ScioMind: Cognitively Grounded Multi-Agent Social Simulation with Anchoring-Based Belief Dynamics and Dynamic Profiles

2026-05-13

AI Summary

This work addresses the limitations of current large language model (LLM)-driven multi-agent social simulations, which either lack cognitive grounding or rely excessively on unconstrained interactions, thereby failing to authentically capture human opinion dynamics. To overcome this, the authors propose a cognitively grounded multi-agent framework that innovatively integrates structured opinion dynamics with LLM-based reasoning. The framework incorporates personality-modulated anchoring strength in belief updating, a hierarchical memory architecture that supports experience-driven belief formation, and dynamically generated agent profiles based on corpus retrieval and generation. Experimental results demonstrate that the proposed approach significantly outperforms baseline methods across key dimensions—including polarization, diversity, extremization, and trajectory stability—yielding more realistic and behaviorally consistent simulations that align closely with empirical findings in political psychology.

Abstract

Large language model (LLM)-based multi-agent simulation offers a powerful testbed for studying social opinion dynamics. Yet current approaches often adopt two contrasting methods: either relying on fixed update rules with limited cognitive grounding or delegating belief change largely to unconstrained LLM interaction. We introduce ScioMind, a cognitively grounded simulation framework that bridges these paradigms by combining structured opinion dynamics with LLM-based agent reasoning. ScioMind integrates three key components: 1) a memory-anchored belief update rule that modulates susceptibility to influence via personality-conditioned anchoring strength; 2) a hierarchical memory architecture that supports persistent, experience-driven belief formation; and 3) dynamic agent profiles derived from a corpus-grounded retrieval pipeline, enabling heterogeneous personalities, rationales, and evolving internal states. We evaluate ScioMind on multiple case studies in a real-world policy debate scenario. Across metrics including polarisation, diversity, extremization, and trajectory stability, the proposed components consistently yield improvements in behavioural realism. In particular, dynamic profiles increase opinion diversity, memory and reflection reduce unstable oscillation, and anchoring induces persistent belief trajectories that better align with patterns reported in political psychology. These results suggest that our cognitively grounded design provides a novel solution to LLM-based social simulation that improves both stable and behavioural realism

Problem

• multi-agent simulation
• opinion dynamics
• cognitive grounding
• belief update
• LLM-based social simulation

Innovation

• cognitive grounding
• belief dynamics
• multi-agent simulation
• dynamic agent profiles
• memory-anchored update