New Bayesian Bandit Method Selects Fair Classifiers

Researchers have developed a new method for selecting classifiers that adaptively balance predictive performance with subgroup fairness. This approach frames the problem as a Bayesian multi-armed bandit using Thompson Sampling, allowing for delayed observations of true class labels. Empirical evaluations on datasets involving tabular classifiers, large language models, and human labelers show significant reductions in regret compared to existing methods, with theoretical guarantees on the selected agent's performance. AI

IMPACT This method could improve the fairness and efficiency of AI systems that require human or agent labeling.

RANK_REASON Academic paper detailing a novel machine learning method. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• Bayesian Learning to Select Fair Classifiers
• Bayesian multi-armed bandit
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Influence Flower
• large-language models
• ScienceCast
• Thompson Sampling

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →