AI can read emotions as well as a therapist does
A groundbreaking AI model has demonstrated the ability to accurately recognize and interpret facial expressions, conveying emotions, even the most fleeting ones. This could potentially enhance psychotherapy practices. The innovative technology, available for free, was trained to detect six basic emotions: happiness, surprise, anger, disgust, sadness and fear, utilizing an artificial neural network trained on over 30,000 facial photos. However, researchers were amazed by its ability to consistently and reliably match facial expressions to their corresponding emotional states.
AI as reliable as therapists to assess facial expressions
The AI model was put to the test at the University of Basel's Center for Scientific Computing in Switzerland. It analyzed over 950 hours of video recordings from therapy sessions involving 23 patients with borderline personality disorder. Researchers then compared model-generated analyses with those provided by three trained therapists and found a "remarkable level of agreement." They mentioned the AI's ability to assess patients' facial expressions was as reliable as a trained therapist.
It detected emotions lasting less than a millisecond
As per the research team, the model was able to identify the most fleeting emotions, lasting less than a millisecond, such as a brief smile or an expression of disgust. This also suggests that the AI exhibits greater sensitivity than therapists in detecting momentary emotional displays, which might be overlooked by therapists or only be perceived subconsciously, the team said.
AI aids therapy but human connections remain essential for success
While the researchers believe that AI could become a valuable tool in therapy and research, they emphasize that the therapeutic process is still fundamentally about human connections. The use of AI in this context could help support therapists in their work, but it is essential to remember that the human element remains at the core of successful therapy sessions.