Mozilla Democracy x AI Cohort: Call for Proposals x vTaiwan

--- tags: vTaiwan --- # Mozilla Democracy x AI Cohort: Call for Proposals x vTaiwan ## 主筆記 - 從摩茲社群的 Irvin 處取得資訊 - 於 [20260225 小松](/wBFeFu4HQVOSrpvtx8Wyow) 確認以之前申請 TWNIC 的企劃書改投稿來申請 - 時程： - Initial Project Proposal Opens: February 12, 2026 - Initial Project Proposal Deadline: March 16, 2026, 11:59pm PT - Finalists Notified: April 2, 2026 - Full Proposals Due: April 15, 2026 - Selections Announced: June 1, 2026 - Cohort Begins: 12 months beginning June 2026 - [20260304 小松](/6ndubFQDRZqGUQ9BzXtgSQ) 確認需要整理以下資料 - 見第一階段提案 - 本次申請案的性質：加速器，提供工具的啟動資金 [name=irvin] - 建議：提到台灣的獨特性：作為 AI 不實訊息操弄的前線 - 如何利用 AI 去面對與處理相關的議題？ - 選擇台灣的原因： - 台灣是不實訊息與資訊操弄的實驗場 - OpenAI 最新的報告：IO (Influence Operation)的威脅 - 台灣面臨社會極化問題，而地緣政治風險讓前述的社會極化變得更加嚴重 - 在面臨衝突與極化的環境下，仍然能夠透過網路快速地進行民主程序 [name=josh] - 可以參考[動態討論視覺化：讓口頭討論可以及時轉成 polis 意見的工具](/MwEMrUEiSMCx9EAnIcPjZw) ## 第一階段提案 - OCF 作為 fiscal sponsor 的可能性 - 目前在跟 OCF 確認的可能性 - 根據這個 proposal 的設計，要通過第一階段才有第二階段。 ### 0304 小松討論 - 參考 [動態討論視覺化：讓口頭討論可以及時轉成 polis 意見的工具](/MwEMrUEiSMCx9EAnIcPjZw) - 潛在諮詢對象：Maximilian Kroner Dale [name=yiting] - https://www.youtube.com/watch?v=EzVN2IJhP7Q - 過去使用 polis 工具的經驗與限制 - 結合 vTaiwan 社群過去使用 polis 進行審議討論的實際成績 - 工具做完之後的討論 - 利用 AI Agent 可以更進一步將相關的意見徵集與發表結果發布在社群媒體上，去擴散討論與 - 過去 polis 的痛點：沒有辦法擴散的問題 - 目前希望解決的問題：可以持續的擴散並且延展討論 - recursive - 運用OpenClaw龍蝦大戰假訊息機器人 - 加分 - Mozilla 社群的長期參與 - g0v 的公民科技社群基礎 - 公民科技社群長期有與政府協作，甚至改變政府政策的相關經驗 - vTaiwan 跟 Join 公共政策參與平台 - 開放政治獻金與政治獻金線上公開 - 可以應用的議題： - 民防是否需要加強？ - 死刑？ - 民間對於軍購的看法？ - 後期的功能 - 身份驗證的機制 - 參考文獻： - https://journals.sagepub.com/doi/10.1177/13540661251382639 ### Proposed Project Title #### 暫時結論 Making Consensus‑Building Alive: AI‑Assisted Real‑Time Deliberation from the vTaiwan Experience in Digitally Polarized Societies ### other preliminary discussion 1️⃣ 最接近你原句的修正版 Making Consensus‑Building Alive: A Real‑Time Deliberation Tool from the vTaiwan Experience 2️⃣ 更清楚一點（推薦） Making Consensus‑Building Alive: Real‑Time Deliberation Tools from the vTaiwan Experience 3️⃣ 加上極化社會背景 Making Consensus‑Building Alive: Real‑Time Deliberation Tools for Polarized Societies 4️⃣ 最像 grant proposal 的版本（我最推薦） Making Consensus‑Building Alive: Real‑Time Democratic Deliberation Inspired by vTaiwan 5️⃣ 如果想強調 AI Making Consensus‑Building Alive: AI‑Assisted Real‑Time Deliberation from the vTaiwan Experience #### 其他選項 1️⃣ 比較標準、穩健型（最安全） AI-Assisted Democratic Deliberation: Real-Time Consensus Discovery for Public Decision-Making 2️⃣ civic tech 風格（我個人很推薦） Sensemaking for Democracy: AI Tools for Real-Time Public Deliberation 3️⃣ Pol.is 延伸型（讓 reviewer容易理解） From Conversation to Consensus: AI Infrastructure for Democratic Deliberation 4️⃣ Mozilla 風格（比較有敘事感） Scaling Democratic Deliberation with AI 5️⃣ 技術特色型（突出你的即時系統） Real-Time Civic Sensemaking: AI-Assisted Deliberation for Complex Public Decisions 6️⃣ 我最推薦的一個（很符合你這個 project） From Conversation to Consensus: AI-Assisted Infrastructure for Democratic Deliberation 1️⃣ 最清楚版本（最安全、grant reviewer 很好懂） AI‑Assisted Democratic Deliberation Infrastructure 2️⃣ 強調公民參與 Scaling Democratic Deliberation with AI 3️⃣ 強調台灣與全球影響 AI for Democratic Deliberation: Lessons from Taiwan’s Civic Tech Ecosystem 4️⃣ 強調工具與平台 Building AI‑Assisted Infrastructure for Public Deliberation 5️⃣ 比較有 vision（Mozilla 會喜歡） Rebuilding Public Deliberation in the Age of AI ### Project Summary :::info We are building an AI‑assisted deliberation system designed for real‑time consensus discovery in civic discussions. The system is intended for citizen assemblies, public policy forums, hybrid meetings, and other time‑limited deliberation settings where groups must quickly identify areas of agreement and disagreement. It can also support public debates and policy presentations during elections, helping audiences better understand competing viewpoints and identify areas of public consensus. The project builds on lessons from Taiwan’s vTaiwan civic technology process including Po.lis, which has facilitated deliberation on dozens of national policy issues through open collaboration between citizens, government, and civil society — including the regulatory framework for ride‑sharing services such as Uber. Taiwan is one of the world’s most highly connected and digitally active societies, but it is also on the frontline of AI‑driven influence operations and large‑scale cognitive manipulation campaigns. These conditions make Taiwan a critical testing ground for new democratic technologies. The system we propose is designed specifically to address this challenge: helping societies maintain meaningful public dialogue in environments shaped by misinformation, algorithmic amplification, and political polarization. The system operates in three stages. 