papert - Warning Concerning Copyright Restrictions The Copyright law of the United States (Title 17, United States Code) governs the making of photocopies or other reproductions of copyright material. Under certain conditions specified in the law, libraries and archives are authorized to furnish a photocopy or other reproduction. One of these specified conditions is that the photocopy or reproduction not be “used for any purposes other than private study, schohu- ship, or research.” If a user makes a request for, or later uses, a photocopy or reproduction for purposes in excess of “fair use,” that user may be liable for copyright infkingement. The Children’s Machine RETHINKING SCHOO, LINTHE AGEOFTHECOMPUTER l *o I I Seymour Papert BasicBooks A Division of HarperCalli~ i~ c’ Copyright Q 1993 by Seymour Papert., PubIished by BasIcBooks, A Division of HarperColIIns Publishers, Inc. AII rights reserved. Printed in the United States of America. No part of this book may be reproduced in any manner whatsoever without written permission except in the case of brief quotations embodied In critical articles’and reviews. For information, address BasicBooks, 10 East 53rd Street, New York, NY 10022- 5299. \ 8 Designed by E& n Leuhe Library of Congress Cataloging- in- Publication Data Papert, Seymour. The children’s machine: rethinking school in the age of the computer/ Seymour Papert. p. cm. Includes bibiiographical references (p. ) and index. ISBN 0- 465- 01830- 0 (cloth) ISBN 0- 465- 01063- 6 (paper) 1. Computer assisted Instruction. 2. Education- Data processing. I. Title. LB1028.5. P325 1992 371.3’34- 4c20 91- 59012 CIP / 9495% 97CC/ CW987654321 1 2 3 ,4 5 6 7 8 9 10 Preface vii 4 Acknowledgments xi I i: Contents ’ 0 .. a Yearners and Schoolers Personal Thinking School: Change and Resistance to Change Teachers . . A Word for Learning An Anthology of Learning Stories ,82 106 Instructionism versus ConStructionism Computerists Cybernetics What Can Be Done? ./ 179 a 205 Sources of In@ mation 227 Bfbliography Index 229 E 232 it 1% . THE CHILDREN’S MACHINE. I ,; relationship of knowledge to media. The tmditional epistemology is based on the proposition, so closely linked to the medium of text- written. and especially printed. Bricokage and concrete thinking always existed but were marginalized in scholarly con- texts by the privileged position of text. As we move into the computer age and new and more dynamic media emerge, this will change. Although it might be futile to outguess such radical depar- tures in ways of dealing with knowledge, it will be interesting to keep the question in mind as we turn now to look more directly at some aspects of the history of computers in relation to episte- mology and learning. Computerists T E pioneers who made the first computers knew exactly what kind of work the machines would do and what style of mind they would serve. It was the 1940s. The world was at war. Complex calculations had to be done under time pressures not normally felt by mathematicians: numerical calculations re- lated to the design and use of weapons; logical manipulations to break ever more complex codes before the information became old news. The pioneers were mathematicians and built the ma- chines in their own image. It is unlikely that they gave even a passing thought to making computers user- friendly to people with softer styles than theirs. The conditions were set for the develop ment of a computer culture with no room for pluralism; its episte- mological norms would be firmly planted in the most analytic tradition. It was inevitably a culture of “bards.” Wartime conditions were not the only factor shaping the com- puter culture in this way. The stage of development of the technol- ogy acted in the same direction. The very appearance of the early machines would strike terror into the technologically faint of heart. The first one I saw (the British ACE designed by Alan Turing himself) looked less like a machine than a robots’ library with Ifit? l THE CHILDREN’S MACHINE racks of electronics ‘in place of books. No way of using it would have made it congenial to a technophobic teacher tentatively exploring her first relationship with a machine! In addition to their appearance, the technical weakness of the machines contributed to forcing ,a very hard- edged way of using them. Interfaces like those that make today’s computers more “friendly” require lots of surplus computer power. In those days one always had to squeeze the last ounce of power from the machine to get even the simplest jobs done, and this often meant carrying out contortions of mathematical computation in one’s own mind. I remember my first experiences of programming as being much more like solving problems in number theory than the self- expressive activity I as- cribed to Debbie or Brian or the Costa Rican teachers. The point I am making is not simply that this was a mathematical culture (which it was), but that it was the particular kind of mathematical culture in which precise calculation plays the dominant role and the technical and analytic have more weight than the intuitive and the experiential. Thus, many factors conspired to cast the early computer culture in the hard and analytic shape that for most people remains even today synonymous with the word computer. After the war the computer slowly moved out of the sanctums of high science and the military into a wider world of business and run- of- the- mill industrial and university research. As it did so it took its culture with it, and so the popular image of the computer as “analytical logic engine” grew up and took root. What is significant here is how elements of the original computer culture persisted even when the technology no longer required or favored them. Once launched, the culture acquires a logic of its own. Although some of the mathematical extremes of the early ways to control comput- ers were gradually softened, the hard core remained. When I programmed the ACE I actually had to express instruc- tions as sequences of OS and 1s coded by literally punching holes one by one in an IBM card. I do not remember the