Understanding, explaining, and utilizing medical artificial intelligence

Area of Science:

• Medical Artificial Intelligence
• Human-Computer Interaction
• Behavioral Economics

Background:

• Medical artificial intelligence (AI) offers cost-effectiveness and scalability, often surpassing human performance.
• Despite AI's potential, public reluctance to utilize these technologies persists.
• Understanding the drivers of this resistance is crucial for AI adoption in healthcare.

Purpose of the Study:

• To investigate the psychological factors underlying resistance to medical AI.
• To examine the role of perceived understanding of decision-making processes (both human and algorithmic) in AI utilization.
• To test interventions aimed at increasing acceptance of AI in healthcare.

Main Methods:

• Five pre-registered experiments (N=2,699) assessed understanding of human vs. algorithmic medical decision-making.
• Studies manipulated perceived understanding and measured willingness to use AI providers.
• A large-scale field study (N=14,013) evaluated intervention effectiveness via Google Ads for a skin cancer detection app.

Main Results:

• Participants demonstrated an illusory understanding of human medical decision-making.
• This illusory understanding led to a preference for human providers over AI, increasing reluctance to use AI.
• Interventions enhancing the perceived understanding of AI processes significantly increased willingness to utilize AI healthcare providers.
• Intervention effectiveness was confirmed in a real-world online setting.

Conclusions:

• Resistance to medical AI stems from a combination of the 'black box' problem and an overestimation of understanding human decision-making.
• Brief interventions focused on demystifying AI processes can effectively enhance user trust and willingness to adopt AI healthcare solutions.
• These findings have significant implications for the design and implementation of AI technologies in clinical practice and public health campaigns.