The result of #Aarau’s “Stadtidee”Participatory Budgeting came out today! Thanks to innovative digital voting methods, 10 points voting + Method of Equal Shares, the diverse interests of the city were reflected accurately & more citizens were satisfied 🙌🏾 https://t.co/Q3SvpUNFDT

Incredible. Nepal’s Gen Z elected their prime minister on #Discord. It shows that when politics is blocked, the internet becomes a public square where ppl can self-organise. It’s often the simple, familiar tools that they pick up. Sometimes, that’s enough to change a country.

First day here at the beautifully organised ACM @FAccTConference 2025 in Athens, discussing AI fairness in the birthplace of #democracy. Feels like we’re continuing a conversation that’s been going on for 2000 years. What’s fair? Who decides? Athens, AI, similar human challenges.

Thanks to @benjaminvonwyl at @swissinfo_en for covering our research on LLM Voting and AI in democratic decision-making! 🙏🏽 The goal is citizen-led democracy with AI support, not AI-led democracy. 💪🏽 https://t.co/FDHHMFGbYC

Sehr interessantes Experiment des Computational Social Science-Team @ETH : Wie würde Chat-GPT abstimmen, wenn man es abstimmen liesse? @joshuacyyang erklärt, wie KI die Demokratie verändern kann, auch wenn wir sie nicht wie Bürger:innen behandeln:

Here in San Jose for #AAAI Conf on AI, Ethics, & Society @AIESConf, presenting "LLM Voting: Human Choices & AI Collective Decision-Making" on Wed! tldr: We compare voting patterns of human & LLM agents to explore how AI can harm or support democracy. 👇🏾 https://t.co/4mex14yvie

Certainly the most special talk I’ve hosted. With @OpendataCH, we invited Taiwan’s former Digital Minister Audrey Tang to Zurich to share the @pluralitybook & the TW experience. Loads of great insights, engagement, & learning. Truly brings hope to our democratic future.🇨🇭🇹🇼 🖖🏽

Introducing the Probability ranking by Duncan & Parece! It uses probability to compare the expected number of medals a country should win based on its population to the total medals won. It highlights outperformance & offers a fairer comparison for countries of various pop size.

Using medals per capita allows us to better appreciate the countries that are punching above their weight. New Zealand 🇳🇿 is doing well! However, this ranking can be quite unstable, relying on a few athletes in small countries & making it difficult for larger countries to excel.

The commonly used ranking systems fail to show which countries are truly outperforming in Olympics & providing athletes w/ the best support. China's impressive gold count in the past games does not necessarily provide insights to the quality of sporting culture or infrastructure.

As we all know, different ranking systems can produce vastly different results. While news outlets in various countries present the Olympics medal table ordered by either only gold or all medals, both rankings can unfairly spotlight large countries 🇺🇸🇨🇳. https://t.co/Hxyh7uQVyH

We hope these findings offer some actionable insights for digital governance around the world, contributing to the development of fair and transparent collective decision-making processes for citizens! 🗳✊🏽 Check out this paper for more details:

While the “front-end” voting input format may not change the overall result as much, it influences voter perception of their contribution to the outcome. In contrast, the “back-end” voting aggregation method, though less visible to voters, is critical for a fair outcome. 10/

Overall, an effective voting system consists of a voting input that allows voters to freely express their choices, an aggregation method that supports proportionality, and an explanation that clearly explains both the mechanism and the outcome of the voting process. 9/

Findings on explanations: -Pre-hoc explanations are more effective in improving trust in algorithms than post-hoc explanations. - Increasing “model transparency” can contribute to an algorithm’s trustworthiness, but does not always lead to perceptions of fairness. 8/

Findings on perceptions: - When Greedy is used, voters tend to perceive an outcome as fairer & satisfying if it results in higher utility for them. - For MES outcomes, perceived fairness is not significantly tied to individual success in terms of utility. 7/

- Voters are not cost-conscious; thus, their selections, paired w/ Greedy, tend to fund costly projects. - The economical variants (cardinality utility) fund more inexpensive projects, where as MES strikes a balance. - MES shows lower variation across different voting inputs. 6/

Findings on aggregation: - Participants indicated a desire for a fair distribution across city districts and project categories. - Participants found the outcome aggregated using the Method of Equal Shares to be fairer than that of the conventional Greedy rule. 5/

- Cumulative voting (distributing points) can be a difficult voting method when the points are insufficient (5 in this case). - Select any number (SN) offers too much freedom & it is perceived as overwhelming. Yet, it generates outcomes less concentrated on a few projects. 4/

Findings on input formats: - Participants preferred voting input formats that are more structured & expressive over simpler formats (such as approval voting). - Being able to select first, then rank or distribute points, is highly recommended bc it offers some structure. 3/

Context: We conducted a lab experiment that invited 180 participants to vote in a fictional Participatory Budgeting program. We look into how voting input format & aggregation methods affect outcomes, not only in the distribution of the budget but in citizen perceptions. 2/