Interpretable AI models reveal human decision-making patterns in games

Researchers have explored interpretable machine learning models to understand human decision-making in games, specifically focusing on deviations from predicted independent and identically distributed (i.i.d.) play. By analyzing over 84,000 decisions from a zero-sum card game, the study compared black-box sequence models like LSTMs against transparent alternatives. The findings indicate that predictable behaviors such as repeating or avoiding past actions, particularly concerning players' management of their own recent histories, account for the majority of the strategically relevant signal, with frequency tracking adding minimal predictive value. AI

IMPACT Provides insights into human behavior modeling, potentially improving AI's ability to predict and interact with human decision-making.

RANK_REASON The cluster contains a single academic paper published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• arXivLabs
• Ewa
• Hugging Face
• lasso
• long short-term memory
• O'Neill
• ScienceCast

• paper
• other

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