Investigating Prosocial Behavior Theory in LLM Agents under Policy-Induced Inequities

TL;DR

The paper tackles how prosocial behavior emerges and decays in LLM-based agents operating in socially structured environments under policy-induced inequities. It introduces ProSim, a modular simulation framework with four modules—Individual, Scenario, Interaction, and Intervention—to model trait-driven decisions, repeated social interactions, and normative policies. Across three progressive studies, the work shows that LLM agents can exhibit human-like prosociality, enforce fairness norms through third-party punishment, and experience erosion of prosociality when policies introduce inequities that spread via social contagion. The findings illuminate how institutional dynamics shape pro-social norms in AI-driven societies and offer guidance for mitigating norm erosion through targeted interventions, with implications for AI governance and social simulation research.

Abstract

As large language models (LLMs) increasingly operate as autonomous agents in social contexts, evaluating their capacity for prosocial behavior is both theoretically and practically critical. However, existing research has primarily relied on static, economically framed paradigms, lacking models that capture the dynamic evolution of prosociality and its sensitivity to structural inequities. To address these gaps, we introduce ProSim, a simulation framework for modeling the prosocial behavior in LLM agents across diverse social conditions. We conduct three progressive studies to assess prosocial alignment. First, we demonstrate that LLM agents can exhibit human-like prosocial behavior across a broad range of real-world scenarios and adapt to normative policy interventions. Second, we find that agents engage in fairness-based third-party punishment and respond systematically to variations in inequity magnitude and enforcement cost. Third, we show that policy-induced inequities suppress prosocial behavior, propagate norm erosion through social networks. These findings advance prosocial behavior theory by elucidating how institutional dynamics shape the emergence, decay, and diffusion of prosocial norms in agent-driven societies.

Figures (8)

• Figure 1: Overview of ProSim. ProSim models the emergence and evolution of prosocial behavior in LLM agents through four modules: Individual Simulation assigns agents diverse demographic and psychological traits; Scenario Simulation presents six tasks spanning key prosocial behaviors; Interaction Simulation enables social learning within a small-world network; and Intervention Simulation introduces policies and inequalities to test behavioral sensitivity and norm dynamics.
• Figure 2: Left: Prosocial intention scores of LLM agents and human participants in six prosocial scenarios. Five-pointed stars indicate the average intention scores. Right: Aggregated prosocial intention scores averaged across all six scenarios.
• Figure 3: SHAP analysis of the predictive contribution of psychological traits to prosocial intentions.
• Figure 4: Behavioral shifts under policy interventions. Each heatmap shows the relative change in prosocial intention under one of four policy framings compared to the baseline condition. The rightmost bar chart counts the positive improvements.
• Figure 5: The third-party punishment rates under different allocation plans and penalty costs.