Unassigned · 2026-10-06
Same Predictions, Different Harms: Causal Auditing of Patient World Models
Yicheng Qi, Xiyi Xiong
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors audit “patient world models” for clinical trial simulation that match transition kernels and arm-specific risks yet may disagree on counterfactual harm (patients harmed by switching treatment). They analyze a two-stage shared-response SCM with categorical intermediate state and common terminal care, deriving sharp harm intervals (exact for up to three intermediate states) and sensitivity bounds under declared stage dependence and response-mismatch budgets. Experiments on public simulators find “honest negative” results for reliability claims.
Why it matters
Validation that two simulators agree on transition frequencies and treatment-specific risks is insufficient for high-stakes intervention-aware reasoning, because counterfactual harm depends on how potential outcomes coexist at the individual level. The paper provides an auditing protocol with sharp/outer bounds and uncertainty propagation for finite-sample audits, clarifying when agreement is reliable for harm quantities and when compatibility gaps remain.
Method
- Model patient world models as a two-stage shared-response SCM: an initial randomized treatment maps to a categorical intermediate state, followed by common terminal care with shared binary response mechanisms used across states reached by counterfactual trajectories.
- Derive sharp harm intervals for the harm probability under stage independence (including a cut-envelope approach with exactness for n≤3 intermediate states, and a dominant pooled state regime).
- Relax causal assumptions using budgets bounding stage dependence (ρ) and response mismatch (δ); propagate finite-sample uncertainty via two eight-variable response LPs and use calibrated outer bounds conditional on the stated causal model.
Limitation
Guarantees and confidence intervals remain conditional on the stated causal model and feasibility of the specified primitives. The paper also states that the audit does not establish coverage for unknown realized harm or for unmeasured confounding.
Abstract (from 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, $[0,\min\{1/2,1/3+(\rho+\delta)/2\}]$, 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 papers
- Repeated-Measure Leakage, Distribution Shift, and Reliability under Partial Observation in Patient World Models
- PatientWM: Towards Generalizable Clinical World Models via Unified Patient States
- CaMeWM: A Causal Medical World Model for Generative Tumor Evolution and Treatment Planning
- EHR2Trace: Auditable EHR Data Infrastructure for Patient World Models and Clinical Agents
- From Prediction to Action: Population Health World Model for Digital Medicine
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-04 · Indexed by Pub-AI: 2026-10-06 UTC
Same Predictions, Different Harms: Causal Auditing of Patient World Models
Preprint posted: 2026-10-04 · Original record ↗
- arxiv
- 2610.05198
Explicit source versions (1)
- v1 · 2026-10-04
- Preprint available
Source date: 2026-10-04 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
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, $[0,\min\{1/2,1/3+(ρ+δ)/2\}]$, 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.