Interpreting and Steering LLM Agents for Social Simulations

• URL: http://arxiv.org/abs/2609.16436v1
• Date: Mon, 14 Sep 2026 23:23:23 GMT
• Title: Interpreting and Steering LLM Agents for Social Simulations
• Abstract summary: Simulations based on large language models (LLMs) have proven to be powerful for understanding human behavior.<n>However, LLMs are ultimately black boxes based on deep neural networks which limits their value for social science.<n>Here, we demonstrate how the black box could be opened up to further enrich LLM-based simulations.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Simulations based on large language models (LLMs) have proven to be powerful for understanding human behavior, making them valuable additions to the social scientific toolkit. However, LLMs are ultimately black boxes based on deep neural networks which limits their value for social science. This is because of a lack of (i) interpretability: i.e. the ability to assign clear mechanisms driving observed behavior; and a lack of (ii) steerability: i.e. the ability to mute or amplify specific theoretically meaningful mechanisms of action to drive specific model behavior. Here, we demonstrate how the black box could be opened up to further enrich LLM-based simulations. Specifically, we compare three types of methods: (1) prompt-based manipulation, (2) SAE-derived feature steering, and (3) probe-based direction steering and examine their utility for LLM-based social scientific simulations. We do so by interpreting and steering two foundational components of human behaviors, namely preferences (risk attitudes, altruism) and capabilities (divergent creativity, product innovation), operationalized using four classic economic and creative tasks implemented as natural-language interactions. Overall, our results show that SAE- and probe-based techniques often outperform basic prompt-based methods for steering LLM agents, although this advantage depends on the specific prompting strategy involved. Together, SAEs and probes constitute an effective pipeline for social scientists seeking to interpret and steer agents in social simulations: SAEs decompose agents' internal representations into human-readable features, after which probes can reliably shift agents' behaviors in specified directions. We discuss implications of these methods for future work using LLM agents for social scientific simulations.

Related papers

• LLM-Driven Personalities for Decision Making in Emergency Simulations [1.2248975590229072]
  We investigate the use of Large Language Models to drive decision-making in virtual humans within a simulated evacuation scenario.<n>Our goal is to evaluate how personality, expressed through language-based prompts, influences both individual behaviors and collective simulation outcomes.
  arXiv  Detail & Related papers  (2026-06-30T02:15:22Z)
• Simulate, Reason, Decide: Scientific Reasoning with LLMs for Simulation-Driven Decision Making [50.05403024674141]
  MechSim is a mechanism-grounded neuro-symbolic reasoning framework for executable scientific simulators.<n>Our framework represents simulators through a shared structured schema capturing assumptions, variables, mechanism dependencies, and execution traces.<n>We evaluate our approach across multiple high-stakes domains and show that it improves mechanism-level explanation quality, simulator analysis, and downstream decision-making reliability.
  arXiv  Detail & Related papers  (2026-06-03T06:36:51Z)
• Position: AI Agents Are Not (Yet) a Panacea for Social Simulation [62.891898598784415]
  Recent advances in large language models (LLMs) have spurred growing interest in using LLM-integrated agents for social simulation.<n>This position paper argues that LLM-based agents are not (yet) a panacea for social simulation.
  arXiv  Detail & Related papers  (2026-02-19T06:35:07Z)
• Leveraging LLM-based agents for social science research: insights from citation network simulations [132.4334196445918]
  We introduce the CiteAgent framework, designed to generate citation networks based on human-behavior simulation.<n>CiteAgent captures predominant phenomena in real-world citation networks, including power-law distribution, citational distortion, and shrinking diameter.<n>We establish two LLM-based research paradigms in social science, allowing us to validate and challenge existing theories.
  arXiv  Detail & Related papers  (2025-11-05T08:47:04Z)
• Social Simulations with Large Language Model Risk Utopian Illusion [61.358959720048354]
  We introduce a systematic framework for analyzing large language models' behavior in social simulation.<n>Our approach simulates multi-agent interactions through chatroom-style conversations and analyzes them across five linguistic dimensions.<n>Our findings reveal that LLMs do not faithfully reproduce genuine human behavior but instead reflect overly idealized versions of it.
  arXiv  Detail & Related papers  (2025-10-24T06:08:41Z)
• LLM-Based Social Simulations Require a Boundary [3.351170542925928]
  This position paper argues that large language model (LLM)-based social simulations should establish clear boundaries.<n>We examine three key boundary problems: alignment (simulated behaviors matching real-world patterns), consistency (maintaining coherent agent behavior over time), and robustness.
  arXiv  Detail & Related papers  (2025-06-24T17:14:47Z)
• SocialEval: Evaluating Social Intelligence of Large Language Models [70.90981021629021]
  Social Intelligence (SI) equips humans with interpersonal abilities to behave wisely in navigating social interactions to achieve social goals.<n>This presents an operational evaluation paradigm: outcome-oriented goal achievement evaluation and process-oriented interpersonal ability evaluation.<n>We propose SocialEval, a script-based bilingual SI benchmark, integrating outcome- and process-oriented evaluation by manually crafting narrative scripts.
  arXiv  Detail & Related papers  (2025-06-01T08:36:51Z)
• Can LLM Agents Simulate Multi-Turn Human Behavior? Evidence from Real Online Customer Behavior Data [46.65903742010956]
  We present the first large-scale quantitative evaluation of state-of-the-art LLMs' ability to accurately simulate human behavior.<n>Using real-world data from 31,865 online shopping sessions containing 230,965 user actions, our evaluation reveals that prompt-based LLMs achieve only 11.86% accuracy in generating human actions.<n>We also showcase that strategies as simple as fine-tuning LLMs on real human click-through data augmented with synthesized reasoning traces can greatly enhance models' performance.
  arXiv  Detail & Related papers  (2025-03-26T17:33:27Z)
• A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models [50.34089812436633]
  Large Language Models (LLMs) have transformed natural language processing, yet their internal mechanisms remain largely opaque.<n> mechanistic interpretability has attracted significant attention from the research community as a means to understand the inner workings of LLMs.<n>Sparse Autoencoders (SAEs) have emerged as a promising method due to their ability to disentangle the complex, superimposed features within LLMs into more interpretable components.
  arXiv  Detail & Related papers  (2025-03-07T17:38:00Z)
• Shall We Team Up: Exploring Spontaneous Cooperation of Competing LLM Agents [18.961470450132637]
  This paper emphasizes the importance of spontaneous phenomena, wherein agents deeply engage in contexts and make adaptive decisions without explicit directions. We explored spontaneous cooperation across three competitive scenarios and successfully simulated the gradual emergence of cooperation.
  arXiv  Detail & Related papers  (2024-02-19T18:00:53Z)
• LLM-driven Imitation of Subrational Behavior : Illusion or Reality? [3.2365468114603937]
  Existing work highlights the ability of Large Language Models to address complex reasoning tasks and mimic human communication. We propose to investigate the use of LLMs to generate synthetic human demonstrations, which are then used to learn subrational agent policies. We experimentally evaluate the ability of our framework to model sub-rationality through four simple scenarios.
  arXiv  Detail & Related papers  (2024-02-13T19:46:39Z)
• Systematic Biases in LLM Simulations of Debates [12.933509143906141]
  We study the limitations of Large Language Models in simulating human interactions.<n>Our findings indicate a tendency for LLM agents to conform to the model's inherent social biases.<n>These results underscore the need for further research to develop methods that help agents overcome these biases.
  arXiv  Detail & Related papers  (2024-02-06T14:51:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.