Richard Hamming: Chapter 22. Computer-Aided Learning (CAI)

Original author: Richard Hamming
  • Transfer
“What you learn from others you can use to follow;

- by Richard Hamming
imageIt remains to publish 1 chapter ...

Сomputer-aided instruction (CAI) - machine learning. The use of computers in the educational process. The computer in the dialogue mode offers the student material in the form of texts and graphic images, gives assignments and checks their implementation, allows to carry out educational experiments.
Translation Islam Rasulov

Since computers were installed in many universities, sooner or later there would be a question about using computers for training. Before we move on to the modern statements, it would be useful to study existing theories on this issue.

Since ancient Greek times there is a story about mathematics, telling the ruler about the existence of royal roads for walks, royal envoys to send mail to the ruler, but there was no road to geometry. You also know that money and training will not help much if you want to run a mile in 4 minutes. There is no easy way to do this. And this applies to absolutely all people.

Stories are known to a huge number of people who were looking for easy ways to learn something. Aldous Huxley in his book “On the Brave New World” discusses the idea of ​​learning while you sleep, when you put a microphone under your pillow and play something through it. He also reveals the serious limitations of such a process.

During my work at the Bell Telephone Laboratory, Dianetics emerged in the world, which promised that it could “clear” your mind of all its mistakes, after which you would be able to reason perfectly. There are still institutions of Dianetics, but their main problem - the people created by them, it seems, did not succeed in any sphere of human activity, let alone in all spheres. Another organization promises to reveal the secrets of ancient people (who was, perhaps, much more intelligent than we are now). There are an infinite number of announcements about speed reading, speed learning techniques, etc. around us, each of which promises, in one way or another, to significantly improve our mind, without putting in the heavy efforts that most of us have to put in order to achieve success. Testing such offers has shown that none of them has yet led to the emergence of a significant number of outstanding people (which we actually know at the moment). As Fermi said about super-earth intelligence and UFO aliens: “Where are they and why have we not met them?”

Consequently, the whole story with its numerous statements about easy learning speaks eloquently against the current surge of promises, but this, of course, does not mean that some new trick will not work. You should be wary of each such proposal, but at the same time, it is possible that new things, previously unknown, and new tools, such as cheap computers, which are currently widely available, are not out of the question. And these tools may differ. Often I read or hear that I have to believe that the computer will drastically change our lives in spite of all the past promises that obviously failed. Beware of the temptation to wishful thinking - you would like it to be true, so you assume that this is true!

There is another important factor that needs to be explained - the Hawthorne effect. A long time ago, at the Western Electric plant in Hawthorne, several psychologists tried to increase the productivity of workers through various environmental changes. They painted the walls in an attractive color and the productivity of the workers increased. They made the lighting softer and the productivity of the workers increased. Each change contributed to improved performance. One of the men, a little suspicious, brought the change back to its original state, and the performance increased again! Why? It turned out that when you take care of a person, he responds in the same way. All workers believed that changes were made for good and responded accordingly.

With regard to the learning process, when you tell students that you are using a new method in teaching, they increase their effectiveness in response, which, by the way, the professor does. The new method may or may not be better, in fact, it may even be worse than the existing teaching methods, but the Hawthorne effect, which is not so insignificant in the field of education, probably indicates a new, important, better teaching method. And it doesn’t matter what this method is, if students realize that this method is for their own good, it will certainly lead to improvements.

Thus, the Hawthorne effect distorts most educational experiments. Let us recall my earlier reasoning, discussed in Chapter 20, about the need for “double-blind” experiments in medicine, and so, this is the same feeling when respondents feel a special attitude and special care. Those who later assess the consequences should also be in ignorance as to who felt or did not feel that particular relationship. Such experiments are an indisputable fact in life, but they are usually ignored. Thus, you should never believe the results of carelessly conducted experiments when they involve people. The authority of the experimenter, the complexity of the equipment, the degree of data compression and, especially, your faith should not shake you. Once again, be very,

The Hawthorne effect shows us that the correct teaching method will always be in a state of experimental change, despite what has already been done. All that matters is the professor and the students who believe in change.

