For example, the artificial intelligence (AI) recognised the shortest emotional expressions in the millisecond range, such as a brief smile or an expression of disgust. Such "micro expressions" can escape therapists, as the University of Basel reported.

Overall, the AI judged facial expressions in psychotherapeutic situations as reliably as humans, as the statistical comparison in the study published in the journal "Psychopathology" showed.

Trained with 30,000 photos

The researchers trained a freely available artificial neural network with over 30,000 facial photos to recognise the six basic emotions of happiness, surprise, anger, disgust, sadness and fear. This trained AI then analysed video recordings of 389 therapy sessions of 23 borderline patients.

"We were surprised that relatively simple AI systems were able to interpret facial expressions so robustly in terms of their emotions," says lead author Martin Steppan.

Humans remain important

According to Steppan, AI could serve as a tool for psychotherapists in the future. According to the University of Basel, analysing and interpreting recorded facial expressions for research projects or psychotherapy is very time-consuming. Experts therefore often resort to less reliable indirect methods such as measuring the conductivity of the skin. AI could provide a remedy and thus develop into an important aid in therapy and research.

However, Steppan emphasised that the interpersonal aspect remains important. Therapeutic work is first and foremost relationship work and therefore remains a human domain - at least for the time being.