AI system accurately detects facial expressions of emotions in therapy

Cutting-edge research has revealed that an AI-driven system has the capability to accurately identify facial expressions that convey emotions, including fleeting ones. This breakthrough discovery holds great potential in bolstering psychotherapy.

The study employed a readily accessible artificial neural network, trained on over 30,000 facial photos, to detect six fundamental emotions: happiness, surprise, anger, disgust, sadness, and fear. Artificial neural networks, a subset of AI, are built upon the principles of connections found in biological neural networks in animal brains.

Researchers then utilised the AI model to process and analyze over 950 hours of video recordings from therapy sessions involving 23 patients diagnosed with borderline personality disorder. The sessions were conducted at the prestigious Center for Scientific Computing, University of Basel, Switzerland.

Comparing the model-generated analyses with those of three trained therapists, the international team uncovered a “remarkable level of agreement.” The AI system proved to be as adept as a skilled therapist in accurately assessing patients’ facial expressions. Additionally, the model excelled in detecting fleeting emotions that lasted less than a millisecond, such as a quick smile or a gesture of disgust.

The AI system’s heightened sensitivity toward these momentary displays of emotions surpassed that of therapists, who may overlook or subconsciously perceive them. Such nuanced insights could potentially have a significant impact on therapeutic outcomes. The team’s findings have been published in the esteemed journal Psychopathology.

Martin Steppan, psychologist at the Faculty of Psychology at the University of Basel and corresponding author of the study, explained, “We wanted to determine whether AI systems can consistently ascertain the emotional states of patients in video recordings.”

Furthermore, the researchers discovered that the model’s analysis unveiled an intriguing trend. Patients who displayed emotional involvement, as evidenced by a smile at the beginning of a therapy session, demonstrated greater adherence to psychotherapy compared to those who appeared emotionally indifferent toward their therapist.

Consequently, the smile emerged as a “promising predictor” of therapy session success for individuals diagnosed with borderline personality disorder.

Steppan expressed surprise, saying, “We were truly amazed to find that relatively simple AI systems can accurately attribute facial expressions to their corresponding emotional states.”

The team emphasised that AI has the potential to become an invaluable tool in therapy and research, supporting the supervision of psychotherapists. However, they underscored that human relationships remain paramount in therapeutic work, stressing that the domain of psychotherapy remains inherently human.