Table 2.

Contextual factors for the 5 case studies.

	Environment	Actors	Framing factors	Causes of trust	Effects of trust
Diagnostic AIa (chest x-rays)	Image-driven diagnostics (radiology)	Medical professionals and AI system; patients to a limited extent	Discourse regarding job security and potential AI replacement	Accuracy, design transparency, and human competencies and virtues	Acceptance of systems by physicians, potentially at the cost of deskilling
Predictive AI (ICUb setting)	Clinical setting of an ICU	Physicians, nurses, and AI system; patients to a limited extent; and potentially caregivers and family members	Stressful situations potentially with a need to act under time pressure, risk of severe consequences, the need to synthesize too much information, and alert fatigue	Accuracy, transparency, and explainability; fairness; exclusion of harm; and rigorous testing (eg, in the form of an RCTc)	Acceptance and use of the system, potentially at the risk of erroneous clinical decisions following misleading predictions
Public health AI (disease outbreak model)	Nonclinical setting—publicly accessible web-based tool for the analysis of heterogeneous data	Developers, public health practitioners, policy makers, and the public	Stage and severity of disease outbreak; usability aspects (eg, intuitive interface or data visualization); and, potentially, antiscience sentiments and conspiracy theories with regard to the disease and health service providers	Historical accuracy and endorsement by authorities	Acceptance and use of the system by public decision makers (public health experts and policy makers)
Assistive AI (neurorehabilitation)	Clinical neurorehabilitation; elective use of different technologies for different activities, potentially every day	Patients and their caregivers and social circle, potentially including employers, engineers, and regulators	Clinical setting, science fiction literature and cinema, and public attitudes and policies on related technologies	Accuracy, privacy, lack of conflicts of interest, independence, long-term technical support, and user understanding of the underlying technology	Technology acceptance by users, health care professionals, and health care providers; potentially facilitated reimbursement and increased affordability and accessibility
Resource-allocating AI (predicting costs and needs)	Health service providers and health care system	The developing company providing the algorithm, the health system implementing it, the clinicians interacting with it, the patients having their care influenced by the algorithm, and regulatory bodies and algorithmic auditors	Media reporting on algorithms and their impact and theories of institutional trust	Reliability, accuracy, transparency, design, and model-centric explanations	Acceptance and use in health care systems

^aAI: artificial intelligence.

^bICU: intensive care unit.

^cRCT: randomized controlled trial.