Actuaries Detect Hidden Bias In Insurance AI
Arthur Charpentier will present "Detecting Hidden Bias in Insurance AI Through Counterfactuals" at the III Congreso Universitario Internacional sobre Seguros y Reaseguros in Perú on December 15, 2025. He outlines using causal reasoning and sequential optimal-transport counterfactuals to decompose model disparities into direct and indirect components, even with complex mediators like prior claims or behavioral features. The talk aims to equip actuaries to diagnose hidden bias and improve governance and compliance.
Key Points
- 1Introduces sequential optimal-transport counterfactuals to estimate predictions if a sensitive attribute differed
- 2Highlights indirect discrimination risk from proxy or downstream variables in actuarial models
- 3Enables actuaries to decompose total disparities into direct and indirect components for governance
Scoring Rationale
Practical causal methods drive relevance, but novelty is limited and evidence relies on a single conference presentation.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems