# Same Predictions, Different Harms: Causal Auditing of Patient World Models

arXiv: https://arxiv.org/abs/2610.05198 · HTML: https://mlsanity.com/paper/2610.05198/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Medical and surgical world models, one of 11 directions in World models.
- Builds on 15 of its 19 references that are in the map.
- First posted on arXiv 4 Oct 2026.

## Placed in

- World models / [Medical and surgical world models](https://mlsanity.com/wm/medical-and-surgical-world-models/)

## Builds on (most cited first, 15 of 15)

- [Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models](https://mlsanity.com/paper/1905.05824/) (2019)
- [An Automated Approach to Causal Inference in Discrete Settings](https://mlsanity.com/paper/2109.13471/) (2021)
- [Causes of Effects: Learning individual responses from population data](https://mlsanity.com/paper/2104.13730/) (2021)
- [Causal Inference Through the Structural Causal Marginal Problem](https://mlsanity.com/paper/2202.01300/) (2022)
- [What's the Harm? Sharp Bounds on the Fraction Negatively Affected by Treatment](https://mlsanity.com/paper/2205.10327/) (2022)
- [Hardness results for Multimarginal Optimal Transport problems](https://mlsanity.com/paper/2012.05398/) (2020)
- [Counterfactual Analysis in Dynamic Latent State Models](https://mlsanity.com/paper/2205.13832/) (2022)
- [Sharp sensitivity bounds for mediation under unmeasured mediator-outcome confounding](https://mlsanity.com/paper/1601.05155/) (2016)
- [Bounding the Probabilities of Benefit and Harm Through Sensitivity Parameters and Proxies](https://mlsanity.com/paper/2303.05396/) (2023)
- [Robust Counterfactual Inference in Markov Decision Processes](https://mlsanity.com/paper/2502.13731/) (2025)
- [Decomposition of Probabilities of Causation with Two Mediators](https://mlsanity.com/paper/2505.04983/) (2025)
- [Mediation Analysis for Probabilities of Causation](https://mlsanity.com/paper/2412.14491/) (2024)
- [Optimal Correlation for Bernoulli Trials with Covariates](https://mlsanity.com/paper/2412.03827/) (2024)
- [Bounding Probabilities of Causation with Partial Causal Diagrams](https://mlsanity.com/paper/2602.14503/) (2026)
- [General Probabilities of Causation with Causal Knowledge](https://mlsanity.com/paper/2608.12657/) (2026)

## Abstract (arXiv)

Patient world models used for clinical trial simulation can agree on transition kernels and arm-specific risks, yet disagree on the fraction of patients harmed by switching treatment---the counterfactual quantity that matters for intervention-aware reasoning. We audit this reliability gap in a two-stage shared-response SCM: a categorical intermediate health state is followed by common terminal care. Under independent stages, the sharp harm interval has closed-form endpoints for at most three intermediate states, with an exactness boundary at four states. Declared dependence and response-mismatch budgets yield calibrated outer bounds when stage independence or complete mediation is relaxed; in a symmetric three-state model the entire sensitivity frontier is sharp, [0,min1/2,1/3+(ρ+δ)/2], and shows exactly how budgets erase the gain over endpoint-only bounds. Two eight-variable response LPs propagate interventional uncertainty for finite-sample audits. Exact witnesses verify attainability. On public clinical simulators (EpiCare; sepsis), native configurations show little resolved stage dependence and no additional joint-compatibility gain over pairwise transport---honest negative results for reliability claims. All experiments are locally reproducible; guarantees remain conditional on the stated causal model. The results provide a concrete protocol for deciding when a patient world model is safe to trust for counterfactual harm.