Reduced symptom reporting quality during human-chatbot versus human-physician interactions

Abstract

Artificial intelligence (AI) systems demonstrate remarkable potential in healthcare. However, their integration into clinical practice depends on overcoming barriers to user acceptance. While previous research has identified reservations regarding AI-generated medical advice, earlier stages of human-AI interaction, such as how symptoms are communicated, remain largely unexplored. In a preregistered between-subjects experiment (N = 500), participants were randomly assigned to provide simulated symptom reports for common medical conditions to either an AI chatbot or a human physician. Report quality was assessed using Large Language Model-based suitability metrics, which were validated by physicians. Participants who believed to interact with an AI (vs. physician) provided lower-quality symptom reports. Our findings indicate a bias at the input-stage of seeking medical advice in digital settings. This outcome could compromise the performance of AI systems in real-world applications, regardless of their actual capacity.