Let me turn to the history of using computers to improve the learning process. In 1960, when I was on sabbatical at Stanford, a program for graders worked there. Any programming task for students required the professor to provide a complete computer program, the names of the input parameters, the range of the input parameters, as well as acceptable output values. When the students felt that their programs were ready for delivery, they called the teacher, confirmed their identity, after which the computer generated some valid input parameters and ran both programs (professors and students), and then compared the results. Each output parameter was labeled "True or False." Thus, the teacher could easily estimate the compilation time and program execution time,

This method was flexible, easily adaptable to changes and was applied from year to year in solving certain problems. The program kept the journal in a closed database, at the request of the professor provided the raw data, while leaving the definition of the actual estimates at the mercy of the professor. Of course, from this database the professor could get the average grades, the difference in grades and the distribution of grades by groups.

When I visited Stanford a few years later, I asked about the fate of this program. And it turned out that it is no longer used. You will ask why? Because, as I was told, the very first professor who started using this program has quit, and changes made to the monitoring system would require several changes to the program! Careful observation and questioning showed that a similar situation existed in many campuses. The computer is programmed to help, apparently, the professor, but soon the program was forgotten.

Let me now turn to the PLATO project developed by a friend of mine at the University of Illinois. I often meet him at various conferences, once we flew in a plane for a long time, and each time he tells me about his wonderful project PLATO. For example, he once told me that one student from Scotland, Canada, and Kentucky were studying simultaneously in the PLATO system. I told him that an ordinary telephone company could have done the same thing, but what he said to me was in no way related to how PLATO worked better compared to the work of people. He never, in my memory, provided any serious evidence that PLATO significantly improves the learning process - this is what you would expect from the Hawthorne effect.

One of the statements stated that a student applying the system is 10% higher in the level of his education than someone who does not use the system. When I asked if a 10% growth was typical of the entire educational system or was only inherent in a particular subject, he could not answer this question for me! What did he do with the Hawthorne effect? Nothing! Thus, I do not know whether or not the goals for which millions and millions of dollars were spent from the federal budget were achieved or not.

Once, when I was the chief editor of ACM Publications, the book “Programmable Book” came to me for publication. This book regularly asked questions to its reader, and then, depending on the answer, the reader was redirected to one of the pages of the book. In principle, errors in the answers were caught and explained, after which the correct answers were sent to the reader as new material. Sounds cool! Each student worked on his own program. But look, with this approach, there was no way to track and search for what the student had already read several pages back and it is not clear at the moment how and from where he reached the current position. There was no sensible way to track it through the text. In fact, this is not a book, although it seems so. Another terrible fact of careful observation of students in the process of reading a book was that strong students were choosing obviously wrong answers to questions simply out of boredom or for entertainment. So they wanted to find out where the book would send them. Based on this, the book did not always fulfill the function that it was supposed to perform; strong students do not necessarily show significant progress in learning compared with weak students!

I did not want to refuse to print this book, relying only on my own opinion, so I went to the psychology department of the Bell Telephone Laboratory and found a local expert there. Among other things, he said that next week there will be a big conference about the “Programmable Book” and invited me to visit it, which I did. On the opening day of the conference, we sat down next to each other. He pushed me and said: "Notice, none of them provided any concrete evidence in favor of the book, they only say that the programmed text is better." And he was right - they all only expressed their opinions without providing any evidence. I refused to print the book and, looking back, I think I did the right thing. We now have computer disks that are designed to do the same,

We looked at the negative side of CAI. We now turn to the positive. I have no doubt that teaching a boring arithmetic, such as addition and multiplication tables, is better done by a computer than a teacher, especially after incorporating a simple error trapping program into the process and generating more examples, such as multiplying by 7. I think that You will agree with me on this. Unfortunately, in the future we expect that corporations and other large organizations will refuse such a way of learning (computers will do it better and cheaper), and the decision to adopt such an approach will depend entirely on you.

