Unravelling public preferences for the use of Artificial Intelligence mobile health applications in Australia

Objectives: To explore public opinion on the factors that drive the use of artificial intelligence (AI) mobile health (mHealth) applications for heart disease and mental health, with a particular emphasis on diagnostics and virtual health assistance (VHA). Methods: This study adopted a discrete choice experiment to investigate the preferences of the Australian general public for heart disease and depression. A total of 5 attributes were considered, including anonymized data sharing, human-AI interaction, accuracy of AI results, explanation of results provided by AI, and funding source. Mixed logit and latent class logit models were used to investigate potential preference heterogeneity among respondents. Results: Respondents (n = 1176) showed that AI accuracy was the most crucial factor in AI mHealth applications, followed by human doctor-AI interaction. Preferences for not sharing anonymized data were reported in depression, whereas there were no statistically significant results for heart disease. Results explained by AI and funding source were generally less important. Those who expressed fear of AI were less likely to opt for AI diagnostics and VHA in heart disease. Older adults (60+) were less likely to use AI in both health conditions, whereas younger adults (18-29) were more inclined to use VHA for heart disease. Conclusions: It is evident that beyond the technical feasibility of AI applications, there are nuanced differences in public preferences for AI mHealth applications in Australia. Understanding factors leading to these discrepancies would be valuable for ensuring safe and equitable acceptance and harnessing the full potential of AI in healthcare delivery and outcomes.