The advancement of large language models (such as ChatGPT, Gemini, or Llama) has opened new possibilities for simulating human behaviors in complex social environments. However, using these models as agents in simulations raises crucial questions about their ability to faithfully replicate human dynamics. This line of research evaluates the degree to which LLMs exhibit behaviors and psychological patterns similar to humans through psychometric tests and classic game theory experiments, such as the Prisoner’s Dilemma. Beyond individual responses, the analysis extends to whether these models reproduce the correlations and behavioral strategies that characterize people in diverse contexts. The objective is to establish a solid foundation that allows us to trust artificial intelligence for the rigorous study of social dynamics and decision-making.
DEFE Researcher: Samuel Martín Gutiérrez.