From Blueprint to Reality: Modeling and Applying Putnam's Social Capital Theory with LLM-based Multi-agent Simulations

2026-07-08 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Data Science & Analytics · Depth: Expert, extended

Summary

SocaSim is a novel LLM-based multi-agent simulation framework designed to model and apply Putnam's Social Capital Theory. It constructs an environment integrating social network evolution, trust dynamics, and norm propagation, where agents engage in repeated collective-action experiments. The framework successfully reproduces Putnam's macro-level patterns and demonstrates strong human-agent alignment (Pearson r = 0.974) at the group level. Applied to smart elderly care, simulations showed that increasing low-SES agents' initial trust boosted technology adoption by 15.4% and reduced decision contradictions by 25.5%. SocaSim enables tracing micro-level causal pathways of social network, trust, and norms, offering process-level interpretability for social science and computer science research.

Key takeaway

For research scientists investigating social dynamics or policy impacts, SocaSim offers a reproducible framework to model complex social theories. You can use its LLM-based agents to conduct counterfactual interventions, revealing causal mechanisms of social capital and informing targeted policy designs, such as improving smart elderly care adoption by boosting trust among disadvantaged groups. This approach provides process-level interpretability beyond correlational findings.

Key insights

SocaSim leverages LLM agents to simulate Putnam's Social Capital Theory, revealing micro-level causal pathways and macro-level patterns.

Principles

• Social networks, trust, and norms are interdependent.
• Trust accumulates through interaction, partially bridging SES gaps.
• Reciprocity norms sustain long-term cooperation.

Method

SocaSim employs Social Structure Trait (SST) for agent profiles, Belief-Desire-Intention (BDI) for real-time decisions, and Social Cognitive Memory (SCM) for adaptive learning in a two-phase (Proposal-Execution) round-based simulation.

In practice

• Simulate technology adoption challenges in specific populations.
• Test policy interventions, e.g., boosting trust for low-SES groups.
• Analyze micro-level causal chains of social dynamics.

Topics

• Social Capital Theory
• LLM Agents
• Multi-agent Simulation
• Elderly Care
• Trust Dynamics
• Social Networks

Best for: AI Scientist, Research Scientist

Related on AIssential

• Leadership as Coordination Control: Behavioral Signatures and the Recovery-Advantage Boundary in Multi-Agent LLM Teams — Leadership-like coordination control in LLM teams is contingent, adding value only under specific conditions.
• Social Simulation with LLMs - Fidelity in Applications (CFP @ COLM'26) [R] — LLM-based social simulations require rigorous evaluation and empirical grounding beyond initial demonstrations.
• Emergent Relational Order in LLM Agent Societies: From Collective Affect to Authority Stratification — LLM-based multi-agent simulations can spontaneously reproduce complex social structures like Differential Order through emotion-ethics-belief chains.
• Trust, Lies, and Long Memories: Emergent Social Dynamics and Reputation in Multi-Round Avalon with LLM Agents — LLM agents with cross-game memory develop role-conditional reputations and sophisticated deception strategies.
• Beyond Static Responses: Multi-Agent LLM Systems as a New Paradigm for Social Science Research — A six-tier framework maps LLM agent evolution from static tools to complex adaptive systems for social science research.
• SocialWise: LLM-Agentic Conversation Therapy for Individuals with Autism Spectrum Disorder to Enhance Communication Skills — SocialWise offers scalable, LLM-agentic conversation therapy for ASD, providing instant, structured feedback via a browser-based platform.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.