Three Lessons from Citizen-Centric Participatory AI Design
Drawing on three 2025 participatory workshops, this paper identifies key challenges in citizen-centric agentic AI design—specifically sustaining engagement, bridging expert-lay communication gaps, and translating speculative input into systems—arguing that reflexive, long-term participation is essential for responsible AI development.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a world where computers don't just follow orders but act like little helpers with their own personalities, making decisions and chatting with us. Scientists call these "AI agents." But here's the big question: Who gets to decide what these helpers should be like? Usually, a small group of tech experts in a lab designs them. But what if the people who actually have to live with these helpers—regular folks like students, retirees, and neighbors—could help design them too? This is the heart of "citizen-centric design." It's like asking the passengers to help design the bus, not just the driver. The goal is to make sure these future AI helpers fit our real lives, our values, and our communities, rather than just being cool gadgets that might accidentally cause trouble.
This paper is a report from a team of researchers who tried to build a bridge between the tech experts and the general public. They held three different workshops, bringing together regular people and specialists to dream up what these AI agents could look like. Instead of just talking, the participants got creative: they told stories about waking up to weird AI situations and built "lo-fi" prototypes using Legos, Play-Doh, and magazine cutouts to show what their ideal AI helpers would look like. Later, they brought these wild ideas to a group of experts to see if they could actually be built.
The team found that while this idea sounds great, it's actually quite tricky to pull off. They identified three main hurdles. First, getting people to stay involved is hard. It's like trying to keep a band together; if you only invite people for one song (one workshop), you miss out on the whole album. The researchers realized that to get real value, people need to be part of the whole journey, from the first spark of an idea to the final product, not just show up for a quick chat.
Second, everyone speaks a different language. The experts talk about "agents" and "algorithms," while regular folks might just think of "smart computers." The researchers found they had to be very careful not to accidentally trick the participants by using fancy words or showing them examples that made them think in a specific way. They had to admit that even the researchers themselves might have their own biases, so they had to be extra careful to listen without leading the conversation.
Finally, there's the challenge of turning those fun, creative drawings and stories into real, working technology. The participants came up with 14 different AI ideas, but figuring out which ones could actually work in the real world was a puzzle. The researchers suggest that the best way forward is to keep the conversation going, pairing those creative ideas with experts who can figure out the technical and ethical details. It's not a solved problem yet, but they believe that if we keep the public involved from start to finish, we can build AI that is both innovative and truly helpful for everyone.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.