code, but similar codes still exist for modern machines: For example, the in],; , ‘i era . . . ‘- :.: :! ,. computetlsu l 159 sequence 11000010111010111OOOOO100011100 could be an instruc- tion to the central processor to add the numbers in two given memory positions. But although these codes still have theoretical importance, someone writing a program today rarely uses them as the actual medium of expression. Expressing instructions as binary numbers is too opaque and tedious even for a mathematician to find comfortable. It did not take long before computer languages were developed to allow an instruction to be expressed in a form more like z = x +y, to mean that the numbers in the memory positions x and y are added together and the result placed in memory location z. One of the intellectually powerful facts about computers is that they can ma- nipulate their own programs: Since the computer can be pro- ’ grammed to translate z = x +y into the appropriate binary num- ber, the only time it is absolutely necessary to use the binary code is to write the program that does the translation. The development of more transparent and congenial forms of. expression did not mean an end to the hard- edged analytic style of thinking in programming; it only softened its most obtrusive manifestations. The mark’ of the mathematician was still there in I the algebraic form of the instruction, and it was stamped on the culture of programming in deeper ways than this. As one might have expected, it was mathematicians with a hard- edged bent of mind who were most inclined to create theories of the proper structure of a computer program and make the effort to set up standards for the process of writing one. The result was to consoli- date their view of programming as the only right one. Thus a new kind of factor became visible,, which still buttresses the hard- edged computer culture today. The hards have an advantage in the ability and desire to offer theoretical justifications for their ways of doing things. A similar self- perpetuating factor works through the recruitment of people. The dominance of the hard- edged style in the culture draws new recruits who think in that way, and discourages those who would tend to push its develop- ment in another, softer direction. 160 l THE CHILDREN’S MACHINE ‘, As the computer spread to wider worlds of application, the idea of using it in education was bound to come up. Indeed, by the early l!% Os an unfamiliar set of actors had become visible on the fringes of the education scene. The technology we brought with us (for I was one of these compute& s attracted by the prospect of change in education) was extraordinarily primitive. A typical project of the time would sit a child in front of a clattering teletype machine connected with a distant computer that was too big and expensive to bring to the child. There was none of the graphics, the color, the action, and the sounds that contribute to the excite- ment of the computers children know and love today. Very little of what was actually done or learned under such circumstances is directly applicable today. But in contrast with the ephemerality of the technological forms of those days stands the resilience of the theoretical orientations- the ideologies- we brought with us from the larger computer ,culture. The important and lasting side of what we did was planting the seed of a specifically educational computer culture. The theme of this chapter is the development of this seed into a tree with so many branches that I shall have to be selective in discussing them. In selecting the branches that seem most important I have concen- trated on those in which I have been most active. I hope this is not because I see importance only where I have worked; I prefer to believe that this is because I have tried to work in the areas that are most important. The easiest way to tell the history of the educational computer is quantitative. In the W6Os we were a small handful, mostly of academics who had strayed in from other fields: for example, Patrick Suppes from philosophy and psychology, John Kemeny (who invented BASIC) from physics and university administration, Donald Bitzer (who developed the PUTO system) from engineer- ing, and myself from mathematics and the study of intelligence. There were also a few entrepreneurs who lost money in prema- ture attempts to commercialize the field. In the early 1970s we ) ’ cornpar- . 161 were a larger handful. The big break came with the advent of the microcomputer in the middle of the decade. By the early 1980s the numbers of people who devoted a significant part of their profes- sional time to computers and education had shot up from a few hundred to tens of thousands. By now it is in the hundreds of thousands, most of them teachers, although many thousands are engaged in the research and business wings of the world of edu- cational computing. The story that is harder to tell but also far more important to know is subjective and sociological. It concerns what these grow- ing numbers of people think and how the development of this culture relates to wider trends in society. My overarching message to anyone who wishes to influence, or simply understand, the development of educational computing is that it is not about one damn product after another (to paraphrase a saying about how school teaches history). Its essence is the growth of a culture, and it can be influenced constructively only through understanding and fostering trends in this culture. The first significant move toward taking understanding beyond a quantitative level was the attempt to classify the modes of use of computers in education. The title of one of the first anthologies of papers in the area provides a witty formulation that illustrates the approach. The book by Robert Taylor (professor at Columbia Teachers College and creator of the first Master’s program in com- puters and education) was called 7he Computer in the School: Tutor, Tutee, Tool. The intention of the first and last terms of the subtitle corresponds closely enough to popular models of what computers can do in