Now consider the process of training the pilot of the aircraft on modern simulators. Simulators show good results compared with any life experience and, as a rule, during training, pilots have little experience in training interactive interaction. Flying, to a large extent, and I point this out, is a conditioned reflex that is embedded in a person in the process of learning. This is not very much thought about, although sometimes it is necessary, but training, mental and physical, contributes to a quick and correct response in case of unforeseen emergency situations.

It seems to me that such a training, in which a conditioned reflex is imparted into a person, computers perform very well. It turned out that in childhood I was engaged in fencing. In a duel, there is no time to think, you must quickly act reflexively. In fact, the duel is usually carefully planned, but at the crucial moment it is necessary to act without delay in thinking.

When I first got into the Naval graduate school in 1976, there worked a dean of the department, who was concerned about education. In some of the hot topics related to education, our views were different. Once I came to his office and said that I taught students weightlifting (he knew that was not true). I announced to the students that only the one that can push 250 pounds will pass the exam. My words discouraged many students and they gave up, some went to train and only a few successfully completed the course. I continued my story. Last night, I thought for a long time and decided to simplify the task by simply dividing the weight in half — the students will be considered as having completed the course if they can lift 125 pounds, fix the weight, then again raise the remaining 125 pounds, so the total is 250 pounds.

I waited a bit until he stopped laughing (just like you, anyway), after which I noticed that when I found a simpler proof of the theorem and used students for this, did I reduce the weight in half or not? What will be your answer? Is there no truth that the easier we make the learning process for students, the more we cut the weight in half? It does not follow from this that students should be given easier tasks in order for them to work harder. Evidence of the ability of people to make a significant contribution leads us to the conclusion: the famous professor was a terrible lecturer, and students must work harder to study subjects on their own.

I again propose to draw attention to the rule:

What you have learned from others you can use to follow.
What you have learned yourself, you can use to lead.

To what extent will it be correct to compare physical muscles with "mental" muscles? Undoubtedly, these two types of muscles are not the same, but how long will they be considered analogs? Think about this question yourself.

My other argument in opposing the same dean was his belief that students should be allowed to take independent promotion courses that were in his area of ​​responsibility. In response, I retorted that learning speed plays a significant role for organizations — all other things being equal, students who quickly mastered the material had a higher value than those who slowly mastered the material; One of the objectives of our work was precisely to increase the speed of study and the allocation of such students for society. Again, this is an opinion, but there is no doubt that you don’t want to deal with very slow students. Of course, speed in learning new things is not everything, but it seems to me that this is an important element.

The main problem in assessing the value of CAI is that we are not ready to say what an educated person is and how we are now attaining this state (if we are reaching!). We can say WHAT we do, but this is not the same thing WHAT we should do. Therefore, I can only tell a few more jokes.

Consider this statement: a good graphic material provides significant assistance in the study of basic concepts. It sounds great, but let's turn to the story I told you about my friend Kaiser. He, having studied the theory of filters in terms of time and voltage, could not cope with an independent variable - energy. It should be noted here that Kaiser was a very intelligent man, but his education limited him in what he learned. The better we introduce the idea with the pictures drawn by the professor, the more we prevent the student from extending the ideas to completely new areas that the professor did not think about (and which they did not include in the pictures).

