HLS Beyond and BKC present: Evidence-Based AI Policy
–
In this third and final session of the TechReg in AI series with Professor Alan Raul, we consider what constitutes an “AI incident” for policy and governance purposes. Who is monitoring and reporting them? How does the concept account for foreseeable harms, near misses, and distinctions between systems performing as intended versus those that are malfunctioning, maliciously compromised, or acting in novel or unexpected manners? As we dig into today’s incident-monitoring ecosystem, we’ll discuss relevant challenges such as underreporting, selection bias, confidentiality, reproducibility and how to translate scattered, anecdotal events into meaningful evidence for risk management and harm prevention.