education. Examples of the uses of comput- ers considered as tools will be familiar to everyone. A word processor is considered to be a tool; so is a program that allows one to study ecology through simulations; and so are programs that allow one to use the computer as a calculator. The term tutor names the most common image of the computer in education. The term tutee, on the other hand, refers to a metaphor I have frequently used in thinking about programming as teaching the 9 P D 162 l THE CHILDREN’S MACHINE computer. Every professor knows that a good way to learn a subject is by teaching a course on it, and I half playfully suggested that a child could get some of the same kind of benefit by “teach- ing,” that is to say, programming, the computer. A slightly different classification that has been so frequently used that I have not been able to identify its original author talks about “learning with the computer; learning from the computer; and learning about the computer.” with corresponds neatly to tool andbm to tutor. The relationship between about and tutee is less direct but still exists, in that being able to program a com- puter is synonymous with learning more deeply about how it works than is required by the other two modes of use. In this chapter, however, instead of classifying ways of using computers, I look at the development of ways of thinking about their uses. I suggest a way of thinking about successive periods of their history, defining these as “classical,” “romantic,” “bureau- cratic,” and finally “modern.” Looking back, I think of the earliest period (corresponding very roughly to the l!% Os) in the development of educational comput- i ing as “classical” in a sense intriguingly resonant with Webster’s definition: “conformity to established treatments, taste or critical standards . . . attention to form . . . regularity, simplicity, balance, proportion and controlled emotion (contrasted with romantic).” There was conformity in a double sense. We each came into education from another established field and continued to con- form to a set of methods, tastes, and critical standards that were a meld of the prevailing, hard- edged computer culture and our own home disciplines. At the same time, perhaps because we felt we were guests or immigrants, we structured our work in ways that did not challenge School’s fundamental assumptions. Even I, who was a Yearner of long standing and the maverick of the early community, cast my ideas in what I see now as a remarkably Schoolish mold. Emotion, to continue down Webster’s list of characteristics, was certainly controlled; indeed, it was not even ‘CompUM . 163 acknowledged as a relevant category for thinking about educa- tion. The prevailing computer culture favored keeping our focus firmly on the cognitive side of education. A look at three participants in the early educational computer culture, Suppes, Kemeny, and myself, will be sufficient to show how its “classicism” cuts across ideas and debates about modes of use of computers in education. Patrick Suppes became the intellectual father of CA1 (Computer Aided Instruction), a phrase that has become synonymous with the mode of use of the com- puter I characterized with some polemical overstatement as using the computer to program the student. John Kemeny was one of the fathers of BASIC and therefore a pillar of support for a very different view of the computer: The student programs the computer and so makes it a tool that aids learning rather than a robot teacher that aids instruction. Thus, along one axis Suppes and Kemeny stood at opposite extremes. But on other relevant axes they were very close. They shared a virtually exclusive em- phasis on the cognitive side of learning: They saw learning in terms of facts and skills to be acquired; they had no explicit concern for feelings or for personality or for development of the individual on a level that was not reducible to such spe- cific atoms of learning. They shared an acceptance of School. They kept their views on education separate from their engage- ment with politics, gender, and race. In many such respects they were distinct from the spirit’of the “romantic” period, which would bring hotter social issues and more “intimate” aspects of the computer to the forefront of concern, And so, on the whole, was I. I was certainly the obstreperous maverick of the group. I quar- reled with both CA1 and BASIC and developed Logo as an altema- tive to both. But it would take me five years to understand the anticlassical implications of ideas with which I was grappling. In the meantime I found myself acting like a “person of my time” (or perhaps even like a “man of my time”- my own work has only gradually broken with what I recognize as androcentrism, and 5’ !, i, 1,. : 164 . THE CHI/ IDREN’S MACHINE some of my feminist friends would deny that a male could ever completely break with it). The concept of CAI, for which Suppes’s original work was the seminal model, has been criticized as using the computer as an expensive set of flash cards. Nothing could be further from Suppes’s intention than any idea of mere repetitive rote. His theo- retical approach had persuaded him that a correct theory of leam- ing would allow the computer to generate, in a way that no set of flash cards could imitate, an optimal sequence of presentations based on the past history of the individual learner. At the same time the children’s responses would provide significant data for the further development of the theory of learning. This was seri- ous high science. However, from the beginning several considerations kept the approach from sitting well with me. My gut- level response rejected the status of object given to the child by any theory of this kind. Behaviorists are fond of using the designation “learning theory” for the foundations of their think- ing, but what they are talking about is not “learning” in the sense of something a learner does but “instruction,” in the sense of something the instructor does to the learner. The form in which I was best able to articulate my disagreement at that time was epistemological, that