Can AI teach medicine?

“Can AI do the teaching?” asked Nick Woznitza, consultant radiographer and clinical academic at University College Hospital London. “No. But . . .”

This was the subject of a panel session in February, at the Royal College of Radiologists’ conference on the future of AI in healthcare and the NHS.1 Medical education has long been built on a foundation of mentorship and hands-on learning—and there are still things that only a human teacher would pick up on, Woznitza told the audience.

“I have been a trainee,” he said. “Sometimes I didn’t even say, ‘I don’t understand’: it was the quizzical look on my face that told the tutor I needed help. AI can’t recognise that.”

Artificial intelligence (AI) may excel at pattern recognition and data processing, but it lacks the intuition and empathy that a skilled educator brings to the table, said Woznitza. “If we look at [diagnostic AI that examines images pixel by pixel], it can draw a line around a lesion, but it can’t explain why [that lesion is there],” he said. “It might give a probability score, but it definitely can’t explain it in another way if you don’t understand.” The ability to frame a concept differently depending on the learner’s needs is a skill that remains uniquely human.

James Wang, consultant oncologist at University College Hospital London, …