Artificial intelligence outperformed humans in a series of emotional intelligence tests, a team of Swiss researchers said Thursday.
Teams from the universities of Geneva and Bern administered emotional intelligence tests, originally developed to assess humans, to six different AI systems.
The five-part test series evaluated empathy and the ability to identify, manage and regulate emotions. The tests include emotionally charged scenarios like determining the most appropriate response when an employee’s idea is stolen at work.
Overall, AI models gave correct answers 82% of the time, whereas the accuracy rate for human participants was only 56%.
In the second phase of the study, researchers asked ChatGPT to generate new emotional intelligence tests.
These AI-generated tests were evaluated by over 400 participants and were found to be as realistic as those not created by humans.
"This suggests that these AIs not only understand emotions but also grasp what it means to behave with emotional intelligence,” said one of the researchers, Marcello Mortillaro, in a statement.
Experts say the findings suggest AI, under expert supervision, could play a role in fields traditionally seen as uniquely human, such as education, coaching and conflict resolution.
The results of the study were published in the journal Communications Psychology.