is to say, in terms of differ- ences about the kinds of knowledge being used. Suppes’s instruc- tional theory sought to reduce what children needed to learn in mathematics to a set of precise “factlets” that could be counted and sequenced by his computer programs. In this he was not being idiosyncratic. The logician in him supported a view of knowledge as made up of precise particles; the statistician in him liked to see knowledge as particulate and therefore countable; the neobehaviorist required it to be so. What was expressed’in his work was an all- embracing epistemological paradigm, which was then dominant in large sectors (and is still powerful in some sectors) of the American academic world. From my side too this paradigm was very present in the theoretical worlds from which cotnpum l 165 I came to educational endeavors- but as an obstacle to be chal- lenged. In psychology my mentor Piaget was the most consistent (though in America Noam Chomksy had become known as the most vehement) critic of behaviorism, In artificial intelligence (AI), my work with Marvin Minsky struggled against “logic” as the basis of reasoning and against all forms of “particulate” and “proposi- tional” representations of knowledge. The issue comes out in a stark form by contrasting two views of Debbie. CAI is based on a diagnosis of Debbie’s difficulty as a deficiency of specific items of knowledge about fractions and seeks to cure the problem by supplying them. I see a defi- ciency -or even multiple deficiencies- in relatio~&~ ip: There is debilitating weakness both in Debbie’s own relationship with fractions and in the relationships among the different pockets of’ what knowledge she does have. As a result she is unable to take charge either of making effective use of her existing knowledge or of generating or seeking new knowledge. I pose the educational goal not as giving her factlets but as encouraging her to make connections between different elements ‘of what she already knows: for example, intuitive knowledge about fractions, knowl- edge about the “real world,” and knowledge about strategies of learning. Making the connections is something only Debbie can do. They have to be her connections. . The advocate of CA1 might say, “But we have seen that if you put people like Debbie through our programs their scores will improve. The approach must be right.” Perhaps her scores will improve, but it does not follow that the underlying theory is right. The question always arises: Is there another, equally likely expla- nation? An anecdote points to one. I was observing a child working with a CAI program for multi- plication. There was something strange going on. I had seen the child do several multiplications quickly and accurately. Then I saw him give a series of wrong answers to easier problems. It took me a while to realize that the child had become bored with the program and was having a better time playing a game of his own - t ; 166 l THE CHILDREN’S MACHINE invention. The game required some thinking. It redefined the “correct” answer to the computer’s questions as the answer that . would generate the most computer activity when the program spewed out explanations of the “mistake.” I would bet this child was one of those who would become a statistic showing gain in math ability from the use of the CAI program. Would it follow that the program was in fact a good way to teach math? Yes and no! Yes, because it did, in fact, enable the child to learn; no, because it did so for a reason quite. different from what the programmer. intended. The issue at stake here is whether self- directed activity was better than carefully controlled programmed activity for learning math, and this child supported the self- directed alternative. A CA1 salesperson might still object (though I am sure Suppes would not) that this is of no importance if the child did in fact learn. My reply to that is what 1 say about most learning by rote methods with or without computers: Yes, indeed, children can make a game of anything and learn through t it, but if that’s what we want to see happen let’s say so and work i hard to find contexts in which playfulness is brought out to best I advantage., The anecdote illustrates the difference between the intellectual atmosphere of Suppes’s background and mine. .While he was working in the tightly controlled thinking of logic, I was working in the playful atmosphere of the MIT AI Lab. Of course, neither of r 1 us denied the importance of both formal and intuitive thinking. But we saw a reversal of relationship between them. The logician sees logic as the primary kind of thinking and struggles to explain the intuitive in logical terms. Many of my colleagues in artificial intelligence argued (and some still do) that when we are doing what we think, of as intuitive thought, we are still following (with- out knowing it) precise, logical rules- only they are not the rules we might think that we are following. This is why they are de- lighted whenever anyone programs a computer to do something that resembles intuitive reasoning. The computer is following definite rules, so the task, whatever it is, can be done by following computeris~ 9167 rules. For me the challenge is in the other sense. The basic kind of thought is intuitive; formal logical thinking is an artificial, though certainly often enormously useful, construct: Logic is on tap, not on top. I am delighted whenever something that appeared to be formal, rule- driven behavior turns out to be something else. This is why I was so pleased with the child playing with the CA1 program. Thus Suppes and I differed quite deeply about what kind of knowledge we wanted to foster in children. A mark of authenticity in our debate is that we differed also in our own personal styles- both in how we thought and in our appreciations of how we thought. In one of my first encounters with Suppes he had formulated the issue of styles by summing up a debate between us at a conference on the philosophy of science. I remember his words well, since they came to be emblematic for me of what I often saw as the fundamental problem of teaching: “I would rather be pre- cisely wrong than vaguely right.” 