How mistakes help us effectively study ourselves and the world

A book by cognitive neuroscientist Stephen M. Fleming's "Metathinking" is devoted to self-knowledge. It describes the mechanisms that help people discover and reflect on their thoughts. Fleming explains why we sometimes find it easier to believe a stranger than our own childhood memories. describes what happens to the brain when we try but can't remember a word, and touches on others interesting topics.

With the permission of Individuum, we publish an abridged excerpt from the chapter "Self-Control Algorithms" about why a person could not develop if he did not make mistakes.

---

One of the first studies on how we spot our own mistakes was done by psychologist Patrick Rabbitt in the 1960s. He came up with a complex monotonous a task in which subjects had to press buttons in response to a sequence of numbers. However, the task itself did not matter much. The trick was that the subjects were asked to press a separate button if they noticed that they had made a mistake. Rabbitt accurately measured the time it took to press this button, and found that people are able to correct their own mistakes extremely quickly. In fact, the subjects realized that they had made a mistake, on average, 40 milliseconds faster than they reacted to external stimuli. This elegant and simple analysis proved that the brain is able to track and detect its own errors through efficient internal calculations independent of signals from the outside world.

A quick process of finding bugs can lead to an equally quick fix.

When making a simple decision about whether this or that stimulus belongs to category A or B, after tens of milliseconds after pressing the wrong button, the muscles that control the correct answer begin to contract, trying to correct the mistake. And if corrective processes happen quickly enough, they can even prevent it. For example, by the time the muscles contract to press the button and send thoughtless message, We we have time acquire additional evidence of the failure of this undertaking and at the last moment refrain from a fatal mouse click.

Decades after Rabbitt's experiment, scientists began to discover brain processes that promote internal fault finding. While working on his doctoral dissertation, published in 1992, psychologist William Gering wrote electroencephalograms (EEG) of participants in one experiment who performed complex tasks. An EEG uses a network of small electrodes that non-invasively detects changes in the electric field caused by the combined activity of thousands of neurons within the brain. Hering discovered that less than 100 milliseconds after making a mistake, a special wave appears in the brain. The speed of this reaction helps to explain what Rabbitt discovered, namely the ability of people to rapidly realizethat they have made a mistake, even before they are told about it.

This brain activity has come to be known as error-related negativity, or ERN, which modern psychologists affectionately refer to as the "Damn!" response.

Today we know that this reaction occurs as a result of errors in the performance of a wide variety of tasks (from pressing buttons until reading aloud) and is generated by the brain area located in the center of the frontal lobe, the dorsal zone of the anterior cingulate cortex. This telling neural evidence of self-monitoring is found early on. development person. In one experiment, 12-month-old babies were shown different images on a computer screen while recording their eye movements. Sometimes they were shown a human face, and if the babies looked directly at it, they were rewarded with music and flashing colored lights. If the child did not look at the image of the face, then in the context of the experiment this was considered a mistake - he did not perform an action for which he would have received a reward. In such cases, EEG recordings clearly reflected NSO, even if the reaction was somewhat belated compared to adults.

NSO can be considered as a special case of the "predictive error" signal. The name “predictive errors” is self-explanatory: they are errors in our predictions of the future, which are also a key component of algorithms that help to effectively study the world. To understand how predictive errors help us with this, imagine that a new coffee shop opens near your office. You still don't know how good it is, but its owners made sure to buy a first-class coffee machine and create a great atmosphere. You have high expectations - you assume that coffee will be good, although they have not drunk it yet. Finally, you try it for the first time and find out that it is not just good - you have not drunk such a wonderful espresso for a long time. Because the coffee exceeded your expectations, you update your estimate and the coffee shop becomes your new favorite stop on your way to work.

Now imagine that several weeks have passed. The baristas have relaxed and the coffee isn't as good as it used to be. It may still be good, but given your increased expectations, you perceive what is happening as a negative error in your prediction and may become even more disappointed.

