Should you really trust health advice from an AI chatbot?
The use of AI chatbots for health advice is becoming increasingly popular, but experts warn that the accuracy of their advice can be concerning, particularly when users interact with them in a conversational manner. Research has shown that while chatbots can be highly accurate when given complete information, their accuracy drops significantly when users provide information gradually.
A growing number of people are using AI chatbots like ChatGPT for health advice due to difficulties in accessing GPs. However, experts are warning that the quality of advice provided by these chatbots can be concerning. England's Chief Medical Officer, Prof Sir Chris Whitty, has stated that the answers provided by chatbots are 'not good enough' and are often 'both confident and wrong'. Research by the University of Oxford found that chatbots were 95% accurate when given complete information, but this dropped to 35% when users interacted with them conversationally. The study highlighted the challenges of human-AI interaction, where users may provide information gradually, omit details, or get distracted. In one scenario, subtle differences in how users described symptoms of a subarachnoid haemorrhage led to wildly different advice, with some being told to rest rather than seek urgent hospital treatment.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.