1 I saw this as a fundamental problem for teaching in this sense. It had been obvious to me for a long time that one of the major difficulties in school subjects such as mathematics and science is that School insists on the student being precisely right. Surely it is necessary in some situations to be precisely right. But these situa- tions cannot be the right ones for developing the kind of thinking that I most treasure in myself and many creative people I know. This is not thinking that goes as the logician might like, from truth to truth to truth until it gets from premise to solution. The normal state of thinking is to be off course all the time and make correc- tions that go back sufficiently to keep going in a generally good direction. This kind of thinking is always vaguely right and vaguely wrong at the same time. The teaching dilemma comes from the difficulty in knowing where someone else is in such a process. So how can the teacher give advice to the student? I had devoted much effort to looking for a theory of how 168 l THE CHILDREN’S MACHINE teachers could do this. I now see that I had not been able to find anything very deep for reasons rather like those that blocked the teacher at my Logo workshop: I was fixated on children in School and so was looking for ways to improve the guidance process in traditional schoolwork. The breakthrough that set me on track to what would become my trademark way of using computers came when I was able to “forget the - children” and think about myself. It happened on a visit to Cyprus in 1965. I was still reeling from the culture shock that came with moving (in 1363) from the Uni- versity of Geneva, where there were no computers, to MIT, where I suddenly had free access to the best machines in the world. Here on this remote Mediterranean island I was feeling my first absence from a way of life in which computers were a constant presence. This in turn stirred up thoughts about how much I had learned since coming to MIT, how I had used a computer to make ri breakthrough on a theoretical problem that had bothered me for some time, how concepts related to computers were changing my thinking in many different areas. Then in a flash came the “obvi- ous” idea: What computers had offered me was exactly what they should offer children! They should serve children as instruments to work with and to think with, as the means to carry out projects, the source of concepts to think new ideas. The last thing in the world I wanted or needed was a drill and practice program, telling me to do this sum next or spell that word! Why should we impose such a thing on children? What had launched me into a new spurt of personal learning at MIT wasn’t in the slightest bit like the CAI programs, I became obsessed with the question, Could access to computers allow children something like the kind of intellectual boost I felt I had gained from access to computers at MIT? In a search for good examples of what children might actually do with computers, my mind raced through my own activities, making lists of ways in which I thought I had benefited from computers and asking myself in each case whether something similar could be made available for children. For a while I simply complb- l 169 passed over the first entry on my list: artificial intelligence, the principal interest that had brought me to MIT. “Obviously not for children.” Then I remembered a conversation with Piaget a few years before in which we had engaged in playful speculation about what would happen if children could play at building little artificial minds. I had been saying that the essence of AI was to make theoretical psychology concrete. So (since concreteness is supposedly what children thrive on) in principle perhaps some elementary form of it could become a children’s construction set. If psychologists could benefit from making concrete models of the mind, why shouldn’t children, whose need was even greater, also benefit? Piaget liked the image of taking one of his favorite apho- risms-“ to understand is to invent”- into a new domain. In the hothouse atmosphere of the discussion in Piaget’s incredibly cha- otic study, we were carried away by images of children under- standing thinking through playing with materials needed to invent a thinking machine, an intelligence. Neither of us thought of it as very real- it was just a scenario for a philosophical Gedanh- periment. But now suddenly on a mountain in Cyprus, the idea changed for me from a philosophical speculation to a real project. The difference came from a very concrete picture of (one ver- sion of) what people actually do when they “do AI.” They select a piece of human mental activity, say, playing chess or seeing a cat; then they write a computer program that will do something simi- lar; and finally they discuss, sometimes at very great length, whether the computer program “really” does what the human did. I had been engaged in a lot of this kind of activity and knew it had stimulated me to exciting and productive insights into human thinking. True, I did not often really think that the AI program was successful in fully imitating a person; but even when the differ- ences were more prominent than the similarities, the discussion of the machine still produced valuable insights into how people think- and into how they do not think. It seemed plausible that doing elementary AI could give children, too, a new context for 3 Q 170 l THE CHILDREN’S MACHINE thinking about thinking. Of course I would not expect them to be able to make a program to play even poor chess, so I cast around for a simpler game and fixed on something in the family of games played with piles of matchsticks. The principle, however, could be the same. My hopeful scenario of children doing elementary AI was something as follows. A group of children is studying the matchstick game called “Twenty- one,” in which two players take turns in removing one, two, or three matches from a pile of twenty- one matches; the one who takes the last match loses. The children’s immediate