Our brain works thanks to its prediction capabilities

Neuroscientists have developed artificial neural networks to study this phenomenon: it is the prediction capabilities that make our brain work

The human brain is one of the most fascinating subjects of study for scientists, fake news aside. And some elements of its functioning still remain unknown, despite recent and very "peculiar" studies. For example, it is still not clear how the brain creates perceptions from sensations. In fact, there are many researches that show that the brain is not able to put together information derived from the senses as if they were pieces of a puzzle. So some scientists are advancing the theory that our brains are capable of "prediction".

The Predictive Capabilities of Our Brains

According to this new theory, the brain uses the information it already has about its surroundings to generate hypotheses about the sensory stimuli it receives. Thus, it is the hypotheses, and not the stimuli, that generate our perceptions. And this works especially if the stimuli it receives are difficult to interpret.

"This theory is still in its infancy, and we are open to alternative explanations," says Floris de Lange, a neuroscientist at Radbound University in the Netherlands. The scholars are exploiting a computational model to test the correctness of their hypotheses.

The Neuroscientists' Studies

The scholars have built an artificial neural network that mirrors the behavior of neurons that exist in nature. This network has learned to make predictions about incoming information, and has shown some surprising capabilities. Some experiments conducted on this artificial network suggest that our brains have had to evolve as "prediction machines" to meet the power limits they can receive.

In fact, thanks to these computational studies, neuroscientists are increasingly convinced that the brain works on experience and makes assumptions about the causes of sensory stimuli. While the exact details of this process remain nebulous, the contours seem to be getting clearer.

This is an idea that actually has been circulating for a thousand years: the Arab astronomer Hasan Ibn Al-Haytham talked about it in 1011, then again the German physicist Hermann von Helmholtz in 1860 and then the cognitive psychologists at the beginning of the twentieth century. The first truly scientific study was in 1980, by Richard Langton Gregory, who said that perceptual illusions are essentially faulty assumptions by the brain about the causes of sensory impressions.

Many questions remain, however, about how prediction processing actually works. One theory in view says that there is a hierarchy of levels of information processing in our brains. The highest level represents abstract knowledge: for example, the perception of a snake in the shadows of a grove. This level makes predictions about this perception, and sends them to lower levels that compare it to the information they are processing. If something doesn't add up, it sends the information back up with an error signal so that the higher level can update its predictions. The last level is the one that actually receives the sensory stimuli.

In particular, there are pyramidal neurons in the neocortex of the mouse brain that are thought to be anatomically suited for predictive processing. Indeed, they can receive both sensory signals from the lowest level and predictive signals from the highest level.