Let me tell you a story about the so-called “transfer learning” theory - the application of ideas from one area of ​​knowledge to the ideas of another area of ​​knowledge. At the very beginning of World War II, I taught a computation course at the engineering school of Louisville. The students had problems with the course in thermodynamics, which was taught by the dean of the engineering department, the former submarine commander, and frightening the students. With the permission of the dean, I decided to attend a lesson in order to understand the reasons for poor student performance. And so the dean writes the following formula on the board:


and asks what it means. In the audience silence, no one knew the answer. A few hours later, I wrote on the blackboard in my class


and the students already knew the answer: it was the logarithm of x plus a constant. When I wrote on the blackboard


the students again knew the answer. “Why did you not answer this question in class with the dean?” I asked the students. The fact is that what they knew in one particular class at one particular hour with one particular professor did not translate into knowledge at another particular hour in another class from another professor. It sounds strange, but this is called the “transfer of learning,” that is, the ability to use the same ideas in a new situation. Thanks to this theory, I made a significant contribution to the development of Bell Telephone Laboratories, I did it quite often, although, of course, I do not know how many chances I missed.

Let us now return to my course on computation, which I often taught at the Naval Postgraduate School, although these thoughts were formed long before me. Students are able to remarkably memorize their way through a large number of math classes, and many really do it. But when I turn to analytical integration (I give students a function and ask them about its indefinite integral), they cannot remember their way from the course in which I teach them. They must learn to recognize


in any endless situations. For the first time in their career, they are forced to learn to recognize forms that do not depend on a particular idea, which in turn is the main feature of mathematics and general intelligence. Removal of analytical integration from a course or its transfer to subprograms on a computer means the collapse of the idea of ​​a study phase, in my opinion, of something substantial until something equivalently complex has been invented and implemented. Students should master the techniques of abstract pattern recognition, if they want to further progress, and use mathematical models throughout their future career.

A very similar mistake was made many years ago when I was a student at the University of Chicago. The Department of Education has launched a primary school for research. They found that students learn to read by syllables rather than by letter, and therefore they decided to skip the study of the alphabet and start reading immediately. What they actually did. Everything went pretty well until the end of high school, when they found that students do not know the alphabet completely and therefore cannot effectively use dictionaries, telephone directories, etc. At their present age, it was almost impossible to make them memorize the alphabet so that they could easily use such sources of information. Therefore, I am wary of the proposed changes as long as

Summarize. I believe that in low-level situations, usually associated with training, where a conditioned reflex is required, computers will make a significant contribution to improving the learning process. As for high-level thinking, i.e. education, here I am skeptical, mainly because we ourselves have no idea what we want to do or what we are doing at the moment. Now we just don’t know what we mean when we say “educated person”, not to mention the meaning of this expression by 2020. Without this knowledge, how can I judge the success of the proposed solutions? Organizations and universities should explore the vast area of ​​knowledge that exists between low-level education and high-level education. Experts rarely make a significant contribution to the development of this direction. significant progress usually comes from the side. We will discuss this further with you in Chapter 26. The role of CAI in organizations that widely use training programs will increase, as progress leads to the constant obsolescence of old and the creation of new, usually more technically complex, tools to use.

Consider computer programs that are designed to teach such disciplines as business management or, more seriously, military affairs. Computers can take on a large number of small tasks, isolate the player from them and at the same time provide useful high-level solutions. The process also needs to include some elements of low-level learning, as well as higher-level thinking. As noted in the three chapters on simulation, we also need to know if simulation is suitable for future tasks, for which, in fact, training is intended. Will the availability and, possibly, the spread of game programs lead to a deterioration in the quality of education? However, you can be sure that even if people offering such programs cannot answer these questions, they will still create and advertise relevant programs. You may be the victim of improper learning!

A few hundred years ago, standard higher education consisted in teaching reading, writing, and speaking Latin, as well as, to a lesser extent, teaching Greek and classics. This was the basic education with which the English, for example, created an empire. Our current education system has very, very little in comparison with the classical system. I firmly believe that future education will have as little in common with the current education system as the current system differs from the classical one. Attempting to make minor changes to our current education system will not help to solve the problem that we face in preparing students for 2020, when desktop computers become equally accessible and have huge capacities for storing and processing information. Without a clear idea of ​​what kind of education will be relevant by the time, how should we evaluate the proposed CAI projects? Just because something can be done, especially using computers, does not mean that it should be done. We must create a vision of what an educated person will be in the future, and only then can we confidently approach the problems that arise during CAI.

