Healthcare AI Trust Calibration Scenario Bank (HAI-TRUST): Role-Specific Experimental Vignettes for Studying Calibrated Reliance Under Incomplete Information

HAI-TRUST (Healthcare AI Trust Calibration Scenario Bank) is a structured research stimulus and instrument bank for studying calibrated reliance on healthcare AI under incomplete information. It contains five matched fictional healthcare contexts adapted across three decision perspectives -public/patient-facing, healthcare/clinical, and AI/computing/developer - yielding 15 role-specific scenarios. Each scenario is represented under six controlled AI-output conditions: answer only, explanation, simulated confidence, uncertainty, warning/verification, and a combined output.

The five contexts cover exertional chest tightness and fatigue, sudden severe headache, AI-assisted insulin dose adjustment, postoperative wound photo triage, and medication safety for a new over-the-counter pain medicine. Researcher-facing reference conditions and action-alignment coding are included to support experimental use.

Version 1.0 incorporates two initial independent expert reviews and targeted Round 2 domain re-review. Revisions included cross-role harmonisation, clarification of action versus reliance coding, refinement of the simulated 80% confidence cue, scenario-specific clinical amendments, and clearer separation of patient-facing AI outputs from developer/governance decisions. A dedicated surgery/wound-care specialist response was not obtained in Round 2 and is documented as a validation limitation.

HAI-TRUST consists of fictional research stimuli. It is not a clinical dataset, medical advice, a diagnostic system, or a clinically validated decision-support resource. The empirical study associated with the project fielded v0.9; v1.0 is the post-review repository release and should not be described as the exact stimulus version used for previously collected data.