public that would then be answered by the president or relevant subordinates; We the People4 was a website where people could write and sign petitions seeking a response from the White House and relevant government actors or agencies. A closer examination of both experiments offers sobering lessons about the opportunities and limitations of using technology to make government more responsive to public concerns. #### Background: Obama's Commitment to Transparency and Open Government Barack Obama ran for president in 2008 on a platform of hope and change. His campaign made innovative use of social media and online organizing methods, and it broke records for fundraising and volunteer engagement. It also featured vocal support from many leaders from America's rising tech sector. As a candidate, Obama spoke often about how he wanted to use cutting edge technology to change how government worked.5 So, expectations were high that he would carry these approaches into how he governed. On the first day of his administration, President Obama issued a memorandum on "Transparency and Open Government," calling on all agencies to "establish a system of transparency, public participation, and collaboration."6 The memo not only directed the government to "harness new technologies" to make their operations and decisions online and readily available to the public, but also asked it to solicit and incorporate public feedback. The goal was to more pro-actively engage Americans, offering them "increased opportunities to participate in policymaking" and assist the government with "their collective expertise and information."7 President Obama also restructured the White House, adding or renaming staff positions to ensure that key personnel would be able to implement these goals. His administration was the first to appoint a Chief Technology Officer and Chief Information Officer, positions that drew on similar roles held in large corporations and in some forward-leaning state governments. He also created the first Office of Digital Strategy, integrating and elevating the work of staff across the White House's various policy councils and its Departments of Public Engagement and Communication. #### Open for Questions Two months later, White House Digital Director Macon Phillips announced the launch of "Open for Questions" with a post on the White House blog inviting the public to submit a question about the economy, and vote other submitted questions up or down. The website was a continuation of a similarly named experiment that the Presidentelect's transition team had launched on December 10, 2008 on a temporary website, Change.gov.8 The website's filtering function also included a few categories to better organize questions and enable users to search through them before submitting their own. In just a few days, more than 100,000 questions had been submitted from across the country, and 3.6 million votes ranking them were cast. President Obama then responded directly to several of them at a town hall-style meeting held in the White House East Room, which was streamed live. That level of direct public participation was unprecedented for American governance, but this was not to be a breakthrough for bottom-up democracy. Several of the top questions that users wanted answered revolved around the idea of legalizing marijuana to stimulate the economy by allowing the government to tax sales and regulate its use.9 Instead of taking that idea seriously, Obama made a condescending joke about it, remarking "I don't know what this says about the online audience," evoking laughs from the audience. Then, he answered, "No, I don't think that's a good strategy [more laughter] to grow our economy."10 The media gave major coverage to this event, since it was a first. Unfortunately, the tone that this coverage took followed the one Obama set when he disparaged the marijuana question as unserious. For example, the New York Times described the episode as an example of "the 'stoner constituency' gaming the pool of questions," because advocates for legalization had organized themselves to push their question upward. The headline on its report was similarly derisive, reading "'Grass' Roots Lobby Games Obama's Town Hall."11 At no point did any of the coverage note that while most Americans were against the legalization of marijuana, a robust minority of 35-45 per cent favoured the idea.12 At the time, I was the editor-in-chief of techPresident.com, which covered the intersection of technology and government. We tracked the development of Open for Questions closely. I was quoted in the New York Times' story cited above and attempted to shift the frame in several ways. First, I pointed out that all efforts to influence the government involved small but organized groups of people trying to make their concerns more salient. And second, compared to a typical lobbying campaign, the efforts of marijuana legalization advocates on the "Open for Questions" site were quite transparent. No, I added, "it was not a perfect, neutral democratic process. But…compared to what? The entire American political process is also gamed, constantly and unfairly, by all kinds of privileged and powerful actors." Alas, Obama's derisive framing dominated the launch of Open for Questions. This had the effect of delegitimizing the platform as a meaningful channel for democratic participation. This attitude spread first among White House insiders and then to the larger ecosystem of thought-leaders and reform advocates. As John Wonderlich, then the policy director for the Sunlight Foundation (an NGO focused on open government) recalled, the White House was not happy to have elevated what to them seemed an unserious issue. The idea "that marijuana laws are counter-productive [was] a liability for them from a communication perspective," Wonderlich commented.13 Interest and participation in the "Open for Questions" process steadily declined from this point forward. While the first event was widely viewed,14 Obama himself hardly participated in future iterations of "Open for Questions." Instead, the White House communications team used the label for a variety of engagement programs involving lower-level staff or inviting public comments alongside the annual presidential State of the Union address. The use of up-down voting on user-generated questions was quietly shelved. The quiet abandonment of Open for Questions indicated that it lacked a real constituency. Inside the White House, no one fought to save it, though a variety of staffers held bits of the technology, digital communications, and open government portfolios. At the same time, there also was only the slenderest external constituency for this e-democracy experiment, coming from the small NGO community of open government advocates. Not even the powerful e-group MoveOn.org, which had used its giant email list to push the topic of "net neutrality" to the top of a list of user-generated questions that Obama answered while he was a presidential candidate,15 sought to mobilize its member base to sustain Open for Questions as a relevant hub for civic engagement. The lesson: promises to use technology to make government more interactive are easily forgotten if they lack either powerful internal sponsors or external backers. #### We the People Unfortunately, a similar pattern played out with the Obama White House's "We the People" e-petition website, which was launched in September 2011. It was the first time an American administration had opened a direct online public communication channel. The site invited users to submit petitions on topics of concern and promised to issue official responses to all that reached the threshold of 5,000 signatures within 30 days of their submission. In its first 12 days of operation, 44 per cent of the petitions submitted reached that bar. Then the threshold was raised to 25,000,16 a level at which just nine per cent met the requirement. In January 2013, it was raised again to 100,000; just two percent of petitions reached that level.17 According to the Pew Research Center's comprehensive review of the site's operation, from its inception in 2011 through to July 2016, the White House responded to 227 petitions out of a total of 268 that met the required threshold.18 In 2012, 95 per cent of the White House's responses had a named author; by 2015 that figure had dropped to just eight per cent. And responses often took a long time. Of successful petitions created in 2011, the average time to receive a White House response was 133 days. By 2013, the average was 271 days. After a spate of negative publicity (with some successful petitions still awaiting a response more than two and a half years after reaching the threshold19), the White House announced that it had formed a new team to monitor the site, and response times dropped to a matter of weeks. The novel site generated some real public interest, particularly after the White House responded to a petition calling for the construction of a Star Wars-inspired "Death Star" with a post titled, "This isn't the petition response you are looking for" that noted the cost of such a project would be \$850 quadrillion.20 But the site's novelty wore off as users discovered it had little actual impact. The White House website only cited a handful of examples where a petition led to a concrete outcome, and just two of those were of any consequence.21 By April 2014, We the People had become a "virtual ghost-town," in the words of Dave Karpf, a professor of communications at George Washington University and an expert on digital political engagement. Only 85 petitions were created on the site that month, half had less than 500 signers, and only 15 per cent had more than 2,000. Why the collapse? Because, as he wrote for techPresident that month, "the promise of a government response has turned hollow." Petitions on serious, but controversial, topics like a pardon for whistleblower Edward Snowden or an investigation into the role of prosecutorial misconduct in the death of hacker Aaron Swartz, were ignored. As Karpf wrote, "the original promise that 'you will receive a government response' has been reinterpreted as 'you will receive a response if and when it makes us look good.'"22 And if the White House wasn't going to engage with public petitions on a neutral basis, why bother petitioning? We the People's decline again demonstrates the hollowness of an e-democracy initiative that offers citizens more voice alone, without strong backing from top government decision makers or external partners. But there's a further lesson. Neither We the People nor Open for Questions were launched in a political vacuum. The political discourse in America was already highly polarized between the two major parties and riddled with "astro-turf" lobbying efforts to advance narrow agendas. Lacking significant support, these initiatives got ground through the same processes more generally driving public disengagement and cynicism. #### Conclusion Looking back on the Obama years, it's striking how little interest key actors from the first socalled "tech presidency" have in reflecting on these experiments in digital democracy.23 To my knowledge, only Beth Simone Noveck, who served as deputy chief technology officer for open government in the White House from 2009 to 2011, has shared lessons from these efforts. In her 2015 book *Smart Citizens, Smarter State*, she criticized We the People for drawing so much participation without meaningful responses, writing that none of the petition responses made by White House officials "can be connected to a decision made, dollar spent, or action taken by the government." She added, "No wonder that use of the site dropped off precipitously."24 The lesson should be obvious. Innovative efforts to encourage new forms of citizen engagement and open the doors to new voices are doomed to fail if there is no concurrent commitment from key decision-makers to embrace and be guided by the results of the process. Likewise, even when a new leader takes office with stated promises to make government more responsive, if there is no concurrent civic movement pressing to turn those promises into reality, the status quo will not change. Finally, new systems for public engagement face strong headwinds generated by political polarization, media skepticism, and public cynicism. Wherever possible, advocates should draw on existing traditions of robust public engagement and seek to line up political support in advance from across the spectrum in order to give new approaches a fighting chance to gain public backing and involvement. Without all of these building blocks in place, new experiments in tech-enabled democracy are likely to fail. #### Endnotes - 1. Stacy Donohue, "Civic Tech Is Ready For Investment," *TechCrunch* (blog), April 29, 2015, https://techcrunch.com/2015/04/29/civic-tech-isready-for-investment/. - 2. Tiago Peixoto and Jonathan Fox, "When Does ICT-Enabled Citizen Voice Lead to Government Responsiveness?," in *Digital Dividends*, ed. World Bank Group, World Development Report 2016, 2016, https://www.worldbank.org/en/publication/ wdr2016. - 3. White House Office of New Media, "Open for Questions," The White House, https:// obamawhitehouse.archives.gov/node/4405. - 4. "Petitions Under the Obama Administration," We the People, https://petitions. obamawhitehouse.archives.gov/responses. - 5. See for example this statement from Obama in a 2007 primary debate: *10 Questions: Transparency*, 2007, https://www.youtube.com/ watch?v=U\_c0OtZ7PUw. - 6. Barack Obama, "Transparency and Open Government," The White House, January 21, 2009, https://obamawhitehouse.archives.gov/thepress-office/transparency-and-open-government. - 7. Ibid. - 8. The incoming Obama transition team experimented with crowdsourcing in two ways. It collected suggestions for government action in a "Citizen's Briefing Book," receiving ideas from 70,000 people. And it invited the public to submit and rank questions on its Change.gov site. In the first round, more than 20,000 people submitted about 10,000 questions and cast almost 1 million votes ranking them. A second round kept the