Decision Making under Miscalibration - 8 minute
summary
Defines a regret notion under bounded miscalibration and computes minimax optimal threshold formulas (closed form under several miscalibration metrics) that often differ from the classic therapeutic threshold j when cost asymmetry is large. Shows slight miscalibration can make j harmful and that the optimal adjustment is typically conservative toward fewer treatments; uses analytical derivations rather than just simulations and highlights remaining sample-driven calibration limits and worst-case nature of the solution.
AI-written summary · about
chapters (6)
related talks
45mAgreement and Alignment for Human-AI Collaborative Decision Making
48mIncentivizing Exploration among Behavioral Agents with Unbiased Histories
31mDecision theoretic foundations for human-AI collaboration
56mDecision Making in the Abscence of Ground Truth
31mAccuracy and Interpretability Through the Lens of Human-AI Teaming
35mML: Beyond confusion matrix, model performance analysis
41mFostering Trust via Explainable ML Inferences
22mInterface Design Optimization as a Multi-Armed Bandit Problem