Tuesday, November 24, 2026 2:30 pm
-
3:30 pm
EST (GMT -05:00)
Location
MC 5501
Speaker
Dr. Amir-Hossein Karimi
Title
When AI Says No: Algorithmic Recourse and the Path Toward Actionable AI
Abstract
Machine learning systems increasingly shape consequential decisions in lending, employment, healthcare, and public services. When a system makes an unfavourable decision, what can the person affected actually do to change the outcome? Algorithmic recourse studies how AI systems can provide actionable guidance in response to that question. This talk will introduce the foundations of algorithmic recourse and its connections to explainability, causal reasoning, and responsible AI. It will then discuss recent work on evaluating recourse methods with RecourseBench, designing recourse in multi-agent settings, and Farzan Mirshekari's work on model stealing through recourse queries.