aiqbench analysis
This study demonstrates that statistical realism in LLM-simulated data is a poor proxy for treatment-effect accuracy, as optimizing for surface-level fidelity can actually degrade the ability to estimate causal impacts of interventions. The research reveals distinct error structures between mimicking human responses and predicting behavioral outcomes, warning that reliance on synthetic benchmarks may propagate hidden errors into real-world decision-making frameworks. Consequently, developers must adopt rigorous diagnostic validation protocols rather than assuming that realistic-looking simulations yield accurate policy insights.
Statistical realism is not evidence that LLMs can estimate treatment effects in social science experiments
arXiv:2604.02458v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly used to simulate human responses and estimate treatment effect of interventions when real-world experiments are costly or infeasible. The treatment-effect estimates are often evaluated using statistical realism, the degree to which simulated responses reproduce properties of observed human responses, although whether realism predicts treatment-effect accuracy remains unknown. Here we test this proxy relationship by jointly measuring statistical realism and treatment-effect accuracy on the same simulated responses in a cross-national experiment with 59,508 participants from 62 countries using three LLMs. The correlation between statistical realism and treatment-effect accuracy is weak, and optimizing for statistical realism can even worsen treatment-effect accuracy when selecting models, prompts, and target populations. The pattern replicates in two additional cross-national experiments spanning 12 and 27 countries with 20,785 participants. The divergence between the two reflects distinct error structures and is larger for behavioral outcomes, where models appear to extrapolate behavioral effects from attitudinal patterns. Because this divergence may remain hidden in deployment, errors can propagate into simulation-informed decisions. We introduce a diagnostic framework for LLM-generated synthetic data and discuss how treatment-effect validation should proceed under varying availability of experimental benchmarks. Simulated responses and simulated treatment effects are distinct estimation targets, and evidence for one does not certify the other.
Abstract:Large language models (LLMs) are increasingly used to simulate human responses and estimate treatment effect of interventions when real-world experiments are costly or infeasible. The treatment-effect estimates are often evaluated using statistical realism, the degree to which simulated responses reproduce properties of observed human responses, although whether realism predicts treatment-effect accuracy remains unknown. Here we test this proxy relationship by jointly measuring statistical realism and treatment-effect accuracy on the same simulated responses in a cross-national experiment with 59,508 participants from 62 countries using three LLMs. The correlation between statistical realism and treatment-effect accuracy is weak, and optimizing for statistical realism can even worsen treatment-effect accuracy when selecting models, prompts, and target populations. The pattern replicates in two additional cross-national experiments spanning 12 and 27 countries with 20,785 participants. The divergence between the two reflects distinct error structures and is larger for behavioral outcomes, where models appear to extrapolate behavioral effects from attitudinal patterns. Because this divergence may remain hidden in deployment, errors can propagate into simulation-informed decisions. We introduce a diagnostic framework for LLM-generated synthetic data and discuss how treatment-effect validation should proceed under varying availability of experimental benchmarks. Simulated responses and simulated treatment effects are distinct estimation targets, and evidence for one does not certify the other.
Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET)
arXiv:2604.02458 [cs.CY]
(or arXiv:2604.02458v3 [cs.CY] for this version)
Full text under the publisher's open licence · 226 words
Related coverage
Reviewing failure modes in science as we prepare for AI researchers
LessWrong · Oct 8, 2026
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face
AI Alignment Forum · Aug 3, 2026
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · Sep 18, 2026
Statistical Foundations of LLM-based A/B Testing: A Surrogacy Framework for Human Causal Inference
arXiv · Jun 17, 2026