1st, AI captures and analyzes live discussions or debates, extracting key arguments and opinion statements in real time. 2nd, participants or audiences respond to these statements through simple agree/disagree/pass interactions, allowing the system to rapidly map areas of consensus and disagreement using clustering and opinion analysis techniques inspired by Pol.is. 3rd, AI agents use the resulting consensus map to engage with the broader information ecosystem — sharing summaries, responding to related discussions online, and helping extend deliberation beyond the original event. By combining real‑time interaction, AI‑assisted sense‑making, and network‑scale dialogue, the platform allows democratic deliberation to scale beyond traditional facilitation limits while preserving human verification and collective decision‑making. Developed within Taiwan’s g0v civic tech ecosystem, the project aims to transform proven deliberation practices into reusable infrastructure that helps digitally polarized societies move from confrontation toward dialogue and consensus‑building. While developed in Taiwan, the approach is also relevant to other democracies facing deep political polarization — including the United States and other digitally connected societies where public discourse is increasingly shaped by algorithmic media environments. 不知道申請書的長度限制為何，可以視情況縮短 ::: ### Your Project - What technology are you building? How does it address this cohort's theme? - 使用的元素： - 即時的投票、回應與討論功能 - UI / UX 可以讓在地的 NGO、里長、基層民代可以輕鬆的導入與應用 - 利用 AI 來補足目前討論在 onboarding 其他參與者 - optional: 可驗證為真人與在地居民的功能 - 克服隱私問題的真人驗證 :::info ### V.2 草稿 We are building an AI-assisted deliberation system designed for real-time consensus discovery in civic discussions. The system is intended for use in citizen assemblies, public policy forums, hybrid meetings, and other time-limited deliberation settings where groups must quickly identify areas of agreement and disagreement. The design follows several key principles: the technology should be low-interference (not interrupting ongoing discussions), real-time (producing insights during the conversation rather than after), and assistive rather than directive. Inspired by the philosophy of Pol.is, the system focuses on discovering shared consensus rather than amplifying the loudest voices. The platform operates in two stages. In the first stage, the system captures opinions through real-time, rapid, low-friction interactions. Participants can respond to short statements using simple agree/disagree/pass reactions. These statements may be embedded in online learning materials or generated dynamically from live discussions using large language models that identify key claims in conversation. In the second stage, the system analyzes the resulting opinion matrix using dimensional reduction and clustering techniques. AI then summarizes clusters of viewpoints and asks participants to verify whether these summaries accurately represent their positions. Through repeated validation loops, the system identifies the broadest possible “rough consensus” that participants explicitly confirm. By combining real-time interaction with AI-assisted sense-making, the technology allows democratic discussions to scale beyond traditional facilitation limits while preserving human verification and collective decision-making. ::: - Describe your project's traction. What’s working? What evidence do you have that this technology solves a real problem? :::info ### V.2 草稿 This project builds on more than a decade of civic technology experimentation through the vTaiwan process in Taiwan. Since 2014, vTaiwan has facilitated deliberations on over 30 national policy issues, including the regulation of ride-sharing platforms, fintech regulation, and online alcohol sales. The vTaiwan model combines online opinion gathering with offline deliberative workshops, allowing citizens, government officials, and industry stakeholders to collaboratively explore policy solutions. Tools such as Pol.is and Mentimeter have been used to collect and analyze large-scale public input, helping identify areas of consensus across different stakeholder groups. Recent work has expanded this model through experiments with AI-assisted sense-making. In collaboration with organizations including Chatham House and the AI Objectives Institute in the Democratic Inputs to AI project, Pol.is