According to a new paper on arXiv, researchers are turning to large language models to better understand how people actually behave in the Prisoner's Dilemma. The classic game has long been used to study cooperation and conflict, but human choices often depart from what fully rational models would predict.
To address this gap, the study proposes using LLMs to model human decision-making in this setting. The abstract notes that human behavior consistently deviates from rational predictions, suggesting that LLMs may offer a more flexible way to capture the nuances of real-world choices.
While the abstract does not detail specific results, it highlights the potential of LLMs as tools for simulating social behavior. The work adds to a growing line of research applying language models to behavioral economics and game theory.