Realistic threat perception drives intergroup conflict: A causal, dynamic analysis using generative-agent simulations

2025-12-18

AI Summary

Empirical research on intergroup conflict faces persistent challenges in causal inference, ethical constraints on experimental manipulation, and scarcity of dynamic, multimodal data on threat perceptions. Method: We develop a large language model–driven generative agent–based virtual society to isolate and precisely manipulate realistic (material) and symbolic (identity-based) threats while dynamically tracking behavioral, linguistic, and attitudinal responses across time. Contribution/Results: We establish, for the first time, the direct causal primacy of realistic threat in driving hostile intergroup behavior. Symbolic threat exerts significant effects only in the absence of realistic threat and operates exclusively through ingroup bias mediation. Neutral cross-group contact buffers hostility, whereas structural inequality exacerbates majority-group animosity. Our framework yields an interpretable causal model that identifies structural fairness interventions and non-hostile intergroup contact as critical leverage points—providing a foundation for evidence-based conflict prevention policy modeling.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Knowledge Representation and Reasoning: Action, Change, and Causality
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Tracking, profiling, and countermeasures against them
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Human conflict is often attributed to threats against material conditions and symbolic values, yet it remains unclear how they interact and which dominates. Progress is limited by weak causal control, ethical constraints, and scarce temporal data. We address these barriers using simulations of large language model (LLM)-driven agents in virtual societies, independently varying realistic and symbolic threat while tracking actions, language, and attitudes. Representational analyses show that the underlying LLM encodes realistic threat, symbolic threat, and hostility as distinct internal states, that our manipulations map onto them, and that steering these states causally shifts behavior. Our simulations provide a causal account of threat-driven conflict over time: realistic threat directly increases hostility, whereas symbolic threat effects are weaker, fully mediated by ingroup bias, and increase hostility only when realistic threat is absent. Non-hostile intergroup contact buffers escalation, and structural asymmetries concentrate hostility among majority groups.

Problem

• Investigates how realistic and symbolic threats drive intergroup conflict dynamics
• Uses LLM agent simulations to overcome ethical and causal analysis limitations
• Tests causal effects of threats on hostility, bias, and conflict escalation

Innovation

• LLM-driven agent simulations model threat dynamics
• Manipulate realistic and symbolic threats independently
• Track internal states, actions, and language causally