data and large language models were used to analyze public input from large-scale civic consultations. The community has also begun experimenting with AI-based tools that summarize deliberation outcomes and simulate missing citizen perspectives to improve representation in discussions. These experiments suggest that AI can significantly reduce the time required to process large volumes of public input while helping facilitators identify patterns of agreement that would otherwise be difficult to detect. The proposed platform builds on these existing practices and aims to transform them into a reusable infrastructure for democratic deliberation. ::: ### Your Team - Who is building this? What relevant experience does your team have? - 希望下週小松前可以拿到相關資訊 - Peter Cui - Joshua 楚約, https://joshuacyang.com - Tim, https://www.linkedin.com/in/tim-y-5446ab3b6 - Yiting Lien, https://www.linkedin.com/in/yiting-lien-b95431135 - Irvin Chen, https://www.linkedin.com/in/irvinchen/ - 待補 :::info ### V.2 草稿 This project is developed by members of the vTaiwan and g0v civic tech community in Taiwan, bringing together expertise across law, public policy, software engineering, data analysis, design, and open internet advocacy. The project is supported by the Open Culture Foundation (OCF), a non-profit organization that promotes open technology, digital rights, and open culture initiatives in Taiwan. Peter Cui is a legal researcher and civic tech contributor working on AI governance, digital policy, and democratic participation in technology regulation. Through his work with the vTaiwan community, he has contributed to deliberation processes addressing emerging technology policy issues. Joshua Yang is a software engineer and a postdoc researcher at ETH Zurich and experience in building digital systems for civic engagement and data-driven public discussion. His work focuses on developing scalable technical infrastructures that support large-scale participation and open collaboration. Tim contributes technical development and infrastructure design, helping translate deliberation methods into scalable digital systems. Yiting Lien works at the intersection of technology policy, research, and civic innovation. She has been engaged in international discussions on technology governance and participated in global forums exploring responsible technology development, including initiatives connected with the Oxford Internet Institute. Her work helps bridge research perspectives with practical civic technology experimentation. Irvin Chen is an active contributor in the Mozilla community and has long been involved in initiatives supporting open internet policies and digital democracy, including helping publish the Mandarin version of the Mozilla Digital ID White Paper in 2020. His other contributions include serving as a member of the Mozilla Reps Council from '18 to '20, working as the local project coordinator of the Common Voice OMSF program that helped shipping six Indigenous languages during '24–'25, and serving as a community liaison for the Mozila Taiwan Community and project manager of the Taipei Mozilla Community Space since 2011. His experience in connecting open-source communities and civic technology networks helps situate this project within broader global conversations on democratic technology. Beyond the core team, the project can mobilize a broader contributor network through the g0v civic tech community, which regularly collaborates through weekly meetings, hackathons, and open online workspaces. This ecosystem enables designers, developers, researchers, and civic organizers to collectively experiment with new democratic technologies. With interdisciplinary expertise, strong community infrastructure, and institutional support from organizations such as OCF, the team is well positioned to develop and test AI-assisted democratic tools in real-world policy discussions. ::: ### Your vision for impact and sustainability - What would success look like in 2-3 years? How will you measure success? :::info ### V2 草稿 In two to three years, success would mean that the platform has become a practical civic infrastructure for facilitating democratic dialogue in high-stakes policy environments. Taiwan faces a unique set of challenges that make such infrastructure increasingly important. The country is on the frontline of information operations and disinformation campaigns, while also experiencing growing political polarization, frequent natural disasters, and persistent geopolitical pressure. These