Attitudes Towards the Use of Artificial Intelligence in Healthcare: A Conjoint Analysis Survey in Singapore

Abstract Novel artificial intelligence (AI) is increasingly being used as a clinical decision support tool in healthcare. Despite AI’s growing use and improved quality in clinical decision-making, questions persist about potential harms and the lack of transparency in their algorithms. Implementation of AI technologies in healthcare must align with local norms and ethical standards if the purported benefits are to be achieved in specific contexts. Using choice-based conjoint analysis, we examined how Singaporeans evaluate different principles related to AI decision-making in healthcare. Six attributes were included: decision type, severity, explainability, quality, responsibility, and discrimination. Among 596 respondents, 51% reported fear that AI would unintentionally harm humans, while 87% feared increased surveillance. Responsibility had the highest relative importance (31.5%) for AI use in healthcare, followed by explainability (27.7%) and discrimination (15.9%). The most valued attribute levels were AI recommendations being as explainable as doctors’, doctors retaining responsibility for treatment decisions, and AI systems being tested for discrimination. Our respondents showed higher levels of trust, hope, and fear toward AI, with a stronger preference for explainability over doctor responsibility. While having AI outperform doctors in generating clinical suggestions is desirable, principles such as explainability, human oversight, and fairness are more important for the people whose lives AI will impact.