Mitigating Patient and Consumer Safety Risks When Using Conversational Assistants for Medical Information:...

Area of Science:

• Human-Computer Interaction
• Medical Informatics
• Health Communication

Background:

• Conversational assistants like Siri and Alexa present safety risks when used for medical information.
• Prior studies highlight potential dangers in relying on AI for health advice.

Purpose of the Study:

• To evaluate methods for mitigating risks associated with AI-generated medical information.
• To assess the impact of disclaimers and query paraphrasing on user behavior.

Main Methods:

• Participants posed medical queries to AI assistants, receiving varied responses (paraphrased queries, disclaimers, or none).
• User actions and likelihood of acting on advice were recorded and analyzed.
• Post-interaction interviews were conducted.

Main Results:

• Users were more likely to follow advice when presented with a correct query paraphrase (P=.04).
• Disclaimer messages significantly increased the likelihood of users consulting a physician before acting (P=.001).

Conclusions:

• AI system designers should integrate disclaimers and query understanding feedback for medical advice.
• Unconstrained natural language input is unsuitable for AI systems providing medical guidance.