Investigating human and LLM psychology by prompting LLMs to play experimental economics games: Xie, Mei, Yuan, and Jackson in PNAS

 The great science fiction writer of my youth was Isaac Asimov, who not only wrote space opera (The Foundation Trilogy), but also wrote about intelligent robots, i.e. about robots with artificial general intelligence.  So, like you and me, they had complicated psychological lives, and one of the main characters in these stories was the robopsychologist  Dr. Susan Calvin (see e.g. the short story collection I, Robot, and also several of the robot novels).

I'm reminded of this by the several papers now reporting how large language models respond when asked to play games that have been used to study human behavior.  Those papers are framed as using LLMs to learn about the human behavior on which they were trained. But they can also be read as telling us about the 'psychology' of LLMs. Here's a good one from the PNAS. 

Xie, Yutong, Qiaozhu Mei, Walter Yuan, and Matthew O. Jackson. "Using large language models to categorize strategic situations and decipher motivations behind human behaviors." Proceedings of the National Academy of Sciences 122, no. 35 (2025): e2512075122. 

Abstract: By varying prompts to a large language model, we can elicit the full range of human behaviors in a variety of different scenarios in classic economic games. By analyzing which prompts elicit which behaviors, we can categorize and compare different strategic situations, which can also help provide insight into what different economic scenarios might induce people to think about. We discuss how this provides a step toward a nonstandard method of inferring (deciphering) the motivations behind the human behaviors. We also show how this deciphering process can be used to categorize differences in the behavioral tendencies of different populations.