A Collaborative Approach Of AI Design Criteria With Non-AI Experts

Background: Extreme citizen science aims to take co-creation a step further, with citizens being an equal partner in research projects and being as much in control as experts. In our project (“It Takes a Village To Grow Old”) on designing an Artificial Intelligence (AI) art intervention in the public space to facilitate public familiarity, this necessitates (older) citizens to navigate challenging decisions regarding the design of an AI system. This work aims to: 1) raise awareness of ethical and practical considerations with a new method centered around the 7 Key Requirements from the EU Guidelines of Trustworthy AI (https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai), and 2) use
this approach for formulating design criteria for an AI system together with non-technical citizens.
Methods: Three sessions were conducted in 2023 with non-AI expert participants (n=28) having diverse backgrounds: academics (n=2), creative professionals and researchers (n=18), and older adults (n=8). Initially, participants engaged with a prototype of the "Story Wall" social storytelling application developed during the citizen science project. Subsequently, the 7 Key Requirements from the EU Guidelines were introduced, each serving as a discussion point. Participants generated pro and con arguments on post-it notes in small groups, followed by larger group discussions.
Findings: Engaging participants with limited understanding of AI in discussions about ethical considerations and consequences of the AI system proved insightful. Topics such as privacy and data governance, transparency, diversity, non-discrimination, fairness, and accountability sparked substantial discussion. Participants delved into the implications of these concepts, highlighting concerns and potential consequences associated with the design and implementation of the AI system. These outcomes have led to multiple changes
to the AI art prototype intervention.
Conclusion: The used methodology facilitated collaboration among diverse participant groups in designing an AI system for older non-AI experts. Participants demonstrated informed decision-making and comprehension of ethical considerations, guided by the EU Guidelines for Trustworthy AI. This structured framework showed to be useful in facilitating discussions and decision-making processes around AI interventions for public familiarity.