Schank, Roger C. (1995) What to Know, How to Learn It. In: John Brockman and Katinka Matson (eds) How Things Are. New York: Morrow: 183-190.


Roger C Schank

What should an educated person know? In school, little time is devoted to answering this question. The school already knows what you need to know: vocabulary about phyla, the plots of various pieces of last century's literature, and how to prove a theorem about triangles. When you try to get computers to know things in order to make them act "intelligently," however, it turns out that these are not at all the sorts of things they need to know. Computers need to know how to do things, how to comprehend what others have done and said, and how to learn from the mistakes it makes in doing all this.

To educate a computer means giving it the ability to make inferences (if John hits Mary it means she is probably hurt, angry, and may hit him); to infer beliefs (if the U.S. bombs Iraq the U.S. must believe that violence is justified by the need to cater to one's economic interests); and to learn from failure (when you order filet mignon on an airplane and it is overcooked mush, you might want to remember this the next time you're ready to board a plane and get a yogurt first). These are exactly the same things people must learn how to do. We know that a computer, for instance, can be programmed to have encyclopedic knowledge about random facts, but I don't believe that this would mean it is "intelligent", any more than I would consider a person who merely has the ability to spout random facts to be intelligent. But despite what we know about how people learn and the very make up of intelligence, schools remain firmly grounded in the learning model that emphasizes facts and downplays doing. This separation of learning from doing is very detrimental to all.

It has become fashionable recently to define intelligence by using various "literacy lists." The bookstores are full of lists of different kinds of facts--scientific, cultural, even religious facts--all purporting to explain exactly what it is that a person must know to be "literate." The idea here is that being educated means knowing stuff. Implicit in all this is that we have, as a society, agreed on what stuff everyone should know, and decided that information delivery is the role of education.

Do not believe it. There is no set of stuff that everyone should know. What? No George Washington? No Gettysburg Address? It doesn't hurt to know these things, of course. But it does hurt to adopt the position that since one should know these things, teaching them to students is what learning is all about. This makes school a fairly boring, stressful, and irrelevant place, as you may have already discovered.

Facts are not the currency of learning, nor does mastery of them indicate anything about an educated person. Facts play a big role in the education system because they are so easy to test. And, it is tests (usually highly irrelevant tests) that have helped shape your learning since you were six. Curiously, most important things that people know they cannot explicitly recall or state as facts. What is the right way to get the person of your dreams interested in you? How does one pursue a successful career? Was the United States wrong to believe in "Manifest Destiny"? Is the situation in Bosnia really all that similar to Nazi Germany, or is it more like Vietnam? An educated person might have answers for these questions. But they are not simple questions and there are no simple answers for them. Being educated means being able to understand the questions and knowing enough relevant history to be able to make reasoned arguments. Making reasoned arguments, not citing history, is the key issue here. Learning to think and express what one has thought in a persuasive way is the real stuff of education.

What is the currency of learning? It is a preparedness to be wrong, a willingness to fail, and the ability to focus on one's confusion in hope of being able to create or being able to understand an explanation that will make things clearer.

For this reason, the way that stuff is imparted is far more important than the stuff itself. It isn't what you know but how you come to know it that matters. Typically, what we learn in school comes to us through the process of memorization. But memorizing something doesn't mean being able to remember it later when that information might be useful. Information acquired in one context cannot be readily used in another context. While it is not easy to resist a school's attempts to force you to memorize facts, it is important to recognize that merely memorizing things doesn't mean you'll know much. Being able to articulate facts is useful for passing tests, impressing your friends and doing well on quiz shows, but for little else.

We learn best what we want to learn--information which helps us accomplish goals we have set for ourselves. Computers--intelligent computers--can help us do this by providing safe but exciting environments in which to learn. Such computers can be "taught" to ask questions, provide helpful information, and be endlessly patient as a user tries to solve problems.

The Institute for the Learning Sciences built a computer-based exhibit for the Museum of Science and Industry in Chicago which shows how this process can work. The museum wanted to teach visitors about sickle cell disease; Sickle Cell Counselor does just that by allowing visitors to play the role of genetic counselors. Of course, the visitors have no intention of becoming genetic counselors. But by presenting them with a challenging problem, their natural motivation to learn is captured.

The problem posed by the program is to advise couples worried about having children because they suspect their children may be at risk for inheriting the sickle cell gene. Sickle Cell Counselor provides simulations of actions genetic counselors actually take with their clients, and provides access for the user to human experts (via videotape) who help solve the problems at hand.

