Simulating Society Requires Simulating Thought

TL;DR

This paper advocates for cognitively grounded reasoning in large language model agents to improve social simulation fidelity, introducing a new modeling paradigm and evaluation framework based on cognitive science principles.

Contribution

It introduces Generative Minds, a cognitive science-inspired paradigm for structured belief modeling, and RECAP, a benchmark for assessing reasoning fidelity in social simulations.

Findings

GenMinds enables structured belief representations in LLM agents.

RECAP effectively evaluates reasoning fidelity and causal traceability.

The approach shifts social simulation from surface mimicry to thought-based modeling.

Abstract

Simulating society with large language models (LLMs), we argue, requires more than generating plausible behavior; it demands cognitively grounded reasoning that is structured, revisable, and traceable. LLM-based agents are increasingly used to emulate individual and group behavior, primarily through prompting and supervised fine-tuning. Yet current simulations remain grounded in a behaviorist "demographics in, behavior out" paradigm, focusing on surface-level plausibility. As a result, they often lack internal coherence, causal reasoning, and belief traceability, making them unreliable for modeling how people reason, deliberate, and respond to interventions. To address this, we present a conceptual modeling paradigm, Generative Minds (GenMinds), which draws from cognitive science to support structured belief representations in generative agents. To evaluate such agents, we introduce the RECAP (REconstructing CAusal Paths) framework, a benchmark designed to assess reasoning fidelity via causal traceability, demographic grounding, and intervention consistency. These contributions advance a broader shift: from surface-level mimicry to generative agents that simulate thought, not just language, for social simulations.