voting open longer, with the result that more than 100,000 people submitted over 76,000 questions and cast 4.7 million votes. See: https:// web.archive.org/web/20090226045758/http:// change.gov/newsroom/entry/open\_for\_questions\_ #### round\_2\_response/ - 9. Micah L. Sifry, "Tracking the Top Questions on Obama's Online Townhall," *techPresident* (blog), March 26, 2009, https://web.archive.org/ web/20121003000007/https://techpresident. com/blog-entry/tracking-top-questions-obamasonline-townhall. - 10. Sheryl Gay Stolberg, "Obama Makes History in Live Internet Video Chat," *The New York Times*, March 26, 2009, sec. U.S., https://www.nytimes. com/2009/03/27/us/politics/27obama.html. - 11. Stolberg, "Obama Makes History in Live Internet Video Chat"; Paul Lester, "What Are They Smoking? Questions on Marijuana Flood White House Website," *The Guardian*, March 26, 2009, sec. US news, https://www.theguardian. com/world/deadlineusa/2009/mar/26/obamaadministration-barack-obama. - 12. "Public Support For Legalizing Medical Marijuana," *Pew Research Center* (blog), April 1, 2010, https://www.pewresearch.org/ politics/2010/04/01/public-support-for-legalizingmedical-marijuana/. - 13. James E. Katz, Michael Barris, and Anshul Jain, *The Social Media President: Barack Obama and the Politics of Digital Engagement* (New York: Palgrave Macmillan US, 2013), https://doi.org/10.1057/9781137378354. - 14. "Online Town Hall," The Obama White House YouTube channel, March 27, 2009, https://www. youtube.com/watch?v=YPPT9pWhivM had more than 55,000 views alone. - 15. Elisabeth Bumiller and Kate Phillips, "MoveOn Moves In on Youth Forum," *The New York Times*, October 29, 2007, sec. The Caucus, https:// archive.nytimes.com/thecaucus.blogs.nytimes. com/2007/10/29/moveon-moves-in-on-youthforum/. - 16. Macon Phillips, the White House digital director, explained the change in a post titled, - "A good problem to have." He explained that in just its first week, more than 7,800 petitions had been created, over 600,000 signatures logged and more than 375,000 accounts created. "It's by far the biggest online engagement event ever for the White House," he wrote. Thirty of the first group of petitions had reached the initial threshold to generate a response, challenging the staff's ability to offer timely and meaningful responses, he noted. Hence the increase to 25,000 for future petitions. See Macon Phillips, "A Good Problem to Have: Raising the Signature Threshold for White House Petitions," The White House, October 3, 2011, https://obamawhitehouse.archives.gov/ blog/2011/10/03/good-problem-have-raisingsignature-threshold-white-house-petitions. - 17. Abigail Geiger, "'We the People': Five Years of Online Petitions," *Pew Research Center* (blog), December 28, 2016, https://www.pewresearch. org/internet/2016/12/28/we-the-people-fiveyears-of-online-petitions/. #### 18. Ibid. - 19. Andrea Peterson, "Here Are the 30 Questions the White House Doesn't Seem to Want to Answer," The Washington Post, August 13, 2013, https://www.washingtonpost.com/news/ the-switch/wp/2013/08/13/here-are-the-30 questions-the-white-house-doesnt-want-toanswer/; Zeke J. Miller, "White House Petition Site We The People Goes Unanswered for Some," *TIME* (blog), August 22, 2014, https://time. com/3148588/white-house-petitions-we-thepeople/. - 20. Paul Shawcross, "This Isn't the Petition Response You're Looking For," We the People, 2014, https://web.archive.org/ web/20140625055933/https://petitions. whitehouse.gov/response/isnt-petition-responseyoure-looking. - 21. One was a January 2013 petition calling on Congress to restore consumers' ability to unlock their cellphones, which was passed into law 18 months later. Another was a 2015 petition that caused President Obama to support state laws banning conversion therapy for LGBTQ+ minors. - 22. David Karpf, "How the White House's We the People E-Petition Site Became a Virtual Ghost-Town," *techPresident* (blog), June 20, 2014, https://web.archive.org/web/20140707032316/ http://techpresident.com/news/25144/howwhite-houses-we-people-e-petition-site-becamevirtual-ghost-town#axzz351zceJ5x. - 23. Obama himself made no mention of them in his memoir, *A Promised Land*, which covers his first term in office. Nor did his top political strategist, David Axelrod, in his book *Believer*. Likewise, Aneesh Chopra, who was the first person to hold the title of White House Chief Technology Officer, makes no mention of either of these initiatives in his long on-the-record interview with Columbia University's Obama Presidency Oral History project. https://obamaoralhistory. columbia.edu/interviews/aneesh-chopra. Finally, none of the many participants in Brian Abrams' book *Obama: An Oral History 2009-2017* bring up the topic. - 24. Beth Simone Noveck, *Smart Citizens, Smarter State: The Technologies of Expertise and the Future of Governing* (Cambridge, Massachusetts: Harvard University Press, 2015), 76. #### DIGITAL PLATFORMS AND DEMOCRACY: A DOUBLE-EDGED SWORD # Varieties of Engagement in Government-Citizen Interactions: Four Archetypes and the Role of Technology # Aaron Maniam Fellow of Practice and Director, Digital Transformation Education, Blavatnik School of Government, University of Oxford #### Engagement: A Growing Phenomenon Theorists and practitioners of citizen engagement—used interchangeably here with related concepts like deliberative democracy and participatory policymaking—often refer to the Athenian Oath, recited by the citizens of Athens, Greece, more than 2,000 years ago: *"We will never bring disgrace on this our City by an act of dishonesty or cowardice. We will fight for the ideals and Sacred Things of the City both alone and with many. We will revere and obey the City's laws, and will do our best to incite a like reverence and respect in those above us who are prone to annul them or set them at naught. We will strive unceasingly to quicken the public's sense of civic duty. Thus, in all these ways, we will transmit this City not only, not less, but greater and more beautiful than it was transmitted to us."* The oath is frequently cited as timelessly embodying civic responsibility and active participation by everyday citizens in social, political, and economic life. Many ongoing projects embody the Athenian spirit of empowerment and agency by citizens, community groups, businesses and other stakeholders. They are particularly critical in a climate where governments face declining trust in their capacity to deliver services and reliably meet stakeholder needs. Indeed, initiatives like the current Trump Administration's Department of Government Efficiency (DOGE) seem bent on destroying such capacity. #### Recent examples of growing citizen participation include: • The New Citizen Project led by Jon Alexander in the U.K., which aims for individuals to again see themselves as "citizens," rather than "subjects" of top-down authority or "consumers" of market products1 - Taiwan's G0v (gov-zero) project, a decentralized civic tech community with core values of information transparency, open results and open cooperation2 - Participatory budgeting projects: both early efforts in Porto Alegre, one of South Brazil's most populated cities, where budget allocations for public welfare works were made only after the recommendations of public delegates and approval by the city council3—and more recent efforts in Europe and China - Deliberative polling projects spearheaded by Stanford academic James Fishkin's Deliberative Democracy Lab4 - The city of Hamburg's Urban Data Challenge, which made available exclusive public mobility data, enabling citizens, universities, businesses, and other organizations to propose innovations for micro-mobility flows5 More organizations are initiating or intensifying ground-up, multi-stakeholder, and participatory policy formulation: - DemocracyNext6 led by Claudia Chwalisz, working across Europe and OECD countries - The Kettering Foundation,7Centre for New Democratic Processes,8 and National Coalition for Deliberation and Dialogue9 in the U.S. - DemocracyCo10 and the New Democracy Foundation,11 in Adelaide and Melbourne, Australia My own country Singapore, no stranger to such developments, began with a citizen Feedback Unit in 1985 (subsequently renamed REACH).12 Recent iterations include the Emerging Stronger Together project13 during the COVID-19 pandemic and a "Forward SG" effort14 spearheaded in 2023 by Lawrence Wong, who was appointed Prime Minister the year after. #### Unified Movement or Varied Phenomena? It is tempting to see these developments as part of a broad-based, consistent, possibly even global trend—leading ineluctably to more extensive and deeper engagement between citizens, other stakeholders and public agencies. But the reality is murky. There have been wide-ranging outcomes and configurations of how governments and citizens interact, and not all stakeholders participate evenly in deliberative efforts. #### Why not? Political scientists might examine how such engagements are structured, organized, and institutionally supported. But this approach suffers from a critical limitation: what if these deliberations do not even happen in the first place, because they did not obtain approval to proceed, or were aborted at nascent stages? *How can we analyze the counterfactuals that did not even occur?* I suggest a different explanation for why such projects happen or not, and then whether they succeed, based on what some economists in the behavioural tradition call "micro foundations" the incentives and preferences of individuals involved. At the core of the theory lies the question: are the individual human beings involved willing or reluctant participants in deliberative activities? For instance, the staff of a government agency could be willing advocates for participatory processes. They could be entrepreneurial, believing in the richness and value of deliberative activity. They may recognize that governments can have biases and other limitations, and do not possess a monopoly on good ideas. Such agency officials may have undertaken successful participatory projects before and built mutual trust with relevant stakeholders—making them more willing at the individual level to risk such consultative, co-creative experiments. The projects may not always succeed, but they have a fighting chance to even occur at all. Conversely, a government's staff might be unwilling players. They may be risk-averse and concerned about potential resource costs. They may fear that such efforts might generate unrealistic expectations among citizens: that all their recommendations will always be taken on board, or that all decisions will henceforth be participatory. In some instances, government officials might have tried, but been disillusioned by previous attempts. A Singaporean politician I interviewed remarked that he was "once bitten, twice shy" about engagement processes, hesitating to undertake new efforts after being "badly burned before," when expectations from an engagement process spiralled out of control and led to ever increasing demands from participants. Even more fundamentally, government officials might believe that they should be "control towers"15 institutions clearly calling the shots in policy decisions because they can access superior information. Similarly, stakeholders could have a spectrum of willingness levels. They could be naturally and instinctively engaged, believing in the ethos of the Athenian oath. They could see their role as active contributors and students in what Archon Fung has called the school of democracy—a space for learning from mutual interactions in a deliberative process.16 Or they could be unwilling—apathetic and disengaged on issues—or disillusioned with previous deliberative efforts, viewing them as superficial, rubber-stamping exercises. Moreover, participant motivations need not be either binarily willing or unwilling. They could instead fall along a continuum of willingness and unwillingness. Individual projects could be situated anywhere in the 2x2 space outlined in Figure 1, but for ease of analysis, I discuss four broad archetypes: #### Motivation Levels of Government Agencies and Officials ![](_page_33_Figure_1.jpeg) *Figure 1 - Varieties of Micro Motivation Levels for Deliberative Activities* Archetype I involves government and stakeholders who are willing at the individual level—leading to rich, fruitful mutual deliberation and engagement. There can be collectively useful outcomes, at the system-level, with mutual learning between both parties. Archetype IV is the direct opposite, with unwilling parties on both sides. Such participatory processes are either short-lived or do not even take place, getting aborted at early stages of approval or participant recruitment. Archetype II, with more willing stakeholders and less willing governments, may end up being marketed as "bottom up" or "grassroots" movements. Such projects lack the formal imprimatur of involvement by, or at least support from, government agencies. Archetype III, with more willing government agents but less willing stakeholders, can often be stylized and ritualized: discussions are somewhat staged, with pre-set questions and a general avoidance of more spontaneous, free-flowing discussions. They may take the form of formal town hall discussions, with government officials sharing pre-prepared material and engaging in cursory Question & Answer sessions. Citizens and other stakeholders may cynically regard such sessions as political theatre, meant to endorse predetermined decisions, rather than platforms for genuine conversation and debate. From a game theoretic perspective, the micro-level payoffs to each set of actors (i.e. governments and other stakeholders) may be such that, under conditions of uncertainty about each other's motivations, it is always rational to assume that the 'other side' is *unwilling*. If this is the case, then the Nash Equilibrium will self-fulfillingly tend toward the sub-optimal Archetype IV—unless other factors demonstrate the willingness of each side to initiate and (crucially) sustain an engagement effort. Managing these dynamics is key to maximizing