Morality
An LLM-based agent simulation framework modeling prehistoric hunter-gatherer societies
This project was collaborated with my collegues in BIGAI.
Abstract
The evolution of morality presents a puzzle: natural selection should favor self-interest, yet humans developed moral systems promoting cooperation. We introduce an LLM-based agent simulation framework modeling prehistoric hunter-gatherer societies with agents of varying moral dispositions based on expanding circles of concern. Our findings reveal that kinship-focused morality dominates when moral perception is limited, while higher communication costs favor selfish strategies. The framework demonstrates how moral dispositions interact with environmental pressures and cognitive constraints to produce different evolutionary outcomes. Our approach offers three key contributions: methodologically, it enables psychologically realistic evolutionary simulations; theoretically, it reveals the critical role of cognitive factors in moral evolution; and empirically, it provides evidence for how different moral orientations succeed under varying conditions. This work establishes a novel complementary paradigm to traditional evolutionary biology and anthropological research for investigating complex social evolution.