Reframing AI in healthcare: An experimental study examining low-cost strategies to increase acceptance of AI-based diagnosis and cultural differences in response

Abstract

Artificial Intelligence (AI) is increasingly used in patient diagnosis. However, prior research found that people perceive AI-based diagnosis as dehumanizing and undignified, reducing willingness to adopt such systems. The present research examined low-cost strategies to increase public acceptance of AI diagnosis across four experiments (total N = 723). Experiments 1 and 2 tested whether adding a visual image influenced perceptions of dignity and acceptance. In Experiment 1, positive versus negative anthropomorphised images resulted in a greater acceptance of AI use in healthcare (p < 0.001). but did not affect other dignity perceptions (all ps > 0.5). Experiment 2 replicated this pattern: anthropomorphised, non-anthropomorphised, and no image conditions did not differ on dignity perceptions (all ps > 0.2), however people were less accepting of AI in healthcare when there was no image (p = 0.045). Experiments 3 and 4 examined informational framing. Experiment 3 found that providing information about the benefits of AI increased acceptance (p < 0.001), and perception of it as a decision-maker (p = 0.03). However, it had no other effects (p > 0.1). Experiment 4 found that cautionary information did not reduce acceptance or dignity perceptions (all ps > 0.09). A strong cultural effect emerged, in which Chinese participants reported higher AI acceptance and increased positivity across all measures, compared to British participants (all ps < 0.02). The findings indicate that visual aids and positive educational content enhanced acceptance of AI in healthcare, however this improvement did not generalise across all measures of dignity.