News

    News

    Modern artificial intelligence can respond so sympathetically that it sometimes creates the impression of genuine understanding. It chooses careful words, acknowledges our feelings, and keeps the conversation going. But can it do more than talk about emotions — can it actually recognize them from our voice and movements? On September 3, 2026, Scientific Reports published the results of a study by David Piterman and colleagues on the ability of generative AI to recognize nonverbal expressions of emotion. The researchers tested three multimodal models from the Gemini family and compared their accuracy with that of people performing similar tasks. The models were asked to identify emotions from short video clips of body movements with faces hidden, as well as from meaningless phrases spoken with different intonations. This allowed the researchers to test separately how well AI understands body language and the emotional tone of voice without relying on word meaning or facial expressions. The results showed that in some tasks modern systems are already approaching human performance, but their emotion recognition remains highly uneven. They identify positive states better than negative ones — and errors occur more often precisely where it is especially important to notice fear, pain, or sadness. Emotions Without Words or Facial Expressions The researchers tested three multimodal models from the Gemini family — systems capable of analyzing not only text, but also audio and video. They were asked to identify emotions from two types of material. In one case, these were short videos of professional actors filmed without showing their faces: the emotional state was conveyed solely through posture and body movement. In the other, meaningless phrases were spoken with different intonations. Since the meaning of the words could not provide a clue, only timbre, pitch, rhythm, and other characteristics of speech remained. The models selected one option from six possible answers. Their results were compared with responses from participants in earlier studies that had used the same materials. When recognizing emotions from body movements, people outperformed all three systems. The most advanced model tested came fairly close to human performance, but the difference remained. The pattern was different for voice: two newer models achieved roughly the same accuracy as people. However, this does not mean they understood intonation without error. Both humans and AI gave the correct answer in fewer than half of the cases. Emotions expressed through voice alone proved difficult for everyone. Joy Is Easier to Detect Than Distress The most interesting finding concerned not the total number of correct answers, but the pattern of errors. For people, recognition accuracy did not differ substantially between positive and negative emotions. AI models, by contrast, performed noticeably better with positive states. This was especially clear when analyzing body movements. Excitement, interest, playfulness, and friendliness were identified with high confidence. Negative experiences caused more confusion: fear was mixed up with shame, sadness with disgust, and anger with dislike. An emotion described by the researchers as pain or emotional hurt was not correctly recognized from body movements by any of the models. The authors call this pattern a “positivity bias.” This does not mean that AI feels optimistic or consciously tries to see the world in a better light. A model does not experience what it sees the way we do. It searches the input for patterns that make one emotion label more likely than another. The study does not establish where this bias comes from. Positive expressions may have been better represented in the data used to train the systems. Fine-tuning that rewards friendly and acceptable responses may also have played a role. It is also possible that some negative emotions are simply harder to distinguish from one another when only isolated nonverbal cues are available. Convincing Empathy Does Not Yet Mean Understanding It is easy to take an appropriate response as evidence that the other party has correctly understood our state. In ordinary human communication, that is often reasonable: words, voice, pauses, facial expression, and the history of the relationship all complement one another. But in a conversation with AI, there is an important difference between the impression created by the response and the way that response was produced. A system may say, “It sounds like things are really difficult for you right now,” because it has learned the language patterns of support very well. That does not yet prove that it accurately recognized anxiety in the voice, dejection in posture, or another nonverbal sign of distress. In a friendly conversation, such an error may go almost unnoticed. But the significance changes when AI is proposed for psychological support, remote monitoring of a person’s condition, or assistance to a professional. If a system is particularly good at noticing joy but worse at distinguishing pain, fear, and sadness, its outward warmth may create an exaggerated impression of emotional sensitivity. This does not mean that emotion-recognition technologies are useless. On the contrary, some of the results show how quickly they are developing. But the ability to produce sympathetic words and the ability to reliably notice another person’s distress are not the same thing. What the Study Does Not Yet Prove The experiment was conducted under deliberately simplified conditions. The actors intentionally portrayed emotions, and did so fairly clearly. The videos showed body movements without faces, while the audio recordings contained voice without meaningful words. In real conversation, all of these signals occur together and depend on the situation, culture, an individual’s way of expressing emotion, and how well the people know one another. In addition, only three versions of Gemini were tested. The result cannot automatically be generalized to all AI systems — especially because models are updated rapidly. A format with six predefined answers is also much easier than freely interpreting a complex emotional state. The study therefore does not prove that AI will necessarily miss signs of severe distress in a real conversation or cause harm as a result. It did not study patients, therapy, or the consequences of using such systems. Under controlled conditions, the authors identified a specific feature of emotion recognition; its significance for real psychological support still needs to be tested. Nevertheless, the study highlights an important boundary. AI can reproduce the language of empathy very convincingly, but convincing language is not the same as emotional understanding. When we especially need support, it is better not to treat a warm response from a machine as proof that it has truly noticed everything that is happening to us. Source: David Piterman et al. “Generative AI emotion recognition from bodily gestures and vocal tone reveals modality-specific performance and positivity bias”, Scientific Reports, September 3, 2026