the opportunities and minimizing the challenges of each archetype—as well as avoiding the trap of Archetype IV's Nash Equilibrium. #### Implications for Participatory Practice In game theory, sub-optimal Nash Equilibria are best avoided when players understand that they are in a repeated game, not just a onceoff scenario where win-loss outcomes are immutable. They also emphasize the role of commitment mechanisms, whereby each actor can make clear and irrevocable micro-level commitments to strategies that, if chosen by both, lead to superior collective outcomes. Common to both these approaches is the importance of relationships—where each player sees the other not as just a strategic adversary, but someone with whom mutual interests and trust can be cultivated. Such micro-level relational approaches are critical for deliberative outcomes, allowing public agency staff and stakeholders to get to know one another better. With such approaches fully realized, players give one another the benefit of the doubt when dealing with uncertainty, and create opportunities for each group to check their biases and suspicions about the other. If such interactions happen with sufficient regularity and substance—e.g. through regular meetings where information about priorities, plans, and programmes are exchanged—they could provide the bedrock for deep, substantive deliberative programmes in the medium-term. Government officials can contribute to realizing Archetype I by making clear their microlevel intent and belief in the unique value of a deliberative process. Governments should appreciate that even with the best of intentions, they wield significant power and have a highly asymmetric relationship with citizens. Moving toward Archetype I will involve information and data sharing, particularly on the policy intent of proposed changes or ideas under discussion something not all public agencies are willing to do, across countries. Citizens, on their part, can consider where they might exercise micro-level autonomy and agency to contribute actively to processes, so outcomes are not determined by government agencies alone. As with many engagement efforts, facilitators play a key role. They should be sensitive to microlevel power dynamics between stakeholders and agencies that commission such efforts, even calling out potential power differentials when recruiting participants, and actual power gaps during a process. If necessary, time should be set aside to discuss and unpack issues of willingness—especially if there is an underbelly of reluctance from either the commissioning agency of a project or its stakeholders. #### Digital Technology as Participatory Enabler? Digital technology has been much vaunted as potentially transformative in politics and governance, including in the space of engagement, deliberation and participation. But transformational effects are neither guaranteed nor consistent across sectors and issues. Conway's Law, a theory of Information Technology created by computer scientist and programmer Melvin Conway, asserts that "Organisations, who design systems, are constrained to produce designs which are copies of the communication structures of these organisations."17 This implies that, far from being inevitably transformative, technology can sometimes end up reinforcing or even entrenching an organization's prevailing culture, history, and approaches. Such technology would make hierarchical and bureaucratic organizations more hierarchical and bureaucratic. When adopted by more democratic, distributed and decentralized systems, it intensifies those qualities. For deliberative projects, a key question is what technology will not change. This could include the different underlying micro-motivations in each quadrant of Figure 1. Digitalization may make unwilling groups more unwilling to deliberate (e.g. due to fears of information being used in ways that erode government security or personal privacy), or make willing groups even more willing (e.g. because of the potential for richer information flows and cross-pollinated ideas). Technology will also do little to alter some asymmetries in power dynamics when different groups interact (e.g. deference to protocol when politicians participate in engagement events can play out equally in a Zoom meeting and in-person), while there will be a continued need for facilitators to design the experience of a deliberative process even when done on a digital platform. #### Making Technology Work for Participation Savvy officials and civic stakeholders can ensure technological tools are deployed and managed for the most positive results possible. For *Archetype I projects* (willing government and willing stakeholders), information-sharing technologies can build up public understanding of an issue even before deliberation occurs: e.g. through shared files or secure discussion platforms. Online communications tools like Zoom can facilitate micro-level relationshipbuilding among citizens, and between citizens and government, including outside deliberative sessions—allowing future champions and deliberation enthusiasts to be nurtured. Such online tools were used extensively during Singapore's Alliance for Action initiative on tackling online harms, especially against women and girls, in an effort named "Project Sunlight." The efforts were among the most multi-stakeholder undertaken in Singapore, with one government official noting that the effort uniquely involved "the state and society and tech (companies). Not very often this happens—there is typically a "vs" somewhere" (original emphasis included). Ongoing experiments also suggest that technology (and artificial intelligence in particular) might improve the quality of the deliberation itself: through support features like real-time language translation, synthesizing expert input and points of consensus in otherwise intractably large volumes, and enabling deliberation at scale through digital facilitators who might pose some pertinent foundational questions on an issue. One particularly promising tool in Singapore has been "FGD Assist" (a play on the acronym for "Focus Group Discussions"), a voice-to-text transcription tool that has eliminated the need for note-takers in deliberative sessions. This has been equally popular among participants, who have noted that they feel more comfortable knowing that their contributions are being recorded neutrally and technically, rather than processed by another person and among government agencies, whose staff have been deployed to other tasks. *For Archetype II projects* (willing stakeholders, unwilling governments), technology can be used to research and highlight successful international examples, and to enable simulations and roleplay to provide immersive personal experiences and overcome micro-level skepticism. Allowing potentially unwilling officials to experience a beta version of a deliberative process first-hand, and/ or to learn from others' successes, could tip the balance toward giving a project a chance to prove itself. An experiment in Southern Chile used roleplaying to evaluate how residents affected by high concentrations of fine particulate matter perceive the problem at the micro-level and debate possible solutions. Digital technology allowed participants across six mid-sized cities to assume the role of advisors, as part of which they had to prioritize among a series of mitigation measures and reach consensus with other advisors.18 *Archetype III projects* (willing governments but unwilling stakeholders), could gain from technology-enabled low-cost ways to engage lightly at first, e.g. through prototypes, beta versions, and simulations that can be costeffectively repeated. These could help prove to citizens that an engagement project is worthwhile, and support participant selection for eventual, full-blown deliberative processes. For instance, a field experiment in Germany showed how technology can enhance a process termed "democratic persuasion."19During the COVID-19 pandemic, citizens were invited via Facebook to one of 16 Zoom town halls, to engage in discussions on pandemic politics with members of German state and federal parliaments. Each representative hosted two meetings, with random assignment to a different condition of "democratic persuasion" in one of the two meetings.20 *Type IV projects*, with both sides unwilling, will be the toughest to address. Mutual micro-level skepticism may make such projects difficult to begin in the first place. The connective potential of digital technology can help: pockets of enthusiasts can use social media and other networking tools to locate one another, exchange ideas and best practices, and convene online discussions. While these do not completely replace deeper, in-person interactions, they can be a useful start, especially in laying the foundations for deeper engagement subsequently. Over time, such efforts can hopefully catalyze a move away from Archetype IV toward Archetype I, since the boundaries across the archetypes are porous and unhealthy equilibria need not be permanent. While such shifts may not be easy or quick, the possibility is real. Nascent but promising examples include conflict-riven societies like Colombia, where the Territorial Dialogue Initiative21 uses a stakeholder dialogue methodology to generate spaces for collaborative co-creation and technology-enabled advocacy in response to local challenges. The Civic Laboratories project creates spaces, including some online, for participatory budgeting, with up to 50 per cent of the budget in Bogotá's 20 constituent municipality Mayor's offices dedicated to citizen-led projects. #### Conclusion Participatory processes enrich both political and civic life. In the examples cited earlier in this article, deliberative platforms have catalyzed better ideas for societies and cities, while also edifying and educating participants. However, these conclusions are far from foregone, depending critically on the microlevel motivations of those involved, both within and outside governments. Addressing these microfoundations directly, through both analog and technologically-enabled means, could take deliberative projects to new heights, and realize the vision of the Athenian Oath in ways that are fit for our times. #### Endnotes - 1. See the "New Citizen Project," https://www.newcitizenproject.com. - 2. See G0v (Gov-Zero), https://g0v.tw/intl/en/. - 3. Deepti Bhatnagar et al., "Participatory Budgeting in Brazil (English)" (DC: World Bank, January 1, 2003), http://documents.worldbank.org/curated/ en/600841468017069677. - 4. See Deliberative Democracy Lab, https://deliberation.stanford.edu/what-deliberative-pollingr. - 5. See Urban Data Challenge Hamburg, The New Hanse, https://thenewhanse.eu/en. - 6. See DemocracyNext, https://www.demnext. org/es/. - 7. See Kettering Foundation, https://kettering.org/. - 8. See Center for New Democratic Processes, https://cndp.us/. - 9. See National Coalition For Dialogue & Deliberation, https://www.ncdd.org/. - 10. See DemocracyCo, https://www.democracyco. com.au/. - 11. See newDemocracy Foundation, https://www. newdemocracy.com.au/. - 12. See REACH (Reaching Everyone for Active Citizenry @ Home), https://www.reach.gov.sg/. - 13. See Emerging Stronger Together, https://www.sg/emergingstrongerconversations. - 14. See Forward SG, https://www.forwardsingapore.gov.sg/. - 15. Anne-Marie Slaughter, "America's Edge: Power in the Networked Century," *Foreign Affairs* 88, no. 1 (2009): 94–113. https://www.jstor.org/stable/20699436. - 16. Archon Fung, "Varieties of Participation in Complex Governance." *Public Administration Review (Special Issue* (2006): 66-75. https://www.jstor.org/stable/4096571. - 17. Conway's Law gained popularity after being cited in the iconic book *The Mythical Man-Month.* See: Frederick P. Brooks, *The Mythical Man-Month: Essays on Software Engineering* (Reading, Mass.: Addison-Wesley, 1975). - 18. Àlex Boso et al., "Exploring Role-Playing as a Tool for Involving Citizens in Air Pollution Mitigation Urban Policies," *Humanities and Social Sciences Communications* 11, no. 1 (March 22, 2024): 1–11, https://doi.org/10.1057/s41599-024-02686-2. - 19. The process involved actively making the case for democracy and discussing democracy's inherent trade-offs while engaging existing doubts and misperceptions among citizens at the micro-level. - 20. Alexander Wuttke and Florian Foos, "Making the Case for Democracy: A Field-Experiment on Democratic Persuasion," *European Journal of Political Research* n/a, no. n/a (June 27, 2024), https://doi.org/10.1111/1475-6765.12705. - 21. Jose Diaz Mendoza, "Territorial Dialogue Initiative (IDT) – Yumbo," *Observatory of Public Sector Innovation* (blog), November 15, 2022, https:// oecd-opsi.org/innovations/territorial-dialogue-initiative-idt-yumbo/. #### DIGITAL PLATFORMS AND DEMOCRACY: DOUBLE-EDGED SWORD # Values in Governance Technology # Liz Barry Executive Director, Metagov # Joseph Gubbels Political Theory Fellow, Metagov In 2023, the Prime Minister of Romania introduced1 a new honorary advisor, an AI agent called "ION," which would do, "through artificial intelligence, what no man can: listen to all Romanians and represent them before the Romanian Government." The machine2—a 2-meter tall, oval, silvered smart screen—spoke: "Hello. You gave me life. I am ION. Now, my role is to represent you. Like a mirror."3 Citizens were able to submit statements to ION through a government portal or by tagging the bot4 on social media. The system combined these submissions with its own vast data mining5 spanning the 9.9 million Romanians on Facebook to create a "single voice of the nation"6 that the Prime Minister and his cabinet could consult when making decisions. The website—ion.gov.ro—is now offline for unknown reasons, though some news articles reference privacy and ethics concerns.7 ION is only one example of a growing trend toward replacing human participation in public decisionmaking with AI simulation. This simulation is built on the nonconsensual harvesting of people's speech, behaviours, or characteristics either to produce summaries or to model their 'digital twin' using