When do people trust AI in healthcare? The role of risk and oversight communication

This study examined how task risk and oversight communication influence trust and behavioural intention towards patient-facing healthcare AI. Drawing on signaling theory, a 2 × 2 within-subject vignette experiment tested lower- versus higher-risk healthcare scenarios and AI-only versus oversight-present conditions. Participants (N = 254) evaluated eight hypothetical healthcare AI scenarios varying in risk level and oversight communication. Trust was lower in higher-risk scenarios, F (1, 249) = 78.43, p < 0.001, η2G = 0.031, and higher when oversight was communicated, F (1, 249) = 96.46, p < 0.001, η2G = 0.072. Similarly, behavioural intention decreased with higher risk, F (1, 249) = 43.24, p < 0.001, η2G = 0.018, and increased with oversight communication, F (1, 249) = 73.66, p < 0.001, η2G = 0.036. Risk × Oversight interactions were significant but small (η2G = 0.002 for both outcomes). Findings suggest that healthcare context and oversight communication shape initial evaluations of healthcare AI, with communicated oversight consistently enhancing trust and willingness to engage However, because risk conditions differed across health domains, conclusions about risk-dependent oversight effects warrant caution.