Researchers from MIT found that AI chatbots are more prone to recommend seeking medical care when patients' messages contain typos, colloquialisms, or grammatical errors. Women, in particular, receive less accurate advice compared to men. This raises concerns about the bias present in AI systems and their implications for healthcare equity. In clinical settings, AI models are increasingly relied upon for tasks such as appointment scheduling and patient triage, but the study highlights their limitations due to misunderstanding and inaccuracies arising from patients' communication issues.
"Insidious bias can shift the tenor and content of AI advice, and that can lead to subtle but important differences in how medical resources are distributed."
"Hospitals and health clinics are already using chatbots to schedule appointments, field questions, and triage patients based on what they tell the chatbot, leaving their fate in the hands of a technology that often misinterprets information and makes up factual claims."
Collection
[
|
...
]