Artificial intelligence is increasingly mastering the task of recognizing human emotions through facial expressions, vocal intonation, and physiological signals. However, according to Leiden University researcher Alexandra Prêzan, the very moment when such systems become virtually infallible could turn into a serious risk for society. This is reported by RBC-Ukraine, citing the scholar's fresh dissertation research, published on the official website of Leiden University. Prêzan did not limit herself to theoretical reasoning: to assess the consequences of deploying perfect emotion recognition technologies (ERT — Emotion Recognition Technology), she modeled a number of future scenarios in which an algorithmic “verdict” on a person's inner state begins to determine their professional and social fate.
The Surgeon Scenario: When a Personal Divorce Becomes a “Professional Risk”
One of the most striking examples proposed by Prêzan concerns a surgeon who arrives at the clinic before an extremely complex operation. The ERT system detects his emotional instability, and the hospital administration removes the doctor from duty. Yet the cause of the algorithm's alarm had nothing to do with his professional competence: the surgeon was going through a painful divorce and worrying about his child. Despite the fact that the doctor had consciously kept his personal problems outside the workplace, the technology classified his state as a threat to the performance of his duties. As a result, the person is forced to disclose and justify to their employer the secrets of their own life, because the system has already rendered its verdict, and it now needs to be “refuted” either documentally or emotionally.
The Destruction of the Boundary Between Private and Professional Life
In life, every person plays many social roles: surgeon, father or mother, partner, friend, neighbor. As a rule, we ourselves decide which boundaries and which layers of our inner world to open up in a given situation. The loss of privacy under infallible ERT means that hiding one's inner experiences becomes impossible, and the boundary between private and professional life is finally destroyed. A person loses the right to an “emotional pause,” the right not to explain why they are sad or irritated today — a right that today is an unspoken but fundamental foundation of social interaction.
The Changing of Communication Rituals and the Shifting of Trust Benchmarks
Prêzan points to yet another alarming shift: a public speech or an apology by an official after a tragedy may be judged by society not by the content of the words and actions, but by the “genuineness” of the expressed grief according to the AI. Meanwhile, doctors, police officers, or politicians are often forced to suppress their emotions for self-protection within the framework of professional ethics, which does not make their professional duty any less sincere. Instead of trusting people's words and actions, society may begin to orient itself exclusively on what the technology “reports” about their state. According to the scholar, this shift in benchmarks undermines the very foundation of interpersonal trust, which was built on a person's ability to choose for themselves what and when to show others.
Current Regulation and the Long-Term Outlook
Today, the European AI Act (EU AI Act) already restricts the use of emotion recognition technologies, but primarily because of their low accuracy at this stage. Prêzan's research is aimed at a different, long-term perspective: the moment when algorithms become infallible. In this situation, society will need stronger moral and legal arguments to restrict the application of ERT, because the argument that “the technology makes mistakes” will stop working. The scholar emphasizes that protecting emotional privacy is not just the protection of information in the narrow sense. It is the preservation of a social space in which a person decides for themselves how to position themselves and build relationships with others.
Not Only a Threat: The Positive Potential of ERT
It is important to note that emotion recognition technologies are not an exclusively negative phenomenon. Prêzan acknowledges that ERT can help in the early diagnosis of neurological disorders, as well as support people who have difficulty recognizing their own emotions — for example, in some forms of autism or alexithymia. Thus, the discussion raised by the dissertation does not reduce to a “ban everything” slogan, but concerns finding a balance: where is the line between a useful diagnostic tool and total emotional surveillance that deprives a person of the right to inner privacy.
