Deutsch
 
Hilfe Datenschutzhinweis Impressum
  DetailsucheBrowse

Datensatz

DATENSATZ AKTIONENEXPORT

Freigegeben

Preprint

Playing repeated games with Large Language Models

MPG-Autoren
/persons/resource/persons241804

Schulz,  L       
Department of Computational Neuroscience, Max Planck Institute for Biological Cybernetics, Max Planck Society;

/persons/resource/persons276874

Coda-Forno,  J
Research Group Computational Principles of Intelligence, Max Planck Institute for Biological Cybernetics, Max Planck Society;

/persons/resource/persons139782

Schulz,  E
Research Group Computational Principles of Intelligence, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Externe Ressourcen

https://arxiv.org/pdf/2305.16867.pdf
(beliebiger Volltext)

Volltexte (beschränkter Zugriff)
Für Ihren IP-Bereich sind aktuell keine Volltexte freigegeben.
Volltexte (frei zugänglich)
Es sind keine frei zugänglichen Volltexte in PuRe verfügbar
Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar
Zitation

Akata, E., Schulz, L., Coda-Forno, J., Oh, S., Bethge, M., & Schulz, E. (submitted). Playing repeated games with Large Language Models.


Zitierlink: https://hdl.handle.net/21.11116/0000-000D-3A43-7
Zusammenfassung
Large Language Models (LLMs) are transforming society and permeating into diverse applications. As a result, LLMs will frequently interact with us and other agents. It is, therefore, of great societal value to understand how LLMs behave in interactive social settings. Here, we propose to use behavioral game theory to study LLM's cooperation and coordination behavior. To do so, we let different LLMs (GPT-3, GPT-3.5, and GPT-4) play finitely repeated games with each other and with other, human-like strategies. Our results show that LLMs generally perform well in such tasks and also uncover persistent behavioral signatures. In a large set of two players-two strategies games, we find that LLMs are particularly good at games where valuing their own self-interest pays off, like the iterated Prisoner's Dilemma family. However, they behave sub-optimally in games that require coordination. We, therefore, further focus on two games from these distinct families. In the canonical iterated Prisoner's Dilemma, we find that GPT-4 acts particularly unforgivingly, always defecting after another agent has defected only once. In the Battle of the Sexes, we find that GPT-4 cannot match the behavior of the simple convention to alternate between options. We verify that these behavioral signatures are stable across robustness checks. Finally, we show how GPT-4's behavior can be modified by providing further information about the other player as well as by asking it to predict the other player's actions before making a choice. These results enrich our understanding of LLM's social behavior and pave the way for a behavioral game theory for machines.