Large Language Models (LLMs) are increasingly used not only as analytic tools but as socially situated agents that reason, interact, and generate behavior in simulated environments. This shift enables new forms of computational social science, where LLM-driven agents are used to model decision-making, social norms, cooperation, persuasion, and collective dynamics at scale. Such simulations offer promising opportunities to explore social processes and policy-relevant scenarios when human experiments are costly, infeasible, or ethically constrained. At the same time, they raise fundamental questions about social validity, interpretability, and the limits of using LLMs as proxies for human or institutional actors. The SocialLLM workshop brings together researchers working on social reasoning, social intelligence, and multi-agent behavior in LLM-based systems. The workshop focuses on core conceptual and methodological challenges, including when LLM-based social simulations are appropriate, what kinds of social phenomena they can meaningfully capture, and how their outputs should be evaluated and interpreted. Topics include norm-aware and ethical reasoning, causal and strategic interaction in multi-agent settings, alignment and safety in agentic systems, and the relationship between simulated and real-world social behavior. By combining keynote talks, research presentations, and interactive discussion, SocialLLM aims to foster a shared research agenda and build a community around principled, responsible, and impactful use of LLMs for social reasoning and simulation.

PDF