Sickle Cell Counselor continues to be a popular exhibit at the museum, where visitors often spend up to half an hour exploring its varied opportunities for learning. This is much more time than museum visitors generally spend at an exhibit, but they stay because the learning experience offered is a realistic one, providing goals they become interested in accomplishing.

Not long ago I asked students in an undergraduate class what they had learned recently. They recited facts they had learned in other classes, but they had no idea when they would ever use this information again. When I asked a class of graduate students the same question, they only told me things they had learned about living. They had just rented apartments for the first time, so I heard a lot about how to cook and clean, but they also talked about what they had learned in school that had been of use in projects they were now trying to complete. Graduate students typically focus on accomplishing tasks. They learn what they need that will help them with those tasks. For them, learning means acquiring knowledge in service of a goal. But unless you consider passing tests a goal, you won't find this pattern repeated very much before graduate school.

To make computers intelligent we must teach them to direct their own learning. We can't just pour facts into computer memories, because they won't know what to do with the knowledge they've gained. But if they acquire knowledge in the course of doing something, then the placement of that knowledge in their memory is easy: it gets placed at the point where it was learned so it can modify the procedures that were wrong when that new knowledge was acquired.

You, too, must learn to direct your own learning. Context provides structure for learning, so putting yourself in many different situations, or tying many different things, is very important. Not knowing how to do something in a situation causes a person to focus on learning how to accomplish the unknown, so keep trying to do new things, and attempt to understand why you fail and why you succeed. Don't allow people to lecture at you, imparting information you don't want. On the other hand, you must demand to be taught after you have struggled to accomplish something and have had difficulties.

The key to learning, after you've tried to do something, failed, and availed yourself of needed help, lies in the process of generalization. It is not sufficient to just learn how to function in a given situation, you must also learn to generalize what you have learned so that it applies elsewhere. If you cannot do this, you will have gained a narrow collection of unrelated expertise, useful only for individual domains, but useless otherwise.

It is not possible to make generalizations after acquiring information without consciously attempting to do so. However, attempting to do so involves coming up with generalizations that are inherently untestable, that remain only hypotheses, and that are not in themselves facts. Even so, do not be afraid to test out your generalizations on people you know. They will undoubtedly tell you that you are wrong, but ask them to support their assertions. In general, people are afraid of new generalizations because they cannot know if they are right or not. People fear the unknown, but real learning and real insight depend upon an examination of the unknown and perhaps the unknowable. Propose new generalizations and be prepared to defend them. Ignore new facts that were presented to you unless they were presented in order to help you do something--answer a question, or alter one of your generalizations.

To see why facts don't matter, consider the value of a typical fact that most literacy lists would agree that everyone should know: "Columbus discovered America in 1492." Of what import is this "fact"? Most facts are oversimplifications of very complex events and whey they are learned as facts, they lose all their interesting properties. What difference does it make when this event took place? What matters is that something took place, that we understand the events that led up to that something and the consequences that that something may have had for our lives. There may well be much controversy about Columbus's discovery of America but there is less controversy over the fact that Columbus's act opened a chapter in the history of the world that had quite important ramifications. This would matter if you were thinking about Bosnia, Iraq, or the plight of the American Indians, for example, and that would be when you might want to learn it. Learn it at a different time, apart from any context, and it will be useless.

If there are no facts worth knowing, then what is worth knowing? First, there are skills, in particular basic skills such as reading, writing, and arithmetic. Also, there are other, less basic skills, such as speaking well, relating to others, understanding the world you live in. Processes, too, are worth knowing: political processes, psychological processes, physical processes, and economic processes. Understanding how things work so that you can work with them and make them work for you is also important. Cases are worth knowing. What is a case? For any subject matter in which a student shows an interest, it will be interesting for that student to hear stories that illustrate truths about that subject, exciting things about that subject, the experience of others with that subject, and so on. Getting computers to have a large "case base" in terms of which they can comprehend new cases and propose new generalizations turns out to be the critical issue in creating artificial intelligence.

One's own experience is, of course, worth knowing. Since we learn best by doing, students must be given real things to do from which they can acquire their own "case base." The best way to learn about a political process for example, is to engage in one. So, a lot is worth knowing, but no facts are worth knowing in and of themselves. To become educated you must direct your own education. To learn, you must do, and dwell on what prevents you from doing, so you can change your knowledge base and try again. Seek out confusion in order to learn from it, and pay only as much attention to school as you have to, remembering that school and learning have little to do with each other.