Imagine the brain as a vast city at night. Thousands of lights flicker on and off simultaneously in every district. At first glance, it seems like chaotic twinkling. But if you observe for a long time, you can notice patterns: when a person is happy, the 'light pattern' is one, when they are anxious, it's entirely different.
This is precisely what a new study published in Nature Computational Science is dedicated to.
Scientists worked with 18 epilepsy patients who, for medical reasons, already had electrodes implanted in their brains. While participants viewed emotional images and video clips, they continuously rated their feelings: how pleasant or unpleasant the emotion was, and how strong it was. Simultaneously, a computer recorded the brain's electrical activity.
Then, researchers trained an artificial intelligence to find the connection between these signals and what the person was feeling. As a result, the system learned to quite accurately determine the emotional state in almost real-time.
The most interesting part turned out to be something else. Previously, most similar studies only analyzed the activity of the brain's gray matter – where neuron cell bodies are located. But the authors decided to also include signals from white matter – the 'cables' that connect different brain regions.
It turned out that it is along these nerve highways that a huge amount of information about emotions is transmitted. When the model considered both gray and white matter simultaneously, its accuracy significantly increased.
Just a few years ago, it was believed that emotions were primarily generated in specific structures, such as the amygdala. The new work shows a more complex picture: a feeling does not reside in a single brain area. It arises from the coordinated work of an entire network, uniting the cortex, thalamus, limbic system, and the nerve pathways connecting them.
But an interesting philosophical question arises.
If a computer can determine that a person is anxious or happy, does that mean it has 'read' their feelings?
Not quite.
The algorithm recognizes patterns of electrical activity that typically accompany certain experiences. It's like a weather forecast. Based on clouds, humidity, and pressure, you can predict rain almost unerringly, but the instruments themselves don't feel the smell of wet earth or the coolness of the first drops.
It's the same here. Artificial intelligence has learned to predict emotions, but this doesn't mean it knows what joy, fear, or love is from the inside.
This is a very important result for modern neuroscience. It aligns well with predictive processing theories, according to which the brain constantly builds predictions about what is happening inside the body and in the external world. Emotions, in this context, do not arise as separate 'happiness buttons' or 'fear centers,' but as a result of continuously comparing the brain's expectations with what is actually happening.
The study also has limitations. All participants were epilepsy patients, so the results cannot be automatically extrapolated to the entire population. Furthermore, the participants themselves assessed their emotions, and the real-time mode could only be tested on four individuals so far.
Nevertheless, the work demonstrates how rapidly the field of neurotechnology is advancing. In the future, similar systems could help detect the onset of depression, anxiety attacks, or emotional breakdowns even before a person becomes aware of what is happening. This opens up new possibilities for treating mental disorders.
But simultaneously, a new ethical question arises: if technology one day learns to recognize our emotions very accurately through brain activity, where should the line be drawn between medical assistance and a person's right to the privacy of their inner world? This very question may become one of the main challenges for neuroscience in the coming years.