goal is exactly that of people making what would later come to be called expert systems: Carefully watch someone engaged in the activity you want your program to imitate, and try to come up with rules you can put into a program to make the computer act similarly. The physical side of the process did not seem important. Today the means exist for children to build a robot that would play by actually picking up sticks. I actually built, one for relaxation (and because I like that kind of joke) while writing this chapter, using an extended Lego kit called Lego- Logo, which was a fallout (twenty years later!) of the very work I am describing here. In state- of- the- art present- day educational computing, the game would be played using computational objects visible as icons on a screen- the computer would play by moving the icon from the row into a bin and the human opponent by dragging it with a mouse or by using keys. Back in the 1960s we used clattering teletype machines; when we got the scenario to work, the matches were Xs and the machine retyped the row after each move. The human player responded by typing a number. What seemed im- portant was how the children would do the programming and, indeed, whether the idea ran counter to established knowledge on stages of intellectual development. My friends in the developmental psychology business were cynical about whether anything that could significantly be called programming could be managed by children who had not yet reached the so- called formal stage of development, which means about junior high school age. I saw the question as more subtle because I was more aware of how much it would depend on what is meant by “programming.” It seemed intuitively obvious that nothing good would come of trying to have third- graders or even fifth- graders make game- playing programs from scratch in any of the then current programming languages such as FOR= or LISP. (Had BASIC or PASCAL existed I would have included them as well.) But was this because the languages were designed for adults and presupposed some elements of mathematical sophis- tication, or was it inherent in the concept of programming? In- deed, is there such a thing as “the concept of programming,” or is “programming” something that can be constructed in radically’ different ways? One could go around in circles forever with such questions. The only sensible approach was to take a first shot at making a programming language that had a better chance of matching the needs and capabilities of younger people than the existing ones. At the time, while working at MIT, I was also doing some part- time consulting for a group led by Wally Feverzeig, head of educational technology at the research firm Bolt, Beranek, and Newman, which was working on one of the earliest attempts to teach pro- gramming in a school. The group did not need much persuasion. to change its goal from trying to teach existing programming languages to developing an entirely new language. We formed a team, and the next year the first language bearing the name Logo was up and running -though few of the million or so children who work with a modem form of Logo on any school day would recognize it. We decided that it was prudent to make the first trial at the junior high level just inside the “formal” boundary; the idea was to descend to lower ages as we developed techniques for teaching and improvements to the language. It took two years and a lot of work to get from Cyprus to a place where young people (seventh- graders) could actually do some- thing like the scenario. Not only did they do so, they even pro- duced an unexpected twist that made the reality more interesting 172 . THE CHILDREN’S MACHINE than the fantasy. Rather than follow strictly in the path of the so- called 46knowledge engineers” who build expert systems, these students followed in the path of psychologists who deliberately construct a series of “inexpert” systems that make the computer act like a “novice” and then pass through a progression of levels of increasing expertise. Unsurprisingly (in hindsight!) our young students were more intrigued by what some of them called “dumb programs” than by “smart programs.” It might be fun to make a program that would play impeccably and always win, but some found it more enjoyable to make one they could beat and laugh at for commit- ting the blunders they saw their friends make. The real case ex- ceeded my original fantasy scenario in giving rise to good talk about much more than computers and programming. In one class the use of the words dumb and s) nart became a subject of intense discussion triggered by an interchange in which A said B’s program was dumb and B countered with something; like: “It’s not dumb, I did it specially like that so I could add more rules. Wait and see, it’ll be the smartest! Real dumb is when you make it so you can’t add anything, it can’t get better.” In another class, discussion led to arguments about whether these judg- / mental words should be applied to the program or to the pro- \ grammer and ended up with a consensus: Using words like dumb and smart is what’s really dumb. B reminded me of Patrick Suppes’s comment about being vaguely right or precisely wrong, by defending a strategy of delib- erately designing a program that would be only vaguely right but capable of being redirected, instead of shooting for being pre- cisely right on the first shot and risking a complete miss. In this he expressed the same thought that underlies Voltaire’s maxim, “the best is the enemy of the good,” which Herbert Simon, Nobel Prize- winning economist and one of the founders of. AI, takes as his life motto. All three thinkers, Voltaire, Simon, and B, suggest a slant on what is really wrong with School’s epistemology: the very little room it leaves for being vaguely right. conlpule* Lrts 9 173 It is not merely an intolerant style of teaching or testing that is responsible for the insistence in so much of schoolwork on being exactly right. The content of the cu, rriculum and the medium of pencil and paper are inherently biased toward a true/ false- right/ wrong epistemology. What B discovered was that programming is inherently biased toward evaluation not by “is it right?” but by “where can it go from here ?” In this he is not alone: Many virtuoso programmers insist on starting a job by making a “quick and dirty” program that is vaguely what is wanted and then seeing how to go from there. Of course, the same is true in other (perhaps all) domains of creative work. My interpretation of such stories as B’s will be that while he could have made this discovery in other domains (obviously!