Decoding GPT’s hidden 'rationality' of cooperation

  • In current discussions on large language models (LLMs) such as GPT, understanding their ability to emulate facets of human intelligence stands central. Using behavioral economic paradigms and structural models, we investigate GPT’s cooperativeness in human interactions and assess its rational goal-oriented behavior. We discover that GPT cooperates more than humans and has overly optimistic expectations about human cooperation. Intriguingly, additional analyses reveal that GPT’s behavior isn’t random; it displays a level of goal-oriented rationality surpassing human counterparts. Our findings suggest that GPT hyper-rationally aims to maximize social welfare, coupled with a strive of self-preservation. Methodologically, our esearch highlights how structural models, typically employed to decipher human behavior, can illuminate the rationality and goal-orientation of LLMs. This opens a compelling path for future research into the intricate rationality of sophisticated, yet enigmatic artificial agents.

Download full text files

Export metadata

Author:Kevin BauerORCiDGND, Lena Liebich, Oliver HinzORCiDGND, Michael KosfeldORCiDGND
Series (Serial Number):SAFE working paper (401)
Place of publication:Frankfurt am Main
Document Type:Working Paper
Year of Completion:2023
Year of first Publication:2023
Publishing Institution:Universitätsbibliothek Johann Christian Senckenberg
Release Date:2023/10/06
Tag:cooperation; economic rationality; goal orientation; large language models
Edition:September 19, 2023
Page Number:34
We gratefully acknowledge research support from the University of Mannheim, the Leibniz Institute for Financial Research SAFE, and the Goethe University Frankfurt.
Institutes:Wirtschaftswissenschaften / Wirtschaftswissenschaften
Wissenschaftliche Zentren und koordinierte Programme / House of Finance (HoF)
Wissenschaftliche Zentren und koordinierte Programme / Center for Financial Studies (CFS)
Wissenschaftliche Zentren und koordinierte Programme / Sustainable Architecture for Finance in Europe (SAFE)
Dewey Decimal Classification:3 Sozialwissenschaften / 33 Wirtschaft / 330 Wirtschaft
Licence (German):License LogoDeutsches Urheberrecht