pressures will become especially visible during Taiwan’s upcoming election cycles in 2026 and 2028, which are widely watched internationally as indicators of democratic resilience in the digital age. In this context, Taiwan’s elections increasingly function as a global testing ground for how democratic societies respond to disinformation, polarization, and rapidly evolving technology. Strengthening public deliberation capacity before and during these moments is therefore not only relevant for Taiwan, but also for other democracies facing similar challenges. Our vision is that the platform becomes a widely used tool for facilitating structured public discussions around complex public issues, including digital governance, disaster preparedness, and responses to emerging technologies. By enabling real-time opinion gathering and consensus discovery, the platform helps communities process disagreement constructively rather than allowing polarization to deepen. We would measure success through several indicators. First, adoption: at least 10–15 deliberative processes organized by civic groups, research institutions, or public agencies using the platform. Second, participation: several thousand participants contributing opinions and interacting with the system across multiple deliberation processes. Third, deliberation quality: measurable identification of consensus areas and validated summaries of public opinion generated through the platform’s iterative consensus discovery process. Fourth, ecosystem growth: an active open-source contributor community and reuse of the platform by organizations beyond Taiwan. Ultimately, success would mean demonstrating that AI-assisted deliberation tools can strengthen democratic resilience by helping societies navigate disagreement, misinformation, and crisis without fragmenting their democratic institutions. ::: - What do you need to make this project sustainable long-term? What are the biggest barriers to getting there? :::success ### V1 草稿 Long-term sustainability for this project depends on building the platform as open public infrastructure that can be widely reused across civic and policy contexts. First, the platform will be developed as free and open-source software (FOSS). This allows local civic organizations, NGOs, and local elected officials in Taiwan to adopt the system for their own public discussions and community consultations. By making the technology easy to deploy and adaptable to different contexts, the project aims to lower the barriers for grassroots actors to organize structured public deliberation. Second, sustainability requires collaboration with institutions involved in digital governance. We plan to explore partnerships with organizations such as TWNIC and other internet governance stakeholders to experiment with how AI-assisted deliberation tools can support discussions around internet governance, online harms, and digital policy. These collaborations can help integrate the platform into real policy and governance processes. Third, the project aims to diversify funding sources beyond a single grant. In Taiwan, there are emerging funding channels supporting civic technology and open-source projects, such as twgrant.tw and various public-interest technology foundations. By demonstrating the platform’s practical value through pilot deployments, the project will seek additional support from these funding programs and philanthropic initiatives. Despite these opportunities, several barriers remain. One challenge is ensuring that the technology is simple enough for non-technical organizations to use. Another is building trust in AI-assisted analysis within democratic processes, which requires strong transparency and human validation mechanisms. Addressing these challenges will be essential to ensuring long-term adoption. If successful, the platform could become a sustainable piece of democratic infrastructure that supports public dialogue and participatory governance in Taiwan and potentially other democratic societies. ::: - What would $50,000 unlock for your project? :::success ### V.1 草稿 A $50,000 grant would allow the project to move from experimental prototypes to a functional and deployable deliberation platform. The most important impact would be enabling a dedicated part-time developer to focus on building and maintaining the core system. Civic technology projects are often developed by volunteers with limited time and resources, which makes sustained development difficult. This funding would provide the capacity