The ability to make and update predictions depends on a well-known brain chemical called dopamine.

Dopamine is not only famous, but often misunderstood - in the popular media it is called the "pleasure hormone". It is true that dopamine levels rises from what we like: money, food, sex and so on. However, the notion that dopamine merely signals the rewarding nature of the experience is misleading. In the 1990s, neuroscientist Wolfram Schultz conducted an experiment that has become a classic. He registered in monkeys signals sent by midbrain cells that produce dopamine and deliver it to other areas of the brain. Schultz taught the monkeys that after turning on the light in the room they were given some juice. At first, dopamine cells responded to the juice, which was consistent with the pleasure theory. But over time, the animals began to understand that turning on the light always precedes the juice - they learned to expect pleasure - and the dopamine response disappeared.

An elegant explanation of the dopamine response pattern in these experiments is that it helped the brain track errors in monkey predictions. At first, juice was a surprise to the monkeys, just as good coffee in a new place was a surprise to you. But over time, the monkeys began to expect juice every time the lights turned on, just like we expect good coffee every time we walk into a coffee shop. Almost simultaneously with Schulz's experiments, computational neuroscientists Peter Diane and Reed Montague worked on the development of one classic psychological theory of learning by trial and errors.

According to this famous theory, the Rescorla-Wagner model, learning occurs only if events are unexpected.

This is understandable even intuitively: if today's coffee is the same as yesterday, we do not need to change the rating that we gave the coffee shop. You don't need to learn anything. Diane and Montagu demonstrated that variants of this algorithm are in excellent agreement with the response of dopamine neurons. Shortly after the publication of the work of Schulz and Diane and Montague, a series of studies by my former supervisor, Ray Dolan, revealed that the reaction neurons in areas of the human brain that receive a dopamine signal is fully consistent with what happens when a predictive signal is received. error. These studies have shown that calculating predictive errors and using them to update our perception of the world lie at the core of how the brain works.

Armed with an understanding of predictive errors, we begin to see how important such calculations are for self-monitoring. Sometimes we directly receive positive or negative feedback about our activities − for example, when we complete a school assignment or find out that we broke a personal record in a half marathon distances. But in many areas of daily life, feedback may be less noticeable or non-existent. Therefore, it is reasonable to consider that the NSO reflects an internal signal about remuneration Or, more accurately, its absence. It expresses the difference between what we expected (we succeeded) and what actually happened (an error occurred).

Imagine yourself sitting down at the piano to play a simple melody. Each note has its own sound, but it would be strange to say that one of them is “better” or “worse” than the other. Played alone, A is no better than a G-sharp. But in the context of the melody that opens the Piano Concerto in A Minor by Edvard Grieg, the mistakenly played G-sharp instead of A will make the listeners shudder. Even if there is no external feedback, the wrong note is an error against the background of the expected execution. Tracking such errors, the brain can appreciatewhether he performs well or poorly, even in the absence of explicit feedback.

By definition, mistakes usually don't happen when we expect them to, otherwise we might be able to prevent them.

This feature of human error is used for comic effect in one of my favorite sketches from "Fast Show". His character, Old Man Unlucky Alf, turns to the camera and says in a thick Northern English accent, “See that over there? They're digging a damn big hole at the end of the road. With my luck, I'm sure I'll fall into it." We watch tensely as he slowly wanders along the road, until suddenly a strong gust of wind comes up and blows him into a hole. Preparedness, foresight, and yet the inevitability of disaster is what makes this sketch funny. We are surprised at mistakes precisely because we do not expect them, and, like Homer Simpson, exclaim "D'ow!", already being confronted with a fact.

Thus, an effective way to do self-monitoring is to make predictions about how we are doing well and see if we are doing well.

The book "Metathinking" will help to understand how the human mind works from the point of view of neuroscience. It is useful for those who want to learn to better understand themselves and others.

Buy a book