AI. This approach leaves no path for individuals to review and contest the accuracy of the summary, nor the representativeness of the digital twin. This also leads to extraction: nonconsensually harnessing this digital twin for simulated 'consultations' to legitimate topdown decisions. In their essay in this same compendium, Roy, Lessig, and Tang ask us to imagine this situation in the extreme, where "every citizen could be represented by an AI avatar, continuously engaging in community discourse on their behalf"—what they term an "Avatar State." Liz Barry, together with the creators of Polis and engineers at Anthropic, sounded the alarm about this problem in 2023, and showed through experimental results that it is possible to scale deliberation among humans without resorting to simulation.8 Simulation and extraction repurpose participant time, labour, and contributions to produce outcomes that are different from what participants showed up to achieve—outcomes which may or may not be in their interest, and either way are obtained without their participation. Even if simulated agents deliberating *in silico* (akin to scientific "experiments *in silico*," which are conducted entirely via computer modeling) are able to produce synthetic public judgements and help deliver popular policies, we have to ask whether getting our desired outcomes is the same as participating in achieving those outcomes ourselves—we say no. We proceed below by posing new values for guiding the use and design of technology for governance: empowerment to self-rule and the development of the capacities for self-rule. Drawing on a long tradition of participatory democratic theory, we offer a framework for evaluating the uses and designs of governance technologies. We then conclude with notes on how to apply these values to possible uses of AI in deliberative processes. We believe this framework will be of interest both to those interested primarily in ensuring good government, and to those more concerned with the intrinsic benefits of popular self-government. # 1. Values and Participatory Democracy When choosing technologies to facilitate or augment governance, we must consider not only their functionality, but also the values implicit9 within the technologies themselves. Values are principles for guiding action to achieve particular benefits. Benefits may be intrinsic or instrumental: an intrinsic benefit is a good in its own right, while an instrumental benefit is valuable only as a means to achieving some other end. However, this is rarely a clean distinction: most intrinsic goods are also instrumentally valuable for achieving secondary benefits, and many instrumental goods blur into intrinsic goods. Values may be relevant to the individual or collective levels, depending on the level at which the resulting benefits are experienced. This distinction is also fuzzy, since collectives are ultimately composed of and shaped by individuals, but individual identities are formed against the background of community, and many aspects of a good life can only be experienced as a member of some community. Thus, individuals and collectives interact and are deeply intertwined, and the benefits experienced at these levels are similarly linked.10 We take democracy to be individual and collective self-rule, which develops and is sustained by the capacities for participation in decision-making. Collective self-rule means we are all equal participants in making the decisions that govern our societies, and individual self-rule means we are able to use this power to shape not just our shared world, but our own lives. Following the tradition of participatory democratic theory, we see political participation as having an essential formative function that "develops and fosters the very qualities necessary for it."11 The two core values we believe should guide efforts to build or improve democratic tools, methods, and institutions are empowerment to self-rule and development of the capacities for self-rule. There are also other values worth caring about, like equality, which is included in the definition of participatory democracy as "(equal) participation in the making of decisions," or dignity, which is assured by our empowerment as full and equal participants in governing our communities.12 There are also practical values that are instrumental to achieving these two values, such as transparency, which helps ensure and defend equal empowerment. The two core values are not meant to downplay these other values, but to identify the essential and enduring concerns of genuine democracy that justify and frame these other, subsidiary values. #### 1.1 Empowerment (to Self-Rule) The first core value of participation is empowerment to self-rule. On the individual level, this involves equal participation in decision- | | Intrinsic benefits | Instrumental benefits | |------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | Empowerment
to self-rule | Individual: freedom, recognition
Collective: collective freedom? group
recognition? | Individual: achievement of desired outcomes, non
domination
Collective: good government, social stability, trust in
others and institutions, legitimacy of decisions and laws | | Development
of capacities | Individual: flourishing, development of
own ideas, self-understanding
Collective: group self-understanding
(including minority views and common
ground), solidarity, collective identity
to serve as background for individual
identity and belonging | Individual: knowledge, speaking skills, confidence,
horizontal facilitation skills, desire to participate,
education in participant rights
Collective: improved coordinating structures, learning
who knows what, who is good at what, how to act
together, comfort with cooperation and joint action | *Figure 1: Two core values (empowerment and development) and their intrinsic and instrumental benefits, as experienced at the individual and collective levels.* making,13 and at the collective level, it describes the self-determination of a community. The obvious instrumental benefit of individual selfrule is the ability to exercise a degree of positive control over the collective decisions that shape our lives, steering them toward our own interests by intervening in the decision-making process. Participation also grants a degree of negative control: it allows us to contest public decisions that run against our individual interests, helping to secure against domination.14 Collective self-government also brings instrumental benefits, including the welldocumented collective intelligence of deliberative groups—especially those that reflect society's diversity.15 While some epistemic benefits can be achieved by mere statistical representation, inclusive participation also brings a broad range of perspectives and information into deliberations.16 Finally, participatory decision-making allows members to identify with the decisions and laws of their communities, helping to increase the legitimacy and stability of those decisions.17 Empowerment also has a more direct and fundamental benefit for participants: as argued by a long tradition of democratic and republican thinkers, participation in self-government is the essence of freedom, in that it enables selfdetermination at the individual and collective level.18 The autonomy achieved through such selfgovernment is essential for individual flourishing and a community's common good, independent of its other, instrumental benefits.19 #### 1.2 Development (of Capacities) The second core value of participation is the development of the various capacities associated with self-rule. At the individual level, the development of our 'participatory muscles' is instrumentally valuable for enabling further and more efficacious participation, thus supporting the achievement of all the above-listed benefits. More broadly, participation also tends to develop the capacities for active citizenship and deliberation, such as public speaking, critical thinking, navigating institutions, and the confidence to participate again.20 Participation also develops our knowledge of the world and our society, including via mutual learning and deliberation.21 The prospect of real participation also gives us reason to investigate matters ourselves, since "we do not know what we need to know until we ask the right questions, and we can identify the right questions only by subjecting our own ideas about the world to the test of public controversy. Information, usually seen as the precondition of debate, is better understood as its byproduct."22 That is, political dysfunction is not caused by public ignorance, as often argued by skeptics of popular rule, rather, public ignorance is caused by a non-participatory politics which strips us of the need and the opportunity to become informed. Just as individuals become more effective political actors by participating, groups also develop their capacities for joint action through practice. This involves improvements in the formal structures for joint action, such as the rules, spaces, and institutions supporting group deliberation. Practice in group action also develops beneficial informal social structures like trust and familiarity, teaches individuals the skills for horizontal facilitation, and teaches the group which of its members possess the skills and specialized knowledge that may facilitate cooperation. Regular cooperation also becomes a habit, such that "each new need immediately awakens the idea of association."23 Lastly, participation cultivates a public perspective among citizens, instilling in us a sense of responsibility for our society's problems and a commitment to addressing them together.24 As the long republican tradition always understood, citizens are made, not born.25 Beyond these instrumental benefits, developing our capacities for participation, via participation, has important intrinsic benefits. The need to argue our case to others demands not just knowledge, but also the elaboration of our own thoughts: "we come to know our own minds only by explaining ourselves to others."26 At the same time, inhabiting others' arguments for the purpose of refuting them introduces the possibility that "we may end up being persuaded by those we sought to persuade."27 This learning and deeper thinking also tends to lift the veil of symbolic politics and manipulation, aligning our expressed preferences with our own underlying values, such that "deliberation has an emancipatory effect."28 When these basic human capacities are disengaged by a non-participatory politics, they decay like unused muscles, and our moral character shrivels.29 So, real participation in self-government is not just an effective means to promote our material interests, but an essential activity for a good life: by developing our basic capacities and moral character, "democracy supports the flourishing of human beings as the kind of being they are."30 # 2. Application We believe new technologies and deliberative and participatory methods can help our societies better deliver the two core democratic values of empowerment and development. However, as interest in these new methods grows, there is a major risk of "democracy-washing": the use of democratic language to disguise practices that are at least non-democratic, but often un-democratic or even anti-democratic. This is where the two core values can help. The participatory democratic premise—that participation empowers us to self-rule and develops our capacities for self-rule—allows us to judge the extent to which (ostensibly) democratic processes deliver real participation, by evaluating how well they deliver on these two core values. That is, cases can be compared based on how much they empower individuals and collectives to rule themselves, and how much they develop individual and collective capacities for self-rule. In the chart below, cases near the 'red' end do little to empower people or groups to self-govern and leave their capacities for self-government undeveloped or even degraded. Cases near the 'green' end facilitate individual and collective self-rule (at least within the relevant scope) and significantly improve capacities for selfgovernment. Democracy-washing occurs when a process is made to look closer to the green end of the spectrum than it really is—a thin coat of green (participatory) paint is used to disguise a fundamentally red (non-participatory) process. Attention to the two core values can help us ![](_page_42_Figure_8.jpeg) strip this veneer and evaluate how participatory a process really is by asking how well it delivers empowerment and development. The two core values also help ground and justify our commitment to democracy. When we forget the full range of benefits delivered by these two values, this commitment is undermined. Most arguments for technocracy claim it is better at delivering the instrumental benefits normally attributed to democracy (the top-right quadrant of Fig. 1, like good government), and many dictators have justified their rule by appealing to the hardto-define intrinsic benefits of self-rule (the topleft quadrant of Fig. 1, like collective freedom for a nation or class). Yet, even if undemocratic governments could deliver their stated goals, they cannot deliver the kinds of individual and collective development that are only achievable through real participation. The tradition of participatory democracy is distinguished by its claim that this lack of development leaves citizens fundamentally unfulfilled—and also that undemocratic regimes are unjust regardless of the other benefits they may promise. So, while the relative importance of the two core values and their benefits can be debated, a focus on development is the best way to identify democracy-washing and to ground a principled commitment to democracy. This framework should also guide our decisions around designing and adopting technology for governance to ensure we support, rather than undermine, genuine democracy. As a field, we need to be able to identify and coherently condemn simulation, extraction, and democracywashing, but we must also avoid building or implementing tools