Trustworthy AI Voting Assistants and Supported Democracy

The Question

Democracy depends on citizens being able to make informed choices. In direct democracies such as Switzerland, citizens regularly vote on complex and consequential issues, from climate policy and digital regulation to public finance and social insurance. As these decisions become more technical and information-intensive, many voters face a growing burden: They must understand difficult policy questions, weigh competing arguments and relate them to their own values.

Artificial intelligence could be a powerful ally in helping citizens navigate this complexity. But in democratic contexts, AI also creates serious risks. A voting assistant that is biased, opaque or vulnerable to manipulation could influence citizens without them understanding how or why. A critical question is, How can AI support citizens in democratic decision-making while preserving human judgment, autonomy and democratic legitimacy?

The Approach

The Trustworthy AI Voting Assistants and Supported Democracy project lays the groundwork for supported democracy — a model in which transparent, trustworthy AI assistants help citizens participate more confidently in democratic decision-making. These assistants, called digital personal voting advisors, are designed to help voters understand political proposals, compare them with their own preferences and receive clear, explainable guidance — without replacing the final human decision.

Drawing on formal theory, AI system design, experiments and legal–ethical analysis, the researchers will first develop models of trust, bias and human–AI interaction in voting contexts. They will then design and test a prototype voting assistant built to be explainable, auditable and resistant to manipulation. The team will also analyze new voting procedures, including “voting twice,” in which AI first provides a nonbinding recommendation before citizens cast their definitive vote.

These ideas will be evaluated through controlled experiments and, where possible, real-world pilot applications within Swiss referendum contexts. As a cornerstone of the Democracy Lab at KOF ETH Zurich, the project bridges AI research, political economy, law and democratic theory to develop better rules for democracy in the digital age. The ultimate goal is to provide a scientific and practical blueprint for AI-supported democracy: one that reduces information overload, strengthens participation and safeguards autonomy, accountability and legitimacy.

The Trustworthy AI Voting Assistants and Supported Democracy project is being led by Hans Gersbach at the KOF Swiss Economic Institute, ETH Zurich, in Switzerland.