The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study

• URL: http://arxiv.org/abs/2605.20767v1
• Date: Wed, 20 May 2026 06:09:41 GMT
• Title: The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study
• Abstract summary: Large language models (LLMs) show potential as simulators of human behavior.<n>LLMs can induce unintended shifts in latent user attributes, causing user drift.<n>We formalize the confounding or selection bias that can arise due to user drift.
• Score: 56.649987197786096
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) show potential as simulators of human behavior, offering a scalable way to study responses to interventions. However, because LLMs are trained largely on observational data, interventions in experiments with LLM-simulated synthetic users can induce unintended shifts in latent user attributes, causing user drift where the implicit simulated population differs across treatment conditions, potentially distorting effect estimates. We formalize the confounding or selection bias that can arise due to user drift and show how intervention-dependent shifts can inflate or attenuate observed differences in user responses under intervention. To diagnose confounding, we propose using negative control outcomes--attributes that should remain invariant under intervention--to identify distribution shifts across intervention conditions, providing evidence of user drift. To mitigate drift, we study adjusting the persona specification by eliciting additional confounders, finding that targeted, setting-relevant confounders can substantially reduce bias across survey-style and multi-turn agent evaluations.

Related papers

• Evaluating LLMs as Human Surrogates in Controlled Experiments [4.233590639227564]
  Large language models (LLMs) are increasingly used to simulate human responses in behavioral research.<n>It remains unclear when LLM-generated data support the same experimental inferences as human data.<n>We evaluate this by comparing off-the-shelf LLM-generated responses with human responses from a canonical survey experiment on accuracy perception.
  arXiv  Detail & Related papers  (2026-03-08T16:03:27Z)
• DSO: Direct Steering Optimization for Bias Mitigation [12.033608044339717]
  Generative models are often deployed to make decisions on behalf of users, such as vision-language models (VLMs) identifying which person in a room is a doctor to help visually impaired individuals.<n>Yet, VLM decisions are influenced by the perceived demographic attributes of people in the input, which can lead to biased outcomes like failing to identify women as doctors.<n>We propose Direct Steering Optimization (DSO) which uses reinforcement learning to find linear transformations for steering activations, tailored to mitigate bias while maintaining control over model performance.
  arXiv  Detail & Related papers  (2025-12-17T19:43:46Z)
• Can Finetuing LLMs on Small Human Samples Increase Heterogeneity, Alignment, and Belief-Action Coherence? [9.310571879281186]
  Large language models (LLMs) can serve as substitutes for human participants in survey and experimental research.<n>LLMs often fail to align with real human behavior, exhibiting limited diversity, systematic misalignment for minority subgroups, insufficient within-group variance, and discrepancies between stated beliefs and actions.<n>This study examines whether fine-tuning on a small subset of human survey data, such as that obtainable from a pilot study, can mitigate these issues and yield realistic simulated outcomes.
  arXiv  Detail & Related papers  (2025-11-26T09:50:42Z)
• Drift No More? Context Equilibria in Multi-Turn LLM Interactions [58.69551510148673]
  contexts drift is the gradual divergence of a model's outputs from goal-consistent behavior across turns.<n>Unlike single-turn errors, drift unfolds temporally and is poorly captured by static evaluation metrics.<n>We show that multi-turn drift can be understood as a controllable equilibrium phenomenon rather than as inevitable decay.
  arXiv  Detail & Related papers  (2025-10-09T04:48:49Z)
• Predicting Effects, Missing Distributions: Evaluating LLMs as Human Behavior Simulators in Operations Management [11.302500716500893]
  LLMs are emerging tools for simulating human behavior in business, economics, and social science.<n>This paper evaluates how well LLMs replicate human behavior in operations management.
  arXiv  Detail & Related papers  (2025-09-30T20:20:58Z)
• Prediction-Powered Causal Inferences [59.98498488132307]
  We focus on Prediction-Powered Causal Inferences (PPCI)<n>We first show that conditional calibration guarantees valid PPCI at population level.<n>We then introduce a sufficient representation constraint transferring validity across experiments.
  arXiv  Detail & Related papers  (2025-02-10T10:52:17Z)
• Estimating Individual Dose-Response Curves under Unobserved Confounders from Observational Data [6.166869525631879]
  We present ContiVAE, a novel framework for estimating causal effects of continuous treatments, measured by individual dose-response curves. We show that ContiVAE outperforms existing methods by up to 62%, demonstrating its robustness and flexibility.
  arXiv  Detail & Related papers  (2024-10-21T07:24:26Z)
• Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization [60.176008034221404]
  Direct Preference Optimization (DPO) and its variants are increasingly used for aligning language models with human preferences.<n>Prior work has observed that the likelihood of preferred responses often decreases during training.<n>We demonstrate that likelihood displacement can be catastrophic, shifting probability mass from preferred responses to responses with an opposite meaning.
  arXiv  Detail & Related papers  (2024-10-11T14:22:44Z)
• The Challenge of Using LLMs to Simulate Human Behavior: A Causal Inference Perspective [0.27624021966289597]
  Large Language Models (LLMs) have shown impressive potential to simulate human behavior.<n>We identify a fundamental challenge in using them to simulate experiments.<n>When LLM-simulated subjects are blind to the experimental design, variations in treatment systematically affect unspecified variables.
  arXiv  Detail & Related papers  (2023-12-24T16:32:35Z)
• Decoding Susceptibility: Modeling Misbelief to Misinformation Through a Computational Approach [61.04606493712002]
  Susceptibility to misinformation describes the degree of belief in unverifiable claims that is not observable. Existing susceptibility studies heavily rely on self-reported beliefs. We propose a computational approach to model users' latent susceptibility levels.
  arXiv  Detail & Related papers  (2023-11-16T07:22:56Z)
• Efficient Causal Inference from Combined Observational and Interventional Data through Causal Reductions [68.6505592770171]
  Unobserved confounding is one of the main challenges when estimating causal effects. We propose a novel causal reduction method that replaces an arbitrary number of possibly high-dimensional latent confounders. We propose a learning algorithm to estimate the parameterized reduced model jointly from observational and interventional data.
  arXiv  Detail & Related papers  (2021-03-08T14:29:07Z)
• Estimating the Effects of Continuous-valued Interventions using Generative Adversarial Networks [103.14809802212535]
  We build on the generative adversarial networks (GANs) framework to address the problem of estimating the effect of continuous-valued interventions. Our model, SCIGAN, is flexible and capable of simultaneously estimating counterfactual outcomes for several different continuous interventions. To address the challenges presented by shifting to continuous interventions, we propose a novel architecture for our discriminator.
  arXiv  Detail & Related papers  (2020-02-27T18:46:21Z)

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.