to implement core features, integrate existing deliberation tools, prioritize the development and improve the platform’s usability for real-world deployment. In addition to technical development, the funding would support the organization of pilot deliberation activities. These activities are essential for testing the platform in real public discussions and improving the system through practical feedback. The funding would help cover the basic costs required to host deliberative workshops and public forums, including facilitation, documentation, and technical support. Together, these two components—dedicated development capacity and real-world deliberation pilots—would allow the project to demonstrate how AI-assisted deliberation tools can function as practical democratic infrastructure rather than purely experimental technology. The results of these pilots would also generate open documentation and case studies that other civic organizations and governments could reuse when exploring AI-assisted democratic participation. ::: ### Your commitment to openness - How will you share your code, learnings, and data with others? Is there already a larger community that contributes to the project? :::success ### V1 草稿 The project is designed as an open civic technology initiative from the beginning. All core software developed through the project will be released as free and open-source software (FOSS) on GitHub under a permissive license, allowing other organizations to deploy, adapt, and extend the platform. In addition to releasing the source code, we will publish documentation that explains both the technical architecture and the deliberation methodology behind the system. This will include deployment guides, facilitation materials, and case studies from pilot deliberation processes. The goal is not only to share the technology itself, but also the practical knowledge needed to organize AI-assisted public deliberation. The project also builds on an existing community ecosystem. The vTaiwan process and the g0v civic tech community have developed a long tradition of open collaboration through weekly meetings, hackathons, and shared online workspaces. Developers, designers, researchers, and civic organizers regularly contribute ideas, prototypes, and improvements to civic technology tools. With support from organizations such as the Open Culture Foundation (OCF) and connections with international open internet communities, the project aims to contribute to a broader global conversation about open-source democratic technologies. By sharing code, documentation, and lessons learned openly, we hope the platform can be reused and adapted by civic organizations, governments, and researchers in Taiwan and beyond. ::: ### Democratic Impact - How does your technology actively advance democratic practice? What specific democratic outcomes (participation, transparency, accountability, civic space protection) does it enable? :::success ### V1 草稿 The platform advances democratic practice by strengthening four core dimensions of democracy: participation, transparency, accountability, and the protection of civic space. First, it expands meaningful participation by lowering the barriers to organizing deliberation processes. Traditional in-person deliberation workshops often require significant preparation, including agenda design, opinion collection, facilitation, and post-event analysis. These requirements can make deliberative processes difficult to organize for communities with limited resources. By automating parts of opinion gathering, clustering, and summarization, the platform allows civic organizations, local communities, and grassroots organizers to run structured deliberation processes with far fewer resources. This makes democratic deliberation accessible not only to governments or well-funded institutions, but also to local NGOs, community groups, and local elected representatives. Such capacity is especially important in situations involving crisis or collective decision-making under pressure. In Taiwan, for example, communities frequently need to coordinate responses to natural disasters such as typhoons. After a major typhoon last year, volunteers known as “Shovel Heroes” spontaneously organized themselves to travel to disaster areas and assist with cleanup and recovery efforts. Situations like this demonstrate how civic mobilization often emerges organically, but also how communities can benefit from better tools to quickly gather perspectives, coordinate priorities, and