Short Abstract

We present Schweizer Solargeschichten, an AI-powered chatbot that captures and contextualizes citizen perspectives on solar energy. Instead of extracting data, the system fosters dialogue, reduces misinformation, and enables inclusive participation in Switzerland’s energy transition.

Abstract

Schweizer Solargeschichten emerged from a practical dilemma in citizen science: residents were willing to discuss their experiences with solar energy, but refused to share financial data through traditional surveys or forms. Personal interviews proved too resource-intensive to scale. In response, we co-developed an AI chatbot with citizens to facilitate these conversations in an accessible, conversational format.

The chatbot invites users to share motivations, doubts, and lived experiences related to solar installations. It also responds—sometimes normatively—by clarifying misconceptions or introducing factual context. Technically, it relies on retrieval-augmented generation (RAG) to ground its answers in expert knowledge while avoiding hallucinations. Yet this design choice introduces a new layer of mediation: the AI does not only elicit information but also subtly frames what counts as “relevant” or “correct” knowledge.

Through over fifty dialogues, we found that participants appreciated the opportunity to express themselves anonymously and conversationally. However, their responses also reveal ambivalence toward AI as a conversational partner—both trusted and contested. The chatbot thus exposes an inherent tension between inclusion and control: enabling open participation while encoding expert perspectives in its very structure.

Our contribution reflects on this tension and the ethics of “normative design” in participatory AI. We argue that such systems can democratize engagement only when their epistemic framing is made visible and negotiable—turning AI not into an authority, but into a transparent, fallible interlocutor in collective learning about the energy transition.