Understanding the Influence of Adversarial Attack Concerns on Trust in AI-Enabled Wearable Health Monitoring Systems : A Qualitative Study

Sammanfattning

This thesis investigates how concerns about adversarial attacks influence trust in AI-enabled wearable health monitoring systems. Drawingon the Trust in AI framework, the study explores how perceptionsof reliability, vulnerability, transparency, explainability, and privacyshape stakeholder trust and decision-making in wearable healthcaretechnologies. A qualitative research design was adopted using semistructured interviews with two stakeholder groups: experts involvedin AI, IoT, cybersecurity, and wearable health systems, and tech-awareusers with experience using wearable devices such as smartwatchesand fitness trackers. The findings reveal that trust in wearable AI systems is conditional, context-dependent, and continuously calibratedrather than absolute. Participants expressed concerns regarding system reliability, adversarial manipulation, data privacy, and limitedexplainability, which influenced their willingness to rely on automatedhealth outputs. The study also found that users actively engaged inverification and cross-checking behaviours to evaluate system recommendations. While adversarial attacks were not always discussed inhighly technical terms by users, awareness of possible vulnerabilitiessignificantly shaped perceptions of trustworthiness and system safety.The study contributes to the growing literature on Trust in AI by providing a human-centred perspective on adversarial risks in wearablehealthcare technologies and highlights the importance of transparency,explainability, privacy protection, and realistic trust calibration in thedesign of trustworthy AI-enabled health monitoring systems.