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

arXiv: https://arxiv.org/abs/2605.20767 · HTML: https://mlsanity.com/paper/2605.20767/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 45 references that are in the map.
- First posted on arXiv 20 May 2026.

## Builds on (most cited first, 28 of 28)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [Whose Opinions Do Language Models Reflect?](https://mlsanity.com/paper/2303.17548/) (2023)
- [Out of One, Many: Using Language Models to Simulate Human Samples](https://mlsanity.com/paper/2209.06899/) (2022)
- [RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback](https://mlsanity.com/paper/2309.00267/) (2023)
- [Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies](https://mlsanity.com/paper/2208.10264/) (2022)
- [LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals](https://mlsanity.com/paper/2411.10109/) (2024)
- [Quantifying the Persona Effect in LLM Simulations](https://mlsanity.com/paper/2402.10811/) (2024)
- [Can LLM be a Personalized Judge?](https://mlsanity.com/paper/2406.11657/) (2024)
- [LLM Generated Persona is a Promise with a Catch](https://mlsanity.com/paper/2503.16527/) (2025)
- [PERSONA: A Reproducible Testbed for Pluralistic Alignment](https://mlsanity.com/paper/2407.17387/) (2024)
- [Benchmarking Distributional Alignment of Large Language Models](https://mlsanity.com/paper/2411.05403/) (2024)
- [The Challenge of Using LLMs to Simulate Human Behavior: A Causal Inference Perspective](https://mlsanity.com/paper/2312.15524/) (2023)
- [Virtual Personas for Language Models via an Anthology of Backstories](https://mlsanity.com/paper/2407.06576/) (2024)
- [SimulatorArena: Are User Simulators Reliable Proxies for Multi-Turn Evaluation of AI Assistants?](https://mlsanity.com/paper/2510.05444/) (2025)
- [Take Caution in Using LLMs as Human Surrogates: Scylla Ex Machina](https://mlsanity.com/paper/2410.19599/) (2024)
- [Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reasoning, and Conversation History](https://mlsanity.com/paper/2508.04826/) (2025)
- [Personas with Attitudes: Controlling LLMs for Diverse Data Annotation](https://mlsanity.com/paper/2410.11745/) (2024)
- [Deep Binding of Language Model Virtual Personas: a Study on Approximating Political Partisan Misperceptions](https://mlsanity.com/paper/2504.11673/) (2025)
- [Doubly-Robust LLM-as-a-Judge: Externally Valid Estimation with Imperfect Personas](https://mlsanity.com/paper/2509.22957/) (2025)
- [Should you use LLMs to simulate opinions? Quality checks for early-stage deliberation](https://mlsanity.com/paper/2504.08954/) (2025)
- [Identity, Cooperation and Framing Effects within Groups of Real and Simulated Humans](https://mlsanity.com/paper/2601.16355/) (2026)
- [Sleepless Nights, Sugary Days: Creating Synthetic Users with Health Conditions for Realistic Coaching Agent Interactions](https://mlsanity.com/paper/2502.13135/) (2025)
- [LLMs Can Infer Political Alignment from Online Conversations](https://mlsanity.com/paper/2603.11253/) (2026)

## Built on by (most cited first, 5 of 5)

- [Item-Mean Surrogates: Why Richer Persona Data Fail to Improve LLMs as Human Surrogates](https://mlsanity.com/paper/2608.29455/) (2026)
- [Consequential Behaviour and Representational Fairness in the Validation of Synthetic Research](https://mlsanity.com/paper/2609.27690/) (2026)
- [Passing Coarse Marginal Checks Can Be Cheap: Persona Mixtures and Imprecise Treatment-Response Estimates in an LLM Persona Panel](https://mlsanity.com/paper/2608.00979/) (2026)
- [Pretrained Persona Mixture Models and Tandem Models for Human Simulation](https://mlsanity.com/paper/2609.22607/) (2026)
- [Persona Following Is Not Selective Control: The Neutrality Gap in LLM User Simulation](https://mlsanity.com/paper/2609.35036/) (2026)

## Abstract (arXiv)

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.