    News

    Artificial intelligence is becoming an increasingly visible part of our lives, and with it comes talk of a new wave of anxiety. We worry about losing our jobs, about the security of our personal data, about not being able to keep up with new technologies, and about how far the development of AI may ultimately go.

    All of these concerns are often grouped under a single expression — “AI anxiety.” However, a new study suggests that this label hides different kinds of worries that may arise independently of one another.

    Researchers from the Universities of Pisa and Bergen developed a special scale called AISAS, designed to measure stress and anxiety related to AI. The work consisted of two preregistered studies conducted among adults in the United Kingdom. The samples were constructed to reflect the country’s population in terms of age, sex, and ethnicity.

    The first study involved 301 participants. Their responses helped identify four main types of concern:

    • worries related to work and the possible loss of professional relevance;
    • stress caused by the need to learn how to use AI and adapt to new technologies;
    • anxiety about privacy and the use of personal data;
    • existential questions, including the possible emergence of consciousness in artificial intelligence.

    A second study involving 324 participants confirmed that this structure also appeared in another sample.

    These findings reveal an important distinction. We may use AI tools comfortably and have no difficulty learning how they work, while still being seriously concerned about the data they collect. Or the opposite may be true: we may not be especially worried about privacy, but feel that our profession is under threat. That is why the general question “Are you afraid of artificial intelligence?” oversimplifies the range of possible reactions.

    The most common concerns among participants were related to privacy. By contrast, stress about having to learn and adapt to AI was relatively uncommon. This does not entirely match the image of a society universally frightened by the rapid development of new technologies.

    But it would be misleading to treat the study as evidence of a new “anxiety epidemic.” The authors’ main goal was to create and preliminarily validate a measurement tool, not to determine how many people are experiencing psychological problems because of AI. The study also does not show that artificial intelligence itself caused the concerns that were observed.

    The new scale also has limitations. The component related to stress from adapting to AI consists of only two items and cannot yet be regarded as a fully developed subscale. In addition, both validation studies were conducted only in the United Kingdom. It is not known whether the same structure of concerns would appear in countries with different attitudes toward technology, labor markets, and data protection rules.

    The scale is not intended for diagnosis. But it may help future research speak more precisely about our relationship with artificial intelligence. Instead of a broad and alarming concept of “AI anxiety,” it becomes possible to ask what exactly we are worried about: losing our jobs, losing control over personal information, having to constantly retrain, or a future in which the boundary between human and artificial becomes increasingly unclear.

    Source: Enrico Cipriani, Hanna Joy Justesen, Danilo Menicucci & Simone Grassini. “The AI stress and anxiety scale (AISAS): development, initial validation and insight on the diffusion of AI-related stress and anxiety”, Scientific Reports, 2026

    News

    It is commonly assumed that information bubbles are created mainly by social media algorithms. They show us content similar to what we have already been interested in and gradually surround us with familiar opinions. But a new study reminds us that we also help create these bubbles ourselves when we decide whom to stay connected with and whom it may be better to unfollow.

    Researchers from the University of Michigan and Duke University created an experimental social network in which participants could learn about one another’s views on various social issues. The study involved 373 people. Altogether, they made 7,328 decisions about whether to maintain a connection with another user or break it.

    The result was quite expected: participants were significantly more likely to cut contact with people whose beliefs differed from their own. The greater the disagreement, the higher the probability that the connection would be broken.