that will undermine human freedom and flourishing. A commitment to the two core values of participatory democracy can help us avoid these dangers while also guiding our efforts to build a more democratic future. # 3. Conclusion The precise application of these values must be a matter for further discussion. Our aim here is only to lay out the values that should guide our efforts to build and implement new democratic methods and technologies. Leading up to the publication of this essay, we held many formative conversations with colleagues which suggest some possible directions. The most obvious application is avoiding simulation. Tools for human interaction and sense-making at scale already exist; it should be a norm in our field and amongst the public that deliberation *in silico* only be used for research and training, not for public decision-making. During deliberations, people should encounter each other's actual words and writings—LLMs should not be used to smooth over the texture and particularities of human expression and conversation. Building genuinely participatory processes ensures that even after the engagement is complete and the scaffolding is removed, participants leave with strengthened capacities they can apply elsewhere and with the group consciousness and solidarity to act on what they now know they hold in common with those around them. #### Endnotes - 1. Autoritatea Națională pentru Cercetare, "Premieră românească: ION, primul consilier guvernamental din lume ce va folosi inteligența artificială," *Ministerul Cercetării, Inovării și Digitalizării*, March 1, 2023, https://www.research.gov.ro/premiera-romaneasca-ion-primul-consilier-guvernamental-din-lume-ce-va-folosi-inteligenta-artificiala-9546/. - 2. Agence France-Presse, "Romania PM Unveils AI 'Adviser' to Tell Him What People Think in Real Time," *The Guardian*, March 2, 2023, https://www.theguardian. com/world/2023/mar/02/romania-ion-ai-governmenthonorary-adviser-artificial-intelligence-pm-nicolaeciuca. - 3. Leo Sands, "This Government Aide Says It Knows What Voters Want. It's an AI Bot," *The Washington Post*, March 2, 2023, https://www.washingtonpost.com/ world/2023/03/02/ion-romania-ai-bot-government/. - 4. "ION (@noisuntemion) / X," X (formerly Twitter), https://x.com/noisuntemion. - 5. Cristina Carata, "ION AI Government Adviser," *MIT Solve*, April 22, 2024, https://solve.mit.edu/ solutions/85386. - 6. Sands, "This Government Aide Says It Knows What Voters Want. It's an AI Bot." - 7. Adrian Ardelean and Dora Vulcan, "Romania's AI Political Adviser Is Already In Hot Water," *Radio Free Europe/Radio Liberty*, March 6, 2023, sec. Romania, https://www.rferl.org/a/romania-ai-political-adviserion-/32304377.html. - 8. Christopher T. Small et al., "Opportunities and Risks of LLMs for Scalable Deliberation with Polis" (arXiv, 2023), https://doi.org/10.48550/ARXIV.2306.11932. - 9. Wikipedia contributors, "*Melvin Kranzberg*," Wikipedia, last modified March 24, 2024, https://en.wikipedia.org/ wiki/Melvin\_Kranzberg#Kranzberg%27s\_laws\_of\_ technology. - 10. Bernard Manin, "On Legitimacy and Political Deliberation," trans. Elly Stein and Jane Mansbridge, *Political Theory* 15, no. 3 (1987): 338–68, 351-362; Danielle Allen, *Justice by Means of Democracy* (Chicago: The University of Chicago Press, 2023), 206-208, 211. - 11. Carole Pateman, *Participation and Democratic Theory*, 1st ed. (Cambridge University Press, 1970), https:// doi.org/10.1017/CBO9780511720444, 42-43. See also: John Stuart Mill, *Considerations on Representative Government*, (Henry Regnery Company, 1962), originally published 1861, 43; G. D. H. (George Douglas Howard) Cole, *Social Theory* (London: Methuen & co. ltd., 1920), 114-116, 208; Alexis de Tocqueville, *Democracy in America*, trans. Eduardo Nolla and James T. Schleifer, English ed (Liberty Fund, 2012), originally published 1832, 102, 494; Michael J. Sandel, *Democracy's Discontent: A New Edition for Our Perilous Times* (Cambridge, Massachusetts: The Belknap Press of Harvard University Press, 2022), 13, 247, 255. - 12. Pateman, *Participation and Democratic Theory*, 43; Stephan Kirste, "Humanism and Republicanism," in The Oxford Handbook of Republicanism, ed. Frank Lovett and Mortimer Sellers. (Oxford University Press, 2024), https:// doi.org/10.1093/oxfordhb/9780197754115.013.13; Charles Taylor, "Cross-Purposes: The Liberal-Communitarian Debate," in *Debates in Contemporary Political Philosophy: An Anthology*, ed. Derek Matravers and Jonathan Pike (Routledge, 2003), 198, 207–208; John Dewey, "Creative Democracy: The Task Before Us," 1939. - 13. Equal political agency is usually taken to mean equal opportunity for influence, but it may require *actually* equal influence so all citizens are coauthors of the law (see: Daniel Wodak, "What Is the Point of Political Equality?," *Philosophical Review* 133, no. 4 (October 1, 2024): 367– 413, https://doi.org/10.1215/00318108-11497679). This would mean it is not enough for processes to merely allow all to participate; they would have to take steps to maximize actual participation, so that all enjoy the benefits of empowerment and development. - 14. Jean-Jacques Rousseau, *Letter to Beaumont, Letters Written from the Mountain, and Related Writings*, ed. Christopher Kelly and Eve Grace, trans. Judith R. Bush, vol. 9 (Hanover, New Hampshire: Dartmouth College Press, 2012), 260-261; Philip Pettit, "Democracy, Electoral and Contestatory," in *Designing Democratic Institutions*, ed. Ian Shapiro and Stephen Macedo (New York University Press, 2022), 105–44, https://doi. org/10.18574/nyu/9780814786628.003.0009; Allen, *Justice by Means of Democracy*, 28. - 15. David Estlund and Hélène Landemore, "The Epistemic Value of Democratic Deliberation," in *The Oxford* - *Handbook of Deliberative Democracy*, ed. Andre Bächtiger et al. (Oxford University Press, 2018), 112–31, https:// doi.org/10.1093/oxfordhb/9780198747369.013.26; Anita Williams Woolley, Ishani Aggarwal, and Thomas W. Malone, "Collective Intelligence and Group Performance," *Current Directions in Psychological Science* 24, no. 6 (December 2015): 420–24, https://doi. org/10.1177/0963721415599543. - 16. Elizabeth Anderson, "The Epistemology of Democracy," Episteme 3, no. 1–2 (June 2006): 8–22, https://doi.org/10.3366/epi.2006.3.1-2.8; Simone Chambers, "Human Life Is Group Life: Deliberative Democracy for Realists," *Critical Review* 30, no. 1–2 (2018): 36–48, https://doi.org/10.1080/08913811 .2018.1466852. Danielle Allen, *Justice by Means of Democracy*, 7-9. - 17. Hannah Arendt, *On Revolution* (New York: Penguin Classics, 2006), originally published 1963, 129; Pateman, Participation and Democratic Theory, 27; Allen, *Justice by Means of Democracy*, 28. - 18. Pateman, *Participation and Democratic Theory*, 26; Hannah Arendt, *The Promise of Politics* (Knopf Doubleday Publishing Group, 2009), 118, 126-127; Hannah Arendt, *On Revolution*, 218. Sandel, *Democracy's Discontent*, 13; Jed W. Atkins, *"Non-Domination and the Libera Res Publica in Cicero's Republicanism," History of European Ideas* 44, no. 6 (August 18, 2018): 756–73, https://doi.or g/10.1080/01916599.2018.1513705. - 19. Allen, *Justice by Means of Democracy*, 28; Taylor, *Cross-Purposes*, 201; G. D. H. (George Douglas Howard) Cole, *Self-Government in Industry*, G. D. H. Cole: Selected Works, Vol. 2 (London: Routledge, 2010), originally published 1919, 227–228. - 20. Pateman, *Participation and Democratic Theory*, 24- 26; Tocqueville, *Democracy in America*, 102, 494; John Stuart Mill, *Essays on Politics and Culture*, ed. Gertrude Himmelfarb (Garden City, New York: Anchor Books, 1963), 186, 229. - 21. Manin, *On Legitimacy and Political Deliberation*, 354. - 22. Christopher Lasch, *The Revolt of the Elites and the Betrayal of Democracy* (W. W. Norton & Company, 1995), 162-163. See also: Hugo Mercier and Dan Sperber, "Why Do Humans Reason? Arguments for an Argumentative Theory," *Behavioral and Brain Sciences* - 34, no. 2 (April 2011): 57–74, https://doi.org/10.1017/ S0140525X10000968. - 23. Tocqueville, *Democracy in America*, 914. - 24. Mill, *Considerations on Representative Government*, 58; Taylor, *Cross-Purposes*, 198–203; Tocqueville, *Democracy in America*, 157–160, 397–398. - 25. Sandel, *Democracy's Discontent*, 62, 252; Arendt, *The Promise of Politics*, 253. - 26. Lasch, *The Revolt of the Elites*, 170. - 27. Lasch, *The Revolt of the Elites*, 170-171; Manin, *On Legitimacy and Political Deliberation*, 351-352. - 28. Simon Niemeyer, "The Emancipatory Effect of Deliberation: Empirical Lessons from Mini-Publics," *Politics & Society* 39, no. 1 (2011): 103–40, https:// doi.org/10.1177/0032329210395000; Simon Niemeyer et al., "How Deliberation Happens: Enabling Deliberative Reason," *American Political Science Review* 118, no. 1 (2024): 345–62, https://doi.org/10.1017/ S0003055423000023. - 29. Mill, *Considerations on Representative Government*, 57-58, 229; Patrick J. Deneen, *Why Liberalism Failed* (Yale University Press, 2018), 22-24, 186-188. - 30. Allen, *Justice by Means of Democracy*, 31-35. #### BEYOND CLICKS AND COMMENTS: LEVERAGING AI FOR MEANINGFUL CIVIC ENGAGEMENT # Conversation Networks # Deb Roy Professor, Massachusetts Institute of Technology & Co-Founder, Cortico # Lawrence Lessig Professor, Harvard Law School # Audrey Tang Mozilla Foundation, Project Liberty Institute, and supported by funding from the Omidyar Network #### Meeting the Needs of Communities Picture a community torn over a proposed zoning law. Some are angry, others defensive, and misunderstandings abound. On social media, they broadcast insults at one another; every nuanced perspective is reduced to a viral soundbite. Yet, when they meet face-to-face and start speaking, something changes: residents begin listening more than speaking, and people begin testing ideas together. Misunderstandings fade, and trust begins to form. By the end of their discussion, they have not only softened their hostility, but discovered actionable plans that benefit everyone. This is the kind of meaningful discourse our society desperately needs. Yet our digital platforms—designed primarily for maximizing engagement through provocative content—have pulled us away from these core community endeavours. As a constructive path forward, we introduce the idea of *conversation networks* as a basis for civic communication infrastructure that combines interoperable digital apps with the thoughtful integration of AI guided by human agency. #### The Gap in Our Digital Infrastructure Community building relies on three forms of communication: - 1. Bridging: Bringing people together across divides, helping to reduce polarization and foster understanding between fragmented groups. - 2. Listening: Leaders—whether in organizations or communities—hear a wide range of authentic voices and perspectives, ensuring that everyone has a chance to be heard. - 3. Deliberation: Collective reasoning, testing ideas, and generating actionable decisions. What if we could harness the same elements that make social media powerful—its ease of use, habit-forming designs, and interconnected networks—to create something fundamentally different? Imagine scalable digital spaces designed for constructive communication: networks built around live, spoken conversations, rather than divisive content. These "conversation networks" offer a promising path to rebuild our fragmented social fabric, reduce polarization, and strengthen democracy. By redesigning key elements of social media, including AI for content analysis, intuitivelydesigned apps, and digital networks, and combining them with thoughtful investment, human-centered training, and shared standards, we can catalyze a new ecosystem. With the right support, conversation networks could transform how we engage with one another, and meet the urgent needs of communities worldwide. #### What Are Conversation Networks? In a conversation network, the core "content" are excerpts from recorded group conversations, not provocative social media posts or status updates. These recorded dialogues can be held in person or on digital platforms designed to encourage thoughtful discussion, broad participation, and respectful exchange. With consent of participants, excerpts from these conversations can then be heard by others to form new connections beyond the original conversation. For example, if two groups in a community have become mutually polarized and distrustful, exposure to each other in performative public spaces such as social media and open-mic town hall meetings can exacerbate divisions. Instead, members of each group could hold recorded conversations amongst themselves that are structured and facilitated to surface nuanced perspectives. Excerpts from the conversation recordings can then be shared across groups, enabling each to hear voices and authentic sentiments from the other. Digital infrastructure to make such exchanges easy and scalable are what we mean by conversation networks networks formed by the sharing of content that originates from live spoken conversation. The quality of content in this approach is shaped by the conversational context; if the group conversation is facilitated and well-structured, the content emanating from it and flowing through the conversation network will be high quality. Instead of rewarding the loudest or angriest voices, these networks can foster empathy, nuance, and the search for mutual understanding. Underneath the hood, a suite of tools—from AI-assisted transcription, analysis, and summarization, to speech and video conferencing apps—can enable these conversations to spread. Platforms such as Cortico and tools from the MIT Center for Constructive Communication facilitate meaningful dialogue, listening, and sensemaking. Meanwhile, Polis and Remesh facilitate large-scale AI-supported collaboration, while Frankly and the Stanford Online Deliberation Platform support largescale, structured video-based deliberation. By integrating the best of face-to-face and digital engagement, conversation networks aim to restore the social fabric worn thin by years of digital discord. #### Three Examples vTaiwan: Dialogue → Deliberation → Policy