Same Predictions, Different Harms: Causal Auditing of Patient World Models
Abstract
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,min{1/2,1/3+(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.
Related
- Empirical Auditing of Edge-Private Graph Generators
- PDECheck: A Diagnostic Software Suite for Auditing Learned PDE Simulators
- Integrating Network Psychometrics and LLMs: The Ising-Embeddings-Model applied to Reliability Auditing
- A lightweight blockchain-inspired hybrid intrusion detection system with ensemble learning for tamper-proof auditing
- Auditing Conformal Prediction for Language Models Under Inference-Time Model Shift
BibTeX
@misc{qi2026same,
title = {{Same Predictions, Different Harms: Causal Auditing of Patient World Models}},
author = {Yicheng Qi and Xiyi Xiong},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.05198},
eprint = {2610.05198},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.05198},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Same Predictions, Different Harms: Causal Auditing of Patient World Models
AU - Yicheng Qi
AU - Xiyi Xiong
PY - 2026
DA - 2026/10/04
PB - arXiv
DO - 10.48550/arxiv.2610.05198
UR - https://arxiv.org/abs/2610.05198
AB - 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,min{1/2,1/3+(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.
ER -