- many did so before computers came on the- scene), programming in the right supportive context offers especially favorable conditions, and the more so the younger the discoverer. The game of Twenty- one turned out to be simple enough to be played by programs within the grasp of seventh- graders who were indeed able to draw on the experience for discussion of strategies for thinking. Students took with gusto to making programs that would generate sentences in approximate English and through doing so came to a new kind of understanding of grammar. But something was missing, and the idea of children doing AI did not really take off until we marrled it, nearly twenty years later, with Lego to produce a construction set for building programmable robots. The difference between these two situations touches the heart of my story, indeed of this whole book. But I am getting ahead of myself. . At the time 1 was both happy and frustrated. The experiment showed that seventh- graders could learn Logo and do some of the things I had hoped to see. I don’t put that down; it was a signifi- cant finding. There was no doubt that some of the students did get an intellectual boost. Several who had been “average” students became straight A students. 1 felt confirmed and was beginning to dream ambitiously of really making a difference in how children 174 l THE CHILDREN'S MACHINE learn. Seventh- graders are scarcely children, however, and I felt that if contact with computers were destined to have an important effect it would be at a much younger age. Yet it was obvious to me from the texture of the work that extending it downward in age was not simply a matter of developing teaching techniques. The more I got a deep feel for what it was like to work in the language as it was and to work on the kinds of projects we had been using, the clearer it became that my psychologist friends had been ‘right: If this is what programming means, it’s not for pre- formal children, Nevertheless, I knew there had to be another way to approach the problem. It needed a radically different idea. The idea took a while to come and an even longer while for me to recognize what it was. At first I was blocked by looking too hard for something too new in a way that often happens. After- ward you realize that you had the solution to the problem all along, but you couldn’t see it because you were straining your eyes and stressing your mind looking out there into the great blue yonder. In this case I found the solution when I stopped taking myself so seriously and looking so hard for something new. The new idea came from looking in a more relaxed way at what was in hand. I was doodling at the computer as I often do by writing little programs with no particular importance or difficulty in them- selves. You could call it just playing. I don’t know what such activity does for the mind, but I assume it’s the same as what happens when one draws patterns or pictures with pencil on paper while thinking or listening to a lecture. What happened this time came from thinking that writing programs can be like draw- ing in many ways. In a way the Twenty- one program is a represen- tation -might one say a kind of picture- of the form of a mental process, just as a pencil and paper drawing can be a representa- tion of a physical shape. The knowledge engineer’s manner of work even has something in common with the portraitist’s. The artist looks at how a person appears and tries to capture features in the medium of pencil and paper or paint and canvas. The knowledge engineer looks at how a person acts and tries to capture essential features in a computational medium. These analogies quickly become strained, but they led to a shift in my perception of what was important in the Twenty- one program. Previously I would have said that what was important about the program was that it represented a kind of thinking. Now I wanted to say that what counted was that it represented something the programmer does. It didn’t matter that the something was think- ing; it could just as well have been walking or drawing or what- ever. In fact, maybe walking or drawing would be better than playing Twenty- one; children care more and know more about these activities. The turtle came from thinking about how on earth a child could capture in computational form something physical like drawing or walking. The answer was a yellow robot shaped rather like R2D2 and, like him, mounted on wheels. Nowadays we have much smaller robots with computers inside them. We also have turtles that exist only on a computer screen. In those days the turtle was a large object, almost as big as the children who were using it, connected by wires and telephone links to a faraway computer that filled a room. One could order it around by giving instructions in proper Logo grammar. As for words, a few were built- in (in- nate), and one could communicate in Logo to the computer that one wanted to define a new word. What was most remarkable was that by giving Logo the handful of new commands needed to FORUARD 150 u u RIGHT 90 FORURRD 100 RIGHT 45 u FORURRD 150 j& LEFT 90 A turtle path showing commands FORWARD, RIGHT, LEFT. 176 l THE CHILDREN’S MACHINE control the turtle, the spirit of what could be done with it changed dramatically. Where the day before I was worrying about how to descend a year from the seventh grade, now there was an area of “baby AI” that seemed plausibly accessible to children well below school age. We have met the essential commands. Typing FORWARD 50 causes the turtle to move in the direction it is facing a certain