Patients rate AI as warmer, more empathic than doctors

(Web Desk) - A sweeping meta-analysis of 15 studies found patients consistently rate AI chatbots as warmer and more empathic than real clinicians.

In the largest dataset of 2,164 live patient interactions, and across multiple smaller studies, tools like ChatGPT edged out doctors and nurses by roughly two points on 10-point empathy scales.

Overall, artificial intelligence had a 73% chance of being judged more empathic in head-to-head comparisons.

“In text-only scenarios, AI chatbots are frequently perceived as more empathic than human HCPs,” noted the researchers.
The work, led by teams at the Universities of Nottingham and Leicester, undercuts a confident 2019 UK government assertion that empathy is an essential human skill that AI cannot replicate. In text, at least, it looks like AI can.

Across nine separate studies spanning cancer care, thyroid disease, mental health, autism, and general medical questions, ChatGPT-4 routinely outscored licensed clinicians.

On thyroid surgery questions, the AI’s empathy ratings sat 1.42 standard deviations above human surgeons. On mental health queries, 0.97 standard deviations higher than credentialed professionals.

When responding to patient complaints routed through hospital departments, the gap widened dramatically: 2.08 standard deviations in favor of the AI over patient-relations staff.

Crucially, this wasn’t just a patient-only effect. In a lupus set, physicians themselves rated the AI’s tone as more empathic than their peers’ responses to the same questions.

For multiple sclerosis, trained patient representatives using a validated empathy instrument also favored the AI over neurologists.

Dermatology was the outlier. In two studies focused on skin complaints, dermatologists beat ChatGPT-3.5 and Med-PaLM 2 on empathy. Researchers didn’t pinpoint why this specialty bucked the trend, but it’s a useful reminder that the effect isn’t universal.