make collective decisions. Second, the platform increases transparency through real-time visualization of collective opinions. Rather than producing summaries only after discussions end, the system dynamically displays patterns of agreement and disagreement as participants interact with the platform. This immediate feedback allows participants to see how their views relate to others in the group, enabling more responsive and adaptive discussions. Third, the platform strengthens accountability in deliberation processes. Because the system records how opinions were collected, clustered, and validated through participant feedback loops, it creates a transparent trace of how consensus emerges. Policymakers, civic organizations, and participants can therefore understand how deliberation results were produced. Finally, the platform helps protect civic space in polarized or information-contested environments. By identifying areas of shared ground across diverse viewpoints, the system reduces the risk that discussions are dominated by the most vocal participants or fragmented by misinformation and polarization. Instead, it supports constructive dialogue and collective problem-solving. Together, these mechanisms transform public discussion into a more inclusive, transparent, and resilient democratic process. ::: - How is AI essential to achieving your democratic impact at scale or in ways that weren't previously possible? What would be lost without the AI component? :::success ### V1 草稿 AI is essential to this project because it enables the platform to transform unstructured conversations into structured democratic input in real time. The system uses AI at three critical stages of the deliberation process: capturing opinions, synthesizing collective signals, and validating emerging consensus. First, AI enables the system to extract votable statements directly from ongoing discussions. In physical deliberation workshops or hybrid meetings, participants often express ideas verbally rather than through written submissions. Large language models monitor the conversation and identify key claims or policy statements, which are then immediately presented to participants as short propositions that they can respond to using Polis-style voting (agree, disagree, or pass). This allows discussions to continue naturally without requiring facilitators to interrupt the conversation to manually record opinions. Second, AI helps transform the resulting voting data into a map of collective opinion. As participants respond to statements, the platform generates an agreement matrix that can be analyzed using dimensionality reduction and clustering techniques. These methods reveal patterns in how participants align with different viewpoints and allow facilitators and participants to see where groups converge or diverge. AI-assisted analysis helps interpret these clusters and generate readable descriptions of the viewpoints represented in each group. Third, AI supports an iterative consensus discovery process. The system generates summaries of each opinion cluster and presents them back to participants for validation. Participants can confirm whether the summary accurately reflects their views, allowing the system to progressively merge clusters and identify the widest possible areas of agreement. This process ensures that consensus is not imposed by the system but verified by participants themselves. Without AI, these processes would require extensive manual facilitation and post-event analysis. Facilitators would need to manually transcribe discussions, formulate statements, analyze voting data, and summarize viewpoints. This would make it extremely difficult to run the system in real time or to scale deliberation processes to larger groups. The AI component therefore acts as a deliberation infrastructure that allows structured democratic discussion to occur dynamically within live conversations. ::: - Provide a link to your code repository, demo, or website where we can see your project in action. - https://github.com/g0v/sensemaker-frontend - Provide a link to a 2-minute video (a Google Drive link, for example) showing your project and vision. This isn't about production value—we want to see the technology working and hear directly from your team about why it matters. (note, this video submission is not required, but strongly encouraged). - 這個蠻重要的，要上傳一個 2 分鐘的影片 demo 要用的東西 - pitch video - Notebook LLM 的影片生成功能 ## Action Item - 錄製影片： - 3/14 （六）晚上21:00 - 確認 Rosalind 是否可以出席 - Josh - Peter - Irvin - Tim