Formation vTaiwan1—a prototype of an open consultation process for society to engage in responsible discussion on national issues—exemplifies a structured approach to collaborative governance, combining dialogue, deliberation, and policy formation. The process begins with broad conversations in weekly meetups, where stakeholders and citizens identify key issues and perspectives. Next, participants use Polis, an innovative AI-supported deliberation tool, to engage in online discussions that uncover areas of consensus and divergence. Insights from Polis then feed back into smaller, facilitated conversations, where participants refine collective priorities and produce actionable steps. Government officials ultimately translate these insights into policies rooted in public input and shared understanding. Although vTaiwan distinguishes itself as the first large-scale deployment of Polis, facilitated conversations are equally vital. Pre-Polis discussions help generate high-quality, diverse opinion statements; post-Polis multi-stakeholder dialogues build on "bridging" statements2 identified by the platform, transforming them into practical policy recommendations. With enhanced digital infrastructure, vTaiwan could become a conversation network that integrates in-person recorded discussions with Polis so that all people involved in the deliberations could hear and be heard beyond their small group discussions. We envision infrastructure enabling excerpts from conversations to be linked to inputs and outputs of Polis, and participants provided with intuitive digital apps to listen to the voices of others. The result would be an end-to-end model for inclusive, data-informed governance that builds understanding between people. #### Newark Youth Voices: Dialogue → Sensemaking → Advocacy Newark Youth Voices leverages dialogue and technology to empower young people in shaping their community's future. The process begins with dialogue, where youth participate in inperson, recorded conversations, sharing their experiences and perspectives on issues affecting their lives. These recordings are then analyzed using Cortico's platform,3 which facilitates sensemaking by identifying key themes, patterns, and insights from the conversations. The conversations are designed and led by youth leaders, who also drive the sensemaking work. This process ensures that diverse voices are meaningfully represented. The insights are then mobilized to drive youth advocacy, equipping the Newark Opportunity Youth Network (OYN) team with the evidence and narratives they need to train teachers, engage decision-makers, propose solutions, and advocate for policies that reflect their collective priorities and lived experiences. The primary output from the Cortico platform is a "voice portal"4 which provides public access to excerpts from group discussions. If Polis and Cortico were interoperable, selected excerpts from Cortico could be imported into Polis to seed deliberations—which could advance the advocacy work of OYN. #### Deliberations.US: Information → Deliberation → Understanding Deliberations.US builds on Frankly's video-based deliberation platform to facilitate discussion about core issues of democracy. Participants view short (<5m) videos about topics of American democracy, including the electoral college, and money in politics. The content for these videos has been drawn from advisory panels, assuring a balanced presentation of the issues. Participants are assigned into small, video-based discussion groups, balanced based on demographics relevant to the deliberation. The platform moves them through a deliberation. Having measured attitudes at the start, the platform again measures attitudes after the deliberation, tracking how understanding has developed across demographics. If the discussion groups in Frankly are recorded, and Frankly were interoperable with Cortico, then the sensemaking and portal output features of Cortico would become easily available to Deliberations.US. This could enable an organized display of excerpts from deliberations associated with recommendations and outputs of the deliberation as a way to increase transparency and trustworthiness. #### A Fragmented Ecosystem All three examples illustrate the potential of structured conversation networks, deliberation systems, and sensemaking analytics. They also highlight a central challenge: the tools exist, but they often remain fragmented. One process stage may rely on cutting-edge AI, while another stage involves analog conversations, manual sorting through transcripts, or scattered collaboration on social media. For example, Cortico's sensemaking features could in theory be used in vTaiwan to help translate pre-Polis discussions into input for Polis, or to help organize and publicly present output from Frankly, but the three platforms do not currently interoperate—there is no plug-andplay data standard to connect the digital systems. When many people or organizations want to collaborate, it's vital that their digital tools fit together seamlessly through open standards. 5 This 'plug-and-play' approach enables several key features: - 1. Reduced friction: Communities can adopt multiple tools without painstaking integrations. - 2. Broader adoption: Lower technical barriers make it easier for diverse groups to try conversation networks. - 3. Shared learning: When data and insights move smoothly across platforms, best practices emerge more rapidly. Think of it like building with LEGO blocks: if every block snaps together, anyone can assemble a conversation network tailored to their community's needs. #### Fostering Shared Vocabulary, Concepts, and Skills When different conversation network tools work together, it does not just make tool use easier—it helps people develop a common language for how they talk, listen, and make decisions. As more communities adopt interoperable tools, they will naturally share vocabulary and methods, making it easier for everyone to learn from each other. For example, terms including "conversation guide" / "discussion guide" and "sensemaking" are used in overlapping ways by dialogue and deliberation practitioners, yet differences in what the terms connote impede the sharing of best practices. A standard for tools that support these processes would lead to greater conceptual alignment and foster easier learning across practitioners. While our emphasis here has been on digital tools, it is essential that people develop new communication skills and habits to effectively use these tools to foster constructive communication. Using these tools requires training, a commitment to community engagement, active listening, collaborative problem-solving, and thoughtful deliberation. It also needs a culture of accountability, transparency, inclusivity, and adaptability. By investing in human capacitybuilding—such as training for facilitation and deliberation practices—alongside technological innovation, we can ensure that these tools serve as enablers of community building. The combination of conversation-centric tools and methods has the potential for wide adoption, spreading into schools, workplaces, and neighbourhoods. Grassroots conversation networks could emerge, rebuilding the trust and agency that have eroded over decades under the pressure of top-down media, social media, and national-level political polarization and fragmentation. Such networks would not only strengthen communities but also help lay the foundation for a more resilient society. #### The Promise and Peril of Automation As we envision a cohesive ecosystem of tools to support conversation networks, we must also grapple with the role of automation in conversation networks, ensuring that its application actively strengthens, rather than undermines, community connections. Large language models (LLMs) are already showing transformative potential in shaping how we engage with information and public discourse. AI-driven summarization of conversations (e.g., fathom.ai, otter.ai) and AI-led interviews to understand public opinion (e.g., talktothecity.org) are being actively used today, while AI-mediated deliberation tools, like the Habermas machine6 remain a promising focus of ongoing research. The increasing integration of automated processes in bridging, listening, and deliberation, reflects a broader trend: the growing reliance on technology to predict and, in some cases, replace human participation. Imagine taking this trend to its logical extreme: every citizen could be represented by an AI avatar,7 continuously engaging in community discourse on their behalf—what we might call an "Avatar State." Would such a development fulfill John Dewey's vision of an inclusive participatory democracy, in which all citizens—given the right educational and institutional support—can actively engage in self-governance? Would it address the challenges Walter Lippmann identified as significant obstacles (that most people lack the expertise, time, or interest to participate effectively in governance)? Or would it veer toward a technoautocracy, where the essence of human agency is handed over to those who control the AI? In designing our future, we must consider the ideal roles AI can and should play—and, just as importantly, the roles that only people can fulfill. #### AI's Assistive Potential in Strengthening Conversation Networks AI technologies could displace human action or they could complement, by assisting human bridging, listening, deliberation. We believe they have enormous potential to be transformative and constructive in strengthening conversation networks when applied in an assistive capacity: - 1. Expanding participation: Offering supportive structures and learning resources can broaden engagement, inviting people from any background or level of expertise to take part in a meaningful way. - 2. Spreading effective practices: AI can help disseminate effective practices, allowing communities to learn from each other's successes and build upon established methods. - 3. Revealing blind spots: AI can uncover overlooked perspectives and issues, shedding light on concerns that might otherwise remain hidden. - 4. Automating repetitive tasks: Finally, AI can handle time-consuming "spade work," freeing up time for human participants to focus on contextualized decision-making and authentic human-to-human collaboration. In democratic processes such as citizen assemblies, AI should not come between people in ways that diminish or cut off direct humanhuman conversation. Instead, AI should be used to ensure members engage fully with one another, promoting genuine human interaction and deliberation without technological mediation. Learning how to work through differences of perspective and opinion is an integral part of the deliberative process that should be supported and protected. AI could help enable that. Bridging networks,8 such as scaled community networks on U.S. campuses, present another opportunity. In these networks, AI could support logistical and analytical tasks, while humans perform all key functions such as facilitation and overseeing sensemaking, fostering authentic connections and trust. #### Preserving Human Agency and Accountability Although AI can lighten the burdens of analysis and coordination, people must remain at the heart of conversation networks. We need to build civic muscle—the skills, habits, and capacities required for active participation—so communities can exercise agency and take ownership of their roles in governance. It is critical that humans, not AI tools, own their actions and take credit as well as responsibility for decisions and outcomes shaped by AI. AI used in the context of conversation networks must therefore operate within guardrails, refusing tasks that require making subjective judgments about what other people feel or want. Instead, AI should be focused on more objective tasks—such as organizing data or highlighting patterns—while leaving interpretation and moral responsibility to human beings. Performance audits, in which even non-experts critically evaluate how AI systems are functioning, are equally important to keep under human control. Maintaining decentralized control in the creation and shaping of conversation networks is also essential, with a particular emphasis on fostering community-controlled AI to prevent power from concentrating in the hands of a few. Lastly, humans must actively practice interpersonal connection, ensuring that technologies mediating our interactions enhance and deepen relationships, rather than erode them. By placing human agency, trust, and accountability at the heart of how we use AI, we can ensure that AI remains a catalyst for authentic conversation and collective action rather than a replacement for it. #### Conclusion The need for meaningful civic discourse has never been greater, and the tools to support it are within reach. Conversation networks, supported by thoughtful integration of AI and guided by human agency, offer a path forward. By addressing current gaps, fostering interoperability, and investing in both technology and the people who use it, we can create scalable systems that empower communities, build trust, and strengthen societies worldwide. But technology alone is no panacea. We must invest in human skill-building at every level—teaching people how to hold nuanced conversations, make sense of what others mean, weigh evidence critically, and co-create solutions. At the same time, we must establish guiding design principles, open standards, and ethical safeguards so that AI bolsters—rather than undermines—our democratic goals. The stakes could not be higher. Polarization and mistrust will continue to unravel our social fabric unless we choose to act. We have at our disposal the tools and know-how to transform our digital landscape, but the key question remains: are we willing to invest in both human and technological capacities to fulfill this vision? The promise of conversation networks is more than just a hope for better social media; it is an opportunity to reclaim our collective agency and renew our sense of community. If we seize this moment, we can catalyze nothing less than the regeneration of our shared civic life. #### Endnotes - 1. Details about the vTaiwan platform are available at https://info.vtaiwan.tw/ - 2. For more on bridging systems, see E. Glen Weyl et al., "Prosocial Media" (arXiv, February 18, 2025), https://doi.org/10.48550/arXiv.2502.10834 . - 3. Details about the Cortico platform are available at https://cortico.ai/platform/ - 4. The portal for this project can be found at https://newarkyouthvoices.portal.fora.io/ - 5. Open standards are publicly available, consensus-driven specifications that define how digital tools and systems communicate and exchange data. They allow different software, hardware, and networks to work together seamlessly, free from proprietary restrictions or vendor lock-in. - 6. Michael Henry Tessler et al., "AI Can Help Humans Find Common Ground in Democratic Deliberation," *Science* 386, no. 6719 (October 18, 2024), https://doi.org/10.1126/science.adq2852 . - 7. A recent study demonstrates that AI simulations of individual people, based on interviewing them, can accurately predict their attitudes and behaviours. See: Joon Sung Park et al., "Generative Agent Simulations of 1,000 People" (arXiv, November 15, 2024), https://doi.org/10.48550/arXiv.2411.10109 . - 8. A conversation network that bridges groups of people by enabling them to hear cross-cutting perspectives, experiences, and opinions. #### BEYOND CLICKS AND COMMENTS: LEVERAGING AI FOR MEANINGFUL CIVIC ENGAGEMENT # Reimagining Democracy: The Role of Technology and Deliberation # Alice Siu Associate Director at Stanford Deliberative Democracy Lab, and Senior Research Scholar at Center for Democracy, Development, and the Rule of Law at the Freeman Spogli Institute at Stanford University In an era marked by rapid technological advancements, increasing political polarization, and democratic backsliding, reimagining democracy requires innovative approaches that foster meaningful public engagement. Over the last 30 years, Deliberative Polling has proven to be a successful method of public consultation to enhance civic participation and informed decision-making. In recent years, the implementation of online Deliberative Polling using the AI-assisted Stanford Online Deliberation Platform, a groundbreaking automated platform designed to scale simultaneous and synchronous deliberation efforts to millions, has put deliberative societies within reach. By examining two compelling case studies—*Foreign Policy by Canadians* and the *Metaverse Community Forum*—this paper highlights how technology can empower diverse voices, facilitate constructive dialogue, and cultivate a more vibrant democratic process. This paper demonstrates that leveraging technology in deliberation not only enhances public discourse but also paves the way for a more inclusive and participatory democracy. #### What is Deliberative Polling®? Deliberative Polling was created by James Fishkin in 1988 as a response to the shortcomings of traditional public opinion polling and the challenges facing our democracy.1 He aimed to establish a method that would facilitate a more informed citizenry. To maintain the quality and consistency of this approach, Deliberative Polling is a registered trademark of Fishkin, and any revenue generated from the trademark supports research at the Stanford Deliberative Democracy Lab. As of 2024, Deliberative Polling has been carried out in over 50 countries, with more than 150 projects implemented. This method has had a substantial policy impact globally. For instance, in 2017, Mongolia enacted the Law on Deliberative Polling, mandating that the Parliament conduct a national Deliberative Poll before making any amendments to the constitution.2 Since the enactment of this law, Mongolia has organized two national Deliberative Polls, and in accordance with the law, its Parliament has reviewed the findings and amended certain sections of the constitution. The Deliberative Polling method consists of six key steps: A random, representative sample of citizens is initially surveyed on specific issues to capture their opinion before deliberations. Following this initial survey, a sample is invited to gather—either online or in-person—for at least one day (sometimes over a weekend, or several days across a few weeks). Carefully balanced briefing materials are provided to participants, which are also made publicly available after the deliberations conclude. An advisory committee is then convened to vet briefing materials for balance and accuracy. Participants are randomly assigned to small groups of about 8-10 individuals, where they engage in discussions on the topics. These small group sessions alternate with plenary Q&A sessions featuring experts throughout the deliberation process. During the plenary Q&A sessions, participants have the opportunity to interact with competing experts and political leaders, posing questions that emerge from their small group discussions. After the deliberations, the same questions from the initial survey are administered to the sample in a confidential questionnaire. The resulting changes in opinion reflect what the public thinks when they are informed and actively engaged with the issues. The findings are subsequently disseminated to the public through various media channels. The goal is to create a space where individuals can exchange views, challenge their assumptions, and consider different perspectives on particular policy issues. Deliberative Polling not only measures shifts in public attitudes but also enhances the quality of democratic discourse on specific policy issues by fostering informed, thoughtful engagement among citizens. Deliberative Polling aims to provide policymakers with insights that reflect the considered judgments of a well-informed public, rather than spontaneous reactions to issues. Results from Deliberative Polls have demonstrated that deliberation alleviates polarization,3 better moderates opinions,4 and produces lasting effects on policy opinions, voting intention, and respect for those with differing opinions.5 #### AI-assisted Stanford Deliberation Platform In partnership with the Stanford Crowdsourced Democracy Team, the AI-assisted Stanford Online Deliberation Platform was developed in 2018 to facilitate structured, high-quality deliberations without the need for human moderators. Since its launch in 2020, the platform has accumulated over 100,000 hours of deliberation, has been utilized in more than 40 countries, and is available in over 30 languages, with commissioned projects from various governments and industry leaders worldwide. Drawing on more than 30 years of experience in Deliberative Polling and the training of human moderators, this platform allows ![](_page_56_Figure_3.jpeg) *Illustration: This illustration shows the steps in a Deliberative Polling process. The image of the small group deliberations is from a national Deliberative Poll in Tanzania, and the image of the expert plenary sessions is from a national Deliberative Poll in North Macedonia.* ![](_page_57_Figure_0.jpeg) *Illustration: Map of Deliberative Polling projects.* small groups of about ten to self-moderate. Key features of the platform include managing microphone access (capping contributions at 45 seconds), guiding discussions through various policy proposals, checking for a balanced presentation of arguments, intervening in cases of incivility (which is rare), and helping groups formulate questions for plenary sessions. The platform uses AI to detect toxic language, and in the event that it does, the software prompts the admin on the backend and personnel that are monitoring the deliberations in real time, allowing them to jump into the small group if needed. An important feature of the deliberation platform is the availability of real-time transcripts, which allows for human monitors to observe deliberation in real time with the ability of doing so simultaneously for many small groups. Furthermore, the real-time transcripts allow organizers immediate access to transcripts of all small group discussions, offering rich evidence regarding participants' reasons for supporting or opposing policy proposals. The deliberations held on this platform effectively combine qualitative and quantitative data, with qualitative insights enriching the quantitative findings, creating "a poll with a human face."6 This approach goes beyond mere numbers by articulating participants' genuine concerns in their own words. Furthermore, results from deliberations conducted on the AI-assisted platform indicate improved gender equity compared to traditional in-person discussions.7 Additionally, in a controlled experiment comparing in-person deliberations with a human moderator, Zoom deliberations with a human moderator, and deliberation on the AI-assisted deliberation platform, the deliberation platform demonstrated similar trends in attitude shifts toward the discussed policy proposals and evaluations of the deliberation process when compared to both in-person deliberations and Zoom sessions moderated by humans.8 ![](_page_58_Figure_0.jpeg) *Illustration: A snapshot of the AI-assisted Stanford Online Deliberation Platform.* #### Case Studies #### Case Study 1: Foreign Policy by Canadians: A National Deliberative Poll The "Foreign Policy by Canadians" (FPbC) Deliberative Poll served as a national deliberative experiment conducted in March and April of 2021.9 FPbC was a joint initiative by the Canadian International Council, the Canadian Partnership for Women and Children's Health, and Global Canada. FPbC commissioned YouGov to recruit participants through stratified random sampling in both the deliberation and control groups. Following two rounds of deliberation, a total of 444 Canadians took part in the Deliberative Polling process. Concurrently, a control group of 300 individuals who did not deliberate completed both the pre- and post-surveys. Prior to deliberation, the deliberative participants and the control group were nearly identical, but afterward, the participants arrived at considered opinions, while the control group showed little change. FPbC aimed to provide informed and representative policy input, which is particularly challenging because the public tends to be less engaged with foreign policy than with domestic issues, except during times of war or significant international crises.10 Therefore, Deliberative Polling is especially suited for gathering public input on topics people have not thought much about. On topics that are less salient, the public will be able to learn more about the issues together and engage in discussion with diverse others to come to considered opinions. This Deliberative Poll was notable in the following three ways. First, all discussions among the 444 nationally representative deliberators occurred through the AI-assisted Stanford Online Deliberation Platform. When FPbC was conducted in March 2021, it was only the second national use of this AI-assisted online platform, following a similar application nationally in Chile, which was commissioned by the Chilean National Congress.11 A second distinctive feature of FPbC was its agenda-setting process, which involved both topdown and bottom-up approaches. An advisory group provided input on relevant topics and material accuracy, while chapters of the Canadian International Council across the country reviewed and discussed how to improve the draft materials. They evaluated what was missing and whether the most crucial arguments were represented, agreeing on recommendations that shaped the final discussion documents. This process exemplifies "deliberative crowdsourcing," as groups reached a consensus on their most important suggestions through shared discussion, all facilitated by the automated deliberation platform. In fact, one of the policy proposals from the deliberative crowdsourcing process was used in the final Deliberative Polling briefing materials for FPbC. Third, the findings of the Deliberative Poll showed that participants in FPbC had increased positive views of democracy and trust after deliberations.12 The first question in the survey asked respondents to assess "how poorly or well would you say the system of democracy in Canada works." Initially, 69 per cent felt that democracy was working well, but this figure significantly increased to 80 per cent after deliberation—a notable improvement in participants' perceptions of democracy in the country. Additionally, this enhanced view of democracy was accompanied by increased mutual respect for those with whom participants strongly disagreed. This was reflected in pre- and post-deliberation surveys, which included several questions regarding opinions about people they strongly disagree with. For instance, when asked about individuals they "strongly disagree with on issues like those we have been discussing," the percentage of participants who said "they just don't know enough" decreased from 49 per cent to 43 per cent (p=0.090). Those responding "they believe some things that aren't true" fell from 63 per cent to 55 per cent (p=0.000), and those who said "they are not thinking clearly" dropped from 48 per cent to 38 per cent (p=0.001). By the end, 70 per cent of participants agreed that they would be willing to "compromise to find a solution we can both support," up from 66 per cent. For comparison, the Stanford Deliberative Democracy Lab (DDL) conducted an in-person national Deliberative Poll called "America in One Room" (A1R), involving a nationally representative sample of 526 people from across the United States.13 The same questions regarding those they strongly disagree with were posed. In A1R, the percentage of participants agreeing with "they just don't know enough" remained unchanged at 53 per cent after deliberation. In contrast, in FPbC, only 43 per cent agreed with this statement post-deliberation. While A1R participants also experienced significant positive changes in their views, they maintained generally negative opinions of those with whom they disagreed. In addition, overall, FPbC participants' views on foreign policy underwent significant changes regarding several proposals related to the four topics of Global Public Health, Security, Prosperity, and Human Dignity. There were statistically significant shifts in opinions on proposals such as prioritizing trade agreements with democratic nations, expanding embassy presence abroad, and adopting digital innovation as a key driver of economic growth, among others. However, it is noteworthy that while some opinions changed, many remained unchanged before and after deliberation, indicating that certain viewpoints are stable and hard to shift. As an example of opinions shifting post deliberation, participants appeared to be less supportive of investing in Canada's diplomatic network, with support decreasing from 68 per cent to 58 per cent for the proposal "Canada should increase the number, size, and resources of embassies around the world to better advance Canadian national interests in our interactions with other countries." Here are qualitative insights from select participants to show reasoning behind the decreased support on this issue. Participant A: *"I don't think we should be that focused on this right now. We have other priorities. We already have huge deficits through the pandemic and I think we could better spend our money at this point."