Reinforcement Learning to Train Large Language Models to Explain Human Decisions

Researchers have explored using large language models (LLMs) as cognitive models that can both predict and explain human decision-making. By applying reinforcement learning techniques, they guided LLMs to generate reasoning traces for risky choices. This approach resulted in high-quality explanations and accurate predictions of human decisions. The study demonstrates the potential of LLMs to serve as interpretable cognitive models, bridging the gap between predictive performance and explanatory power in understanding human behavior.