PLACO: A Multi-Stage Framework for Cost-Effective Performance in Human-AI Teams

Short summary

PLACO is a research framework for optimizing human-AI team performance in classification tasks by combining human expertise with machine learning. The method uses Bayes rule to integrate deterministic human labels with probabilistic model predictions, leveraging calibrated confidence scores from both parties. This enables cost-effective improvements when neither humans nor models alone achieve acceptable performance.

• •Framework optimizes human-AI collaboration in classification tasks using Bayesian integration
• •Combines deterministic human labels with probabilistic model outputs via calibrated probabilities
• •Targets cost-effectiveness when humans and AI each have complementary strengths

Generated with AI, which can make mistakes.