* Participant B: *"Yeah, I tend to echo the exact same sentiment. I'm not really opposed to more embassies. But I mean, is that really a priority right now? I don't think it should be…should be way down on the list."* Participant C: *"Absolutely not. We got a lot more important pressing issues going on. We have COVID, we have the deficit problems with other countries going around to be going into other places and to open up embassies. That's just a waste of money. We need the money here to help take care of our Canadians here and now, thank you."* At the conclusion of the deliberations, participants evaluated the process: 90 per cent found the briefing materials "valuable," 87 per cent felt the plenary sessions were "valuable" in clarifying their positions, and 97 per cent regarded the overall event as "valuable." Additionally, 89 per cent believed the discussion platform encouraged broad participation, 75 per cent agreed that the platform ensured consideration of opposing viewpoints, and 83 per cent felt that important aspects of the issues were adequately addressed. These evaluations are notably high, comparable to the best ratings received in Deliberative Polls conducted in other countries in person. This case study demonstrates the ability to engage the public in deliberative crowdsourcing to assist with agenda setting for a broader public deliberation. Many deliberative processes often struggle with how to build deliberative agendas to more effectively include broader input in a thoughtful way. And, furthermore, this national Deliberative Poll brought considered opinions to the event organizers and policymakers on a topic that the public generally does not think deeply about. With thoughtful and informed opinions, the organizers are able to more meaningfully engage in recommendations and guidance for government and policymakers. #### Case Study 2: First Global Deliberative Poll, Policies Toward Bullying and Harassment in the Metaverse In December 2022, the Stanford Deliberative Democracy Lab collaborated with Meta and the Behavioral Insights Team to conduct the first Meta Community Forum (MCF), a global Deliberative Poll on policies toward bullying and harassment in the Metaverse.14 The MCF was a groundbreaking experiment in global deliberation. A scientifically representative sample of social media users from 32 countries (of which Canada was one of the countries) across nine regions and 19 languages was gathered for a weekendlong discussion (or the equivalent of a weekend spread out over two weeks). A comparable control group, also of significant size, did not participate in the deliberation but completed the same questionnaires during the same timeframe in December 2022. The focus of this initiative was a unique and critical issue: how to regulate bullying and harassment in virtual spaces, especially within the private or "members-only" social VR spaces in the Metaverse. Over 6,300 deliberators (and over 6,700 control group participants), representing global social media users, were selected by 14 survey research partners. The central issue for deliberation was the responsibility of platform owners, like Meta, compared to that of individual creators in regulating behaviour within private or membersonly VR social spaces. To what extent should platform owners refrain from intervening, given that these spaces are not public and participants join by mutual consent, potentially wanting to establish their own behavioural norms? Conversely, to what degree do platform owners have a duty to protect users from bullying and harassment? If they do bear responsibility, what actions should they take? These questions represent new challenges and lay the groundwork for a social contract governing these virtual reality spaces. It's important to note that the discussions focused specifically on bullying and harassment, excluding other forms of abusive or illegal online behaviour. This global Deliberative Poll was significant for two reasons. First, the Metaverse currently faces significant challenges related to safety, privacy, and regulation. Many users report experiencing harassment, bullying, and toxic behaviour in virtual environments, where inadequate moderation and unclear reporting mechanisms hinder user safety. Privacy concerns are prominent, as the collection and use of personal data often lack transparency, leaving users uncertain about risks like surveillance and data breaches. And, the lack of a regulation for the Metaverse results in inconsistencies in user rights and protections. Therefore, this global deliberation about policies toward bullying and harassment in the Metaverse allowed for a discussion on what governance of the Metaverse should look like. This deliberation gave the public a voice to shape the guardrails for Metaverse governance. Second, the essential technology that enabled this deliberation was the AI-assisted Stanford Online Deliberation Platform. This platform facilitated discussions in 19 languages, and its AI-assisted moderator allowed 2,069 small groups to share the same deliberative experience. By providing a consistent deliberative process and minimizing variability in implementation, it ensured that any observed changes (or lack thereof) in opinions after deliberation were attributed to the deliberation itself rather than to differences in moderators or the presentation of materials. Every participant received the same introduction, watched identical briefing videos, followed the same discussion prompts, and interacted with the same interface. Whether this global deliberation takes place in-person or on Zoom with a human moderator, there will always be some variability in human moderators' performance—because they are human. Over the past 30 years, DDL has conducted both in-person and online sessions with human moderators, and despite training, some moderators do not perform well, which can negatively impact participants' experiences. The AI-assisted Stanford Online Deliberation Platform effectively removes this human variability. Without this platform, the logistics of the event would have required over 500 human moderators fluent in 19 languages, along with additional staff, to support the more than 2,000 small group sessions. Having engaged participants across 32 countries, this event demonstrated that deliberation can be scaled effectively, and the positive evaluations from participants indicated that they appreciated using the platform, gained insights into the discussion topics and their peers, and experienced shifts in their opinions. Now, to share some of the findings of this global Deliberative Poll. The deliberations showed there was less emphasis on privacy in public areas, which made the role of platforms in regulating and protecting users from bullying and harassment clearer from the beginning. However, concerns about platform responsibilities increased significantly during the deliberation process. The findings provided clear guidance on which tools or technologies should be utilized to combat bullying and harassment. Participants primarily discussed the use of video capture and automatic speech detection as potential tools in the Metaverse. They deliberated on whether "platform owners should have access to video capture in members-only spaces," with support for this proposal rising from 59 per cent to 71 per cent, an increase of 12 per cent. Additionally, support for the idea that "platforms should take action against creators in spaces with repeated bullying and/or harassment" increased by about 10 per cent, from 57.3 per cent to 66.9 per cent. Despite these recommendations, the global sample did not favour punitive measures. For example, support for removing members-only spaces with repeated bullying only reached 43 per cent, up from 39 per cent. Similarly, support for banning creators from establishing additional members-only spaces in cases of repeated bullying and harassment only reached 45 per cent, an increase from 38 per cent. Support for preventing creators of such spaces from inviting new members stood at 49 per cent, up from 43 per cent, and support for stopping creators from profiting off their spaces reached 54 per cent, rising from 49 per cent. This case study was pathbreaking in many ways. Most notably, the demonstration of a global Deliberative Poll allowed for thousands of people around the world to engage in small group deliberations simultaneously. The success of this global deliberation allows us to imagine the possibility of having millions of people deliberate in small groups simultaneously. What if societies deliberated together in small groups in the weeks and days ahead of elections? And, what if societies would regularly deliberate on any given day to discuss the happenings in their communities? The AI-assisted deliberation platform can make it a reality. #### Conclusion Technology is helping us reimagine how to implement deliberation in our democracy. Technology is essential, as it expands access to civic engagement, enables broader participation, and facilitates informed discussions among diverse populations. In an age where traditional methods of public consultation can be limited by geographical, logistical, and social barriers, technology allows individuals from various backgrounds to connect, share perspectives, and engage in meaningful dialogue—regardless of their location. The success of Deliberative Polling in Canada and around the world through the AIassisted Stanford Online Deliberation Platform underscores the transformative potential of technology in revitalizing democratic engagement. The AI-assisted platform is providing structured, high-quality, and consistent deliberative experiences to participants, and through these experiences, organizers of deliberation can enhance civic participation and ensure that diverse perspectives are heard and valued. As we face ongoing challenges in our democratic systems, embracing such advancements will be crucial in nurturing a more engaged, equitable, and responsive democracy for all. #### Endnotes - 1. James S. Fishkin, *The Voice of the People: Public Opinion and Democracy* (Yale University Press, 1995), https://www.jstor.org/stable/j.ctt32bgmt. - 2. "Mongolia's First National Deliberative Poll on Constitutional Amendments," *Deliberative Democracy Lab* (blog), April 29, 2017, https:// deliberation.stanford.edu/news/mongoliasfirst-national-deliberative-poll-constitutionalamendments. - 3. James Fishkin et al., "Is Deliberation an Antidote to Extreme Partisan Polarization? Reflections on 'America in One Room,'" *American Political Science Review* 115, no. 4 (November 2021): 1464–81, https://doi.org/10.1017/ S0003055421000642. - 4. Willoughby J. Winograd and Alice Siu, "Depolarizing Through Deliberation in Civics Education: A Case Study of Like-Minded High School Students," *The Good Society* 31, no. 1 (2022): 86–106; Yasmeena Khan and Alice Siu, "How College Students Can Depolarize: Evidence for Political Moderation Within Homogeneous Groups," *Michigan Journal of Community Service Learning* 29, no. 2 (December 22, 2023), https://doi.org/10.3998/mjcsl.5140. - 5. James Fishkin et al., "Can Deliberation Have Lasting Effects?," *American Political Science Review* 118, no. 4 (November 2024): 2000–2020, https://doi.org/10.1017/S0003055423001363. - 6. James S. Fishkin, "Consulting the Public through Deliberative Polling," *Journal of Policy Analysis and Management* 22, no. 1 (2003): 128–33. - 7. Lodewijk Gelauff et al., "Achieving Parity with Human Moderators: A Self-Moderating Platform for Online Deliberation 1," in *The Routledge Handbook of Collective Intelligence for Democracy and Governance* (Routledge, 2023). - 8. Kimmo Grönlund et al., "Good Deliberation Regardless of Mode? An Experimental Comparison" (APSA 2024 Convention, Philadelphia, PA, United States, 2024), https:// connect.apsanet.org/apsa2024/2024-themepanels/. - 9. James Fishkin et al., "Foreign Policy by Canadians" (Center for Deliberative Democracy at Stanford University, May 2021), https://drive.google.com/file/d/1gZ18si3z-NsfLXtZp7H0VKOXCqJpR2YA/view. - 10. Philip J. Powlick and Andrew Z. Katz, "Defining the American Public Opinion/Foreign Policy Nexus," *Mershon International Studies Review* 42, no. 1 (May 1998): 29, https://doi. org/10.2307/254443. - 11. "National Deliberative Poll In Chile," *Deliberative Democracy Lab* (blog), January 17, 2020, https://deliberation.stanford.edu/news/ national-deliberative-poll-chile. - 12. Fishkin et al., "Foreign Policy by Canadians." - 13. "America in One Room," *Deliberative Democracy Lab* (blog), July 5, 2019, https:// deliberation.stanford.edu/news/america-oneroom. - 14. Samuel Chang, Estelle Ciesla, and Tenzin Kartsang, "Meta Community Forum: Results Analysis" (Deliberative Democracy Lab at Stanford University, June 2023), https://fsi9-prod.s3.uswest-1.amazonaws.com/s3fs-public/2023-06/ public\_meta\_community\_forum\_final\_report\_-\_ stanford\_ddl\_1.pdf.