    This is how an information bubble can gradually form. We do not necessarily set out to find a space where everyone thinks alike. Sometimes we simply remove, one by one, those whose posts provoke irritation, anxiety, or the urge to argue. As a result, our social environment becomes increasingly homogeneous, while other viewpoints begin to seem rare, strange, or completely unacceptable.

    However, the study also identified a factor that helped preserve contact even in the presence of disagreement. If participants saw that they shared mutual connections with another user, they were less likely to cut ties with that person.

    Similarity of beliefs and the presence of mutual connections worked independently of each other. Mutual acquaintances did not make participants agree, nor did they eliminate dislike of another person’s position. But they made the connection with that person more stable.

    Perhaps the presence of mutual acquaintances makes it harder to perceive someone we disagree with as a complete outsider. They remain part of a familiar social environment, connected to people we already trust. The disagreement does not disappear, but it becomes only one aspect of the relationship rather than the sole basis for deciding whether to maintain contact or end it.

    Because the researchers themselves manipulated the information participants saw about mutual connections, it is possible to say that, under the experimental conditions, this factor directly influenced the decisions they made. However, the findings should be applied to real social networks with caution. The experiment took place in an artificially created environment, and participants made isolated decisions about unfamiliar users. Real relationships develop over years and depend on many circumstances that cannot be reproduced in a laboratory.

    The study does not prove that a “mutual friend” label can rid society of polarization. But it does show that information bubbles arise not only because of how algorithms are designed. They are also shaped by our own decisions about whom we are willing to keep close when we encounter disagreement.

    And perhaps a shared social connection can sometimes become that thin thread that helps us avoid turning differences in opinion into a complete end to communication.

    Source: Clint McKenna, Ashley Harrell & Ashley Anderson. “Belief consonance and cues of structural embeddedness jointly shape curation in online social networks”, Scientific Reports, 2026

    News

    When we feel bad, we often expect our energy, interest, and desire to do anything to return first — and only then do we think we will be able to start going out again, socializing, working, or returning to our usual activities. However, research on behavioral activation suggests that this sequence can also work in reverse: sometimes a small action comes first, and only then does our state gradually begin to change.

    In August 2026, the journal Clinical Psychology Review published the largest review to date of research on behavioral activation for depression. The authors combined the results of 105 randomized studies involving almost 14,000 patients.

    Behavioral activation is a form of psychotherapy based on gradually bringing back into our lives activities that may be important, useful, or capable of giving us at least a small sense of satisfaction. With depression, these activities usually become fewer and fewer. We meet loved ones less often, abandon familiar routines, move less, and become increasingly distant from what once gave our lives meaning.

    This creates a vicious cycle: a difficult emotional state leads us to withdraw from activity, while reduced activity leaves fewer and fewer opportunities to experience interest, pleasure, closeness, or a sense of accomplishment.

    Behavioral activation helps us gently break this cycle. It is not about being told to “pull yourself together” or forcing ourselves to return immediately to our former life. Together with a specialist, we identify what truly matters to us, observe the connection between our actions and our mood, and begin with small, manageable steps. For one person, that step may be a short walk; for another, cooking a meal, calling someone close, or returning to an activity they gave up long ago.

    The review found that in adults, behavioral activation noticeably reduces symptoms of depression compared with control conditions. On average, its outcomes were no worse than those of other forms of psychotherapy, and the positive effect was still present after 12 months. Self-guided programs without continuous therapist involvement also helped, although their effects were weaker.

    At the same time, the results of individual studies varied considerably, and only 40 of the 105 studies were rated as having a low risk of bias. So it cannot be claimed that the method works equally well for everyone. There is also still insufficient evidence about its effectiveness for children and adolescents and about direct comparisons with medication.

    The main conclusion of the study is not that people with depression should simply force themselves to be more active. Depression is not laziness or a lack of things to do. But waiting for the moment when energy and motivation return first can sometimes become part of the vicious cycle itself. The first careful step is not always taken because we already feel better. Sometimes it is precisely where improvement begins.

    Source: Pim Cuijpers et al. “Behavioral activation for depression: A comprehensive systematic review and meta-analysis”, Clinical Psychology Review, 2026

    News & Articles Newsletter

    Поиск