How AI Disclosure Shapes Patient Trust and Perceived Professionalism : A Quantitative Study of Health Communication in Germany and Uganda

Sammanfattning

This study examines whether disclosing AI involvement in patient-facing health communication affects cognitive trust, affective trust, perceived professionalism, and behavioral intentions, and whether these effects vary between Germany and Uganda and across levels of patient AI literacy. A between-subjects vignette experiment (N = 441) presented post-acute discharge instructions attributed either to a human physician or an AI system, with message content held constant across conditions. AI disclosure significantly reduced trust, perceived professionalism, and willingness to follow advice in both samples. AI-related unease did not significantly increase in either country, suggesting the disclosure effect operates through credibility evaluation rather than emotional discomfort. Parallel mediation analyses revealed that trust carried the significant indirect pathway to behavioral intentions in Germany, while perceived professionalism carried the significant indirect pathway in Uganda — demonstrating that the same disclosure label activates different evaluative concerns depending on institutional and cultural context. AI literacy did not moderate the effect in Germany but strengthened the negative response in Uganda, interpreted through sociotechnical systems theory as reflecting the institutional novelty of AI disclosure in a context where regulatory frameworks governing AI transparency are still developing. A formal condition × country interaction test confirmed that the overall disclosure effect did not differ significantly in magnitude between the two samples, though the explanatory pathways differed in theoretically meaningful ways. The study contributes to trust-in-automation theory, patient-centered communication theory, and sociotechnical systems theory by demonstrating that AI disclosure functions as a context-sensitive credibility signal rather than a universal emotional trigger, and that AI literacy in high-stakes communication domains may activate critical awareness rather than reassurance.