The problem with AI and 'empathy'

If technology redefines what our language means, it could also change our perceptions of ourselves.

- Sarah O'Connor

One after another, the “uniquely human” traits we once thought would remain untouched by the rise of the machines have started to look vulnerable after all. First it was creativity. Is empathy next?

If you have been reading the research of late, you could be forgiven for thinking so. In one study, a team of licensed healthcare professionals compared the responses of chatbots and real doctors with patient questions posed in an online forum. The chatbot responses were rated significantly higher, not just for quality, but for empathy.

In another piece of research, the large language models (LLMs) ChatGPT-4, ChatGPT-ol, Gemini 1.5 Flash, Copilot 365, Claude 3.5 Haiku and DeepSeek V3 outperformed humans on five standard emotional intelligence tests, achieving an average accuracy of 81 per cent, compared with the 56 per cent human average reported in the original validation studies.

This, the authors argued, added to “the growing body of evidence that LLMs like ChatGPT are proficient — at least on a par with, or even superior to, many humans — in socio-emotional tasks traditionally considered accessible only to humans”.

But before we conclude that AI is more empathic than humans, can I suggest that we stop for a moment and give ourselves a shake?