Researchers introduce SocaSim, an LLM-based multi-agent simulation framework designed to study Putnam's Social Capital Theory by bridging theoretical blueprints with simulated reality. The system builds an environment that integrates social network evolution, trust dynamics, and norm propagation, allowing agents to engage in repeated collective-action experiments.
- SocaSim reproduces Putnam's macro-level patterns and exhibits strong human-agent alignment at the group level.
- It traces micro-level causal pathways of social networks, trust, and norms via round-by-round simulations and counterfactual interventions.
- The framework enables process-level interpretability that traditional empirical methods lack due to limits on control and replication.
- The authors apply the three dimensions of the theory to analyze adaptation challenges in smart elderly care.
This research paradigm leverages LLM agents to bridge social science and computer science, offering a method to model core propositions that behavior-driven simulations typically miss.