Comprehensive coverage

Where is the border?

What can people really learn? What is the limit of learning? What laws govern the learning process? What happens in our brain when we learn?

Right: Dr. Elad Schneidman and Jordan Cohen. Learning patterns
Right: Dr. Elad Schneidman and Jordan Cohen. learning patterns

What can people really learn? What is the limit of learning? What laws govern the learning process? What happens in our brain when we learn? A corridor debate on these questions, between two research students, led the members of Dr. Elad Schneidman's group, from the Department of Neurobiology at the Weizmann Institute of Science, to design a computerized learning game that allowed them to "go over the limit". Their findings were published in the journal PNAS.

The game they developed consists of a combination of colored squares. The subjects are shown a combination of six colored squares, in seven possible colors, and they are asked to observe and learn the law that governs the structure of the presented situation. For example: red, blue yellow, red blue yellow. Or: red, red, blue, yellow, blue blue. Since the possible combinations for arranging six squares in seven different colors are very many, the scientists decided to be content with situations representing only six rules. The first interesting result emerged, when it became clear that everyone succeeded in something, different people managed to learn a different number of laws - but none of the subjects was able to identify and learn all six laws.

The next question was, how do we use what we learn; For example, how do we economize our steps when we put together an investment portfolio according to the advice of five advisors, each of whom offers us five stocks. How much do we punish the advisors for poor performance of the stocks they suggested, or reward them for good performance? "Punishment" in this case is the reduction of the advisor's relative share in our investment portfolio, and reward is the increase of his share in the portfolio. Dr. Shneidman says that here other factors come into play, such as the prejudice we came with, and the self-confidence of the consultant, the owner of the case. There are those for whom every performance measurement leads them to redistribute the portfolio among the different consultants; And there are those who rarely make changes, and allow those who fail (up to a certain limit, of course), to improve. The possible combinations between these features create different learning patterns, which characterize different learner types.

Cereal in Brest. Untitled, 1976. This work was part of an installation shown at the Julie M. Gallery. In 2010, from an installation that was included in the exhibition "Bristols", which was presented in 2012 at the Bat-Yam Museum of Contemporary Art, and brought different perspectives on the education system in Israel. Curator: Tali Tamir

Dr. Schneidman and the members of his group developed a mathematical model that allows analyzing the way of learning of different subjects, and knowing what weight they give to prejudice, experience and learning. Such a model, which simulated the learning process of different subjects, was able to correctly predict 88% of the learning processes performed by the subjects. But does the identification of the learning processes make it possible to correctly predict the decisions that the subjects will make following the learning? Here the success rates of the model decreased slightly, but still remained quite high: 75%.

From here the scientists moved on to the next question: How can the learning process be improved? For example, how can the learning process of the scientists themselves be improved, so that the model they assemble reaches more than 75% success in predicting decisions that the subjects will make following learning? What examples can be given to the subjects, to help them learn more quickly the law according to which the colored squares are arranged in the game? It turns out that intervention in only one pattern significantly improves the learning performance of most subjects.

In a similar experiment, carried out in animals, the scientists were able to identify individual neurons that are involved in and define the learning process. This focus hints at future possibilities for intervention that will greatly improve learning ability.

In the last experiment in the series (for now), a difference was discovered between the learning methods of physicists and biologists. So, for example, when a physicist receives a complex combination of colored squares, he immediately calculates the number of possible solutions, recognizes the complexity - and starts looking for complex solutions. On the other hand, biologists who are faced with a complicated system prefer rather to start testing simple solutions, and gradually move up the complexity scale. What does this say about the way different scientists operate? This is a question that still awaits resolution.

2 תגובות

  1. It is not clear if the writer is confused or the researchers are confused or I am confused.

    "Identifying a law" which is actually "finding a pattern" (more precisely, pasting or inventing a pattern) is not primarily a learning process but a process that is primarily the result of the learning a person has learned throughout his life.

  2. Where is zero? If we are trying to find a way to improve the learning process and improve the accuracy of predictions, how can a person reach a dichotomous state like a binary code where there is yes and there is no? that we are talking about different methods that physicists and biologists use on the way to the answer, if they arrive at the same answer how do we know it is the right one? What I'm actually asking, and sorry for the philosophizing, is it possible to create a scale for something that we can't determine is objective, and if not, it doesn't actually make the entire engagement with the matter superfluous

Leave a Reply

Email will not be published. Required fields are marked *

This site uses Akismat to prevent spam messages. Click here to learn how your response data is processed.