distance, which is to be called fifty turtle steps. Typing RIGHT 90 makes it do what in the military would be ordered by Ri- i- i- i- ght TURN! The turtle stays where it is and turns in place. If it is already ‘moving when it gets the instruction (though this was not possible in older forms of Logo that permitted just one instruction at a time), it changes direction and continues moving in what has now become the direction it is facing. But why should a child want to do this? And why should we be happy if a child does it? When I saw children playing with the turtle they told me a simple elemental answer that resonated with the preoccupations I have mentioned in the last few pages. Their first step toward expressing their answer was jumping all over the turtle and de- manding rides. Step 1: They clearly liked it. When the adults held back they asked one another to type commands to make the turtle move. Step 2: They took charge and used it for their own pur- poses. Some time later children began exercising ingenuity in giving commands that would produce interesting paths. FOR- WARD 50 BACK 40 RIGHT 10 (and keep repeating that) does REPEAT 4 [FORUARO 90 RIGHT 901 1 Combining the command REPEAT with the others. something that some people enjoy. Step 3: It leads to invention. Step 4: It leads to the mathematical discovery that the commands FORWARD and RIGHT are a universal set in the sense that they can be combined to produce any possible path or shape. . Intellectual problems about conceptualizing thinking and the role of computers that had been troubling me began to seem tractable. I watched a boy trying to get the turtle to write his name. He wanted an A. This required developing a little theory of the geometry of an A. It is not obvious how much the turtle should turn nor how much it should advance. This is real geometry. But it differed from School geometry in a cluster of important respects. First, it was a real problem that had come. spontaneously to this boy. Of course, that can happen in- regular geometry too. But it is very much more likely to happen here. Second, one visibly works toward the goal by being wrong most of the time. But one can see that one is wrong and ask oneself or someone else what hap- pened. The movements of the turtle extematize one’s conception so one can think and talk about it. One can also do some of the kinds of “problem solving” that people do in the real world, such as solve another problem instead, or borrow a solution from someone else and adapt it to fit your case. The boy trying to make an A did just that. Initially he wanted to make a 45degree angle at the top, so he instructed the turtle FORWARD 50 RIGHT 45 FORWARD 50. This led to a shock, as you see .by following the picture. What happened? He looked at the wrong angle- the angle the turtle turns is what they call the “external angle” in geometty. This kid didn’t know that, but he got the idea. So after some trial and error and computing he typed Fumbling toward an A, following a model of the A as two leaning lines with a crossbar. Successive attempts are closer. 34 c I c 0 178 . THE CHILDREN’S MACHINE fd 20 P. fd 20. repeat 3 [fd 50 rt 1201 rt 60 fd 50 rt 60 fd 20 rrpaat 3 [fd 50 rt 1201 The new model: The A is a triangle with two legs. The plan for the procedure is: Draw 1 leg (fd 20), then draw the triangle, then manuever the turtle to be in place for the other leg. To follow the steps, pay attention to the fact that repeat 3 (fd 50 rt 120) brings the turtle back to the beginning. Why 120? Because the turtle moves 360 in 3 pieces. FORWARD 50 RIGHT 135 FORWARD 50. But now the problem was to make the cross piece. The first step is obvious: BACK 20. But how much should the turtle turn and how much should it go forward? At this point he noticed that another child was pinning up a large equilateral triangle made earlier by the turtle. You could see the “Eureka!” Why did the A need 45 degrees at the top? If Mary there could make a triangle I can too (he explained afterward that this is what he was thinking). And if I have the triangle I’ll just add on extra legs and it will be an A. He still had the problem of making the triangle, but knowing that Mary had done it was a clue that helped him find out himself. In the end, there was the A. It took another eight years before the observation of differences in style as shown by Jeff and Kevin led to the idea that computers would not simply improve school learning but support different ways of thinking and learning. But as I had watched these children my first inklings of this marked the fact that I was seeing the - computer, and educational computing, move out of the classical period in which it had confirmed and strengthened old ways. Soon there would be many manifestations of the romantic period, in which conformity to old ways of thinking would be replaced by the vision of new ways. 0.. . . Cybernetics T LEVISION pictures of the war over Iraq gave millions of people their most vivid view of cybernetic technology, in the form of the “smart” missile, which seemed to hover like an insect before lunging into the entrance of a hangar or other building. It is depressing to feel again that the best way to open a discus- sion is with a military image, but it reflects a real fact of life that has played a big role in the strategies that have guided my work. The people who forge new technological ideas do not make them for children. They often make them for war, keep them in secret places, and show them in distant views. Even when there is no deliberate concealment, there is a trend nowadays toward opaque packaging of instructive technologies. In a distant past everything a society knew might have been open to its children for use or playful imitation. Even in my youth technological objects were far more “transparent” than now. I know that it’was important to my own development that I could see and at least think I understood the inner workings of trucks and cars, and eventually go through the rite of passage of tuning or even “decoking” an engine and reseating its valves. I believe that the fact that so many people ’