To be continued ...

Who wants to help with the translation, layout and publication of the book - write in a personal or mail

By the way, we also started translating another cool book - “The Dream Machine: The History of the Computer Revolution” )

We are looking for those who help to translate the bonus chapter, which is only in the video . ( translate by 10 minutes, the first 20 have already taken )

Book content and translated chapters
  1. Intro to The Art of Doing Science and Engineering: Learning to Learn (March 28, 1995) Перевод: Глава 1
  2. «Foundations of the Digital (Discrete) Revolution» (March 30, 1995) Глава 2. Основы цифровой (дискретной) революции
  3. «History of Computers — Hardware» (March 31, 1995) Глава 3. История компьютеров — железо
  4. «History of Computers — Software» (April 4, 1995) Глава 4. История компьютеров — Софт
  5. «History of Computers — Applications» (April 6, 1995) Глава 5. История компьютеров — практическое применение
  6. «Artificial Intelligence — Part I» (April 7, 1995) Глава 6. Искусственный интеллект — 1
  7. «Artificial Intelligence — Part II» (April 11, 1995) Глава 7. Искусственный интеллект — II
  8. «Artificial Intelligence III» (April 13, 1995) Глава 8. Искуственный интеллект-III
  9. «n-Dimensional Space» (April 14, 1995) Глава 9. N-мерное пространство
  10. «Coding Theory — The Representation of Information, Part I» (April 18, 1995) Глава 10. Теория кодирования — I
  11. «Coding Theory — The Representation of Information, Part II» (April 20, 1995) Глава 11. Теория кодирования — II
  12. «Error-Correcting Codes» (April 21, 1995) Глава 12. Коды с коррекцией ошибок
  13. «Information Theory» (April 25, 1995) (пропал переводчик :((( )
  14. «Digital Filters, Part I» (April 27, 1995) Глава 14. Цифровые фильтры — 1
  15. «Digital Filters, Part II» (April 28, 1995) Глава 15. Цифровые фильтры — 2
  16. «Digital Filters, Part III» (May 2, 1995) Глава 16. Цифровые фильтры — 3
  17. «Digital Filters, Part IV» (May 4, 1995) Глава 17. Цифровые фильтры — IV
  18. «Simulation, Part I» (May 5, 1995) Глава 18. Моделирование — I
  19. «Simulation, Part II» (May 9, 1995) Глава 19. Моделирование — II
  20. «Simulation, Part III» (May 11, 1995) Глава 20. Моделирование — III
  21. «Fiber Optics» (May 12, 1995) Глава 21. Волоконная оптика
  22. «Computer Aided Instruction» (May 16, 1995) Глава 22. Обучение с помощью компьютера (CAI)
  23. «Mathematics» (May 18, 1995) Глава 23. Математика
  24. «Quantum Mechanics» (May 19, 1995) Глава 24. Квантовая механика
  25. «Creativity» (May 23, 1995). Перевод: Глава 25. Креативность
  26. «Experts» (May 25, 1995) Глава 26. Эксперты
  27. «Unreliable Data» (May 26, 1995) Глава 27. Недостоверные данные
  28. «Systems Engineering» (May 30, 1995) Глава 28. Системная Инженерия
  29. «You Get What You Measure» (June 1, 1995) Глава 29. Вы получаете то, что вы измеряете
  30. «How Do We Know What We Know» (June 2, 1995) переводим по 10 минутным кусочкам
  31. Hamming, «You and Your Research» (June 6, 1995). Перевод: Вы и ваша работа

Кто хочет помочь с переводом, версткой и изданием книги — пишите в личку или на почту

Also popular now: