Participatory AI Incident Governance: Embedding Affected Communities in the Reporting and Oversight Loop

Gailmard et al. (2025) stress the need for scalable adverse event (AE) reporting systems in AI but note that current frameworks often underrepresent the perspectives of those most affected by AI failures. Drawing inspiration from the participatory process mapped in Franccois et al. (2025), this research proposes designing and piloting a decentralized, participatory incident reporting infrastructure. This system would empower not only developers and operators, but also downstream users, domain experts, and affected populations (e.g., patients, consumers, marginalized groups) to submit reports, validate incident severity, and help set reporting priorities. The project would also explore incentive mechanisms and the integration of community oversight panels. This approach fundamentally challenges the top-down, expert-driven reporting models, aiming instead for plural, accountable governance. The expected result is richer, more context-aware incident data, faster identification of emergent risks, and greater legitimacy of AI governance regimes.

References:

1. Known Unknowns and Unknown Unknowns: Designing a Scalable Adverse Event Reporting System for AI. Lindsey A. Gailmard, Drew Spence, Christie Lawrence, Daniel E. Ho (2025). Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society.
2. A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety. Camille Franccois, Ludovic Peran, Ayah Bdeir, Nouha Dziri, Will Hawkins, Yacine Jernite, Sayash Kapoor, Juliet Shen, Heidy Khlaaf, Kevin Klyman, Nik Marda, Marie Pellat, Deborah Raji, Divya Siddarth, Aviya Skowron, Joe Spisak, M. Srikumar, Victor Storchan, Audrey Tang, Jen Weedon (2025).