At UZH WP2, we’re laying the groundwork for how artificial intelligence—especially large language models—can genuinely strengthen democratic discussion in digital spaces. We’re the architects of the project, rather than simply adding AI on top of existing platforms, our job is to design the process itself. We’re focused on how technology can help people participate more equally, engage more thoughtfully, and make decisions that feel fair and informed.
Our current framework outlines five key phases of the deliberative process, each with clear roles for AI to play:
1) Participation: Who’s involved shapes everything that follows. That’s why we’re prioritizing inclusion and diversity from the start. AI tools could help identify gaps—groups or voices that are missing—and make targeted outreach easier. We’re also exploring low-barrier entry points, like gamified elements or simple incentives, to invite broader engagement.
2) Group-building: Once people join, the next step is helping them feel connected. Trust and shared norms don’t form automatically. So we’re designing AI-driven onboarding and warm-up exercises to build comfort and establish expectations.
3) Information phase: Deliberation falls apart without reliable information. To support meaningful input, we’re including tools for multi-perspective briefings—think expert summaries alongside stakeholder viewpoints. AI fact-checkers, argument suggestions, and visual maps of stakeholder positions could also helps participants stay grounded and respond with clarity.
4) Deliberation: In the core discussion phase, fairness and balance are key. AI moderation can help make sure everyone’s heard—especially quieter voices. Real-time feedback tools, like heatmaps and argument graphs, can show how views are shifting. Asynchronous features would allow people to take part on their own time, while an AI “devil’s advocate” could bring in alternative angles that keep the dialogue sharp.
5) Decision-making: Finally, deliberation needs to lead somewhere. AI can assist with mapping areas of agreement, managing ranked-choice voting, and collecting justifications or dissenting views to improve transparency. AI-generated summaries can help participants reflect on how the group’s thinking evolved—offering a clear, traceable path from dialogue to decision.
Throughout all five phases, our design emphasizes fairness, accessibility, democratic integrity and trust. That includes explainable AI, strong data protections, AI-quality check, and adaptive interfaces that work across languages and literacy levels. By focusing on the full arc of the process—not just the tools—we aim to build platforms that make digital democracy not just possible, but better.
By Jane Veri, UZH 5 May 2025
At UZH WP2, we’re laying the groundwork for how artificial intelligence—especially large language models—can genuinely strengthen democratic discussion in digital spaces. We’re the architects of the project, rather than simply adding AI on top of existing platforms, our job is to design the process itself. We’re focused on how technology can help people participate more equally, engage more thoughtfully, and make decisions that feel fair and informed.
Our current framework outlines five key phases of the deliberative process, each with clear roles for AI to play:
1) Participation: Who’s involved shapes everything that follows. That’s why we’re prioritizing inclusion and diversity from the start. AI tools could help identify gaps—groups or voices that are missing—and make targeted outreach easier. We’re also exploring low-barrier entry points, like gamified elements or simple incentives, to invite broader engagement.
2) Group-building: Once people join, the next step is helping them feel connected. Trust and shared norms don’t form automatically. So we’re designing AI-driven onboarding and warm-up exercises to build comfort and establish expectations.
3) Information phase: Deliberation falls apart without reliable information. To support meaningful input, we’re including tools for multi-perspective briefings—think expert summaries alongside stakeholder viewpoints. AI fact-checkers, argument suggestions, and visual maps of stakeholder positions could also helps participants stay grounded and respond with clarity.
4) Deliberation: In the core discussion phase, fairness and balance are key. AI moderation can help make sure everyone’s heard—especially quieter voices. Real-time feedback tools, like heatmaps and argument graphs, can show how views are shifting. Asynchronous features would allow people to take part on their own time, while an AI “devil’s advocate” could bring in alternative angles that keep the dialogue sharp.
5) Decision-making: Finally, deliberation needs to lead somewhere. AI can assist with mapping areas of agreement, managing ranked-choice voting, and collecting justifications or dissenting views to improve transparency. AI-generated summaries can help participants reflect on how the group’s thinking evolved—offering a clear, traceable path from dialogue to decision.
Throughout all five phases, our design emphasizes fairness, accessibility, democratic integrity and trust. That includes explainable AI, strong data protections, AI-quality check, and adaptive interfaces that work across languages and literacy levels. By focusing on the full arc of the process—not just the tools—we aim to build platforms that make digital democracy not just possible, but better.