Marvin Minsky spent his career arguing that the mind is machinery—not mysterious, not divine, just an assembly of parts that could, in principle, be understood and built. Inventive Minds gathers six essays in which he turns that conviction toward children and education, asking what schools might look like if they took thinking seriously as a subject. Edited by Cynthia Solomon and Xiao Xiao, with introductions and afterwords from colleagues and former students, the book belongs to the tradition Minsky helped found alongside Seymour Papert: constructionism, the idea that children learn best by building things—physical things, programs, theories about themselves—and that computers are the most powerful construction kit yet devised. The essays grew largely out of Minsky’s involvement with the One Laptop per Child project, and they carry the tone of a man who has watched education settle into habits he considers both wasteful and cruel.
“I believe that the obsessive worship of movie, TV and sports figures is less likely to produce spiritual gain than praying to Thor.”
At the center of the book is a claim that sounds simple and turns out to be radical: schools teach facts well but almost never teach children how to think, and the reason is that we have no shared theory of thinking to teach. Minsky does not propose that psychology become a grade-school subject—he doubts teachers could agree on which theories to present. Instead he wants children to invent their own theories about themselves, using the tools of computation to build models of minds, including their own.
From Tinkertoys to Thinking Machines
The Editors’ Thanks and the Man Behind the Book
The volume opens with acknowledgments that map the human network sustaining it. The editors thank Marvin Minsky for setting down his thoughts on children and education, and Seymour Papert for his long collaboration with him. Two figures anchor the project: Gloria Rudisch, Marvin’s wife and a pediatrician and public health practitioner, who opened the Minsky home in Brookline, Massachusetts, for the work and read drafts with a sharp eye; and Margaret Minsky, Marvin’s elder daughter and close colleague, who served as resource and sounding board. The contributors—Mike Travers, Alan Kay, Hal Abelson, Gary Stager, Brian Silverman, Walter Bender, Patrick Winston, and Margaret—each draw out a different facet of Marvin’s words. Cynthia Solomon thanks a long list of collaborators, Xiao Xiao credits Idit Harel and others, and the extended Minsky family appears by name, down to the grandchildren. The book belongs to Margaret’s larger effort to archive her father’s life and work, aided by Yoshiki Ohshima, Brian Bradley, Richard Amster, and others; the South End Technology Center’s Mel King and Susan Klimczak hosted work sessions, and MIT Press and MIT Libraries staff round out the credits.
A Life in Machines and Minds
The short biography sketches a career of firsts. Minsky held the Toshiba Professorship of Media Arts and Sciences and a professorship in electrical engineering and computer science at MIT, and his research cut across artificial intelligence, cognitive science, neural networks, and the theory of Turing machines and recursive functions, along with graphics, symbolic computation, knowledge representation, machine perception, and learning. A pioneer of robotics and telepresence, he designed mechanical arms and hands with tactile sensors and one of the first Logo “turtles,” and he consulted on Kubrick’s 2001: A Space Odyssey. In 1951 he built SNARC, the first randomly wired neural network learning machine, which learned by reinforcing synaptic connections; in 1956, as a Harvard Junior Fellow, he invented and built the first confocal scanning microscope. From the early 1950s he used computational ideas to characterize human psychology and to give machines intelligence. With Papert he began formulating the Society of Mind theory in the early 1970s, proposing that intelligence arises not from a single mechanism but from the managed interaction of many resourceful agents; The Society of Mind appeared in 1985 as 270 interconnected one-page ideas, and The Emotion Machine in 2006 argued that higher-level feelings, goals, emotions, and conscious thoughts emerge from multiple levels of processes reflecting on one another. A classical pianist who improvised in the styles of Bach and Beethoven until his death on January 24, 2016, he was schooled at Fieldston, Bronx High School of Science, and Phillips Academy, served in the Navy in 1944–45, and took his degrees at Harvard and Princeton.
Solomon’s Preface: Thinking About Thinking
Cynthia Solomon opens with a Minsky quotation: computer science is not only about computers but offers a new world of ways to understand complex processes, including those in our own minds—before programming languages for describing processes and data structures for representing knowledge, people had no expressions for good new ideas about such things. She stresses that for Marvin, AI was never about machines with the surface appearance of intelligence but about machines that mimic how humans think; his vision was a quest to unravel the mechanisms of mind. His books proposed theories of common sense, emotion, perception, action, and how thinking evolves in an individual, and his insights matter not only for intelligent machines but for how we view children’s learning and the computer’s role in schools. Solomon cannot think about children without Papert’s line that “you can’t think about thinking without thinking about thinking about something.” She and Papert were close colleagues for years, codirecting the MIT AI Laboratory, collaborating on countless projects, and co-advising students across the AI Lab and the Media Lab; they thought alike, a kind of proof being how they met at a conference in England in summer 1962.
How the Book Came Together
The editor asked Mike Travers, a former Media Lab student under Marvin, to write the introduction after reading a beautiful essay Travers published on his blog when Marvin died. Margaret Minsky’s afterword reflects her understanding of Marvin as researcher, teacher, musician, and family man; early in her research career she belonged to the AI Lab’s Logo Group, Logo Computer Systems, Inc., and the Atari Cambridge Research Lab, working alongside several contributors, and she now creates haptic interfaces and multimedia artifacts exploring learning, improvisation, and thought. Xiao Xiao began as illustrator and became coeditor: as a Media Lab student she built a piano system creating the illusion of a virtual reflection playing the physical instrument, with Marvin as a featured player; she attended salon meetings in his living room and later played piano for and with him weekly, and the editor, catching her drawing in a sketchbook, enticed her to illustrate the book.
Six Essays on Children and Computation
In six essays Marvin shares wisdom about children, learning, school, and computation, insisting that computers are not merely tools for typical subjects like math but chances for children to learn good ways to think about thinking—for instance, by getting them to picture themselves as programmed computers. The editor’s Logo classes had children “play turtle,” become researchers collecting computer and human bugs, discuss debugging steps, and discover that some bugs are features. Marvin observed that “instead of promoting inventiveness, we focus on preventing mistakes,” and suspected this negative emphasis in arithmetic teaching breeds dislike of the subject and aversion to anything that smells of technology.
Travers’s Introduction: Genius Without Mystique
Travers quotes Marvin’s advice: when ideas seem inadequate, imagine what someone more expert would do. Preconceived ideas treat genius as divine mystery; Marvin, a recognized genius, questioned innate talent and its mystifications—to him the mind being beyond explanation was an affront and a challenge. An inventor, mathematician, and musician, his big trick was facing the mechanical nature of mind without alarm, finding it delightful, which put him at odds with standard humanists. He was warm and welcoming, active on Usenet (chiefly comp.ai.philosophy), arguing AI views with all comers regardless of qualifications, and he had a talent for reflecting on mental processes and devising plausible mechanisms, applied to his own mind and others’. The essays contain little specifically about digital technology; their focus is the psychology of learning and systems capable of learning—technology is merely a tool, and programming the best language for expressing ideas about minds. His last two books were crammed with simple, concrete, powerful ideas, essentially technical books in nontechnical language, which caused reception problems.
Situating the Essays: OLPC and Know Thyself
Five essays emerged from Marvin’s participation in One Laptop per Child, a massive effort to put computational technology in children’s hands, which he saw as a chance to fix ingrained bad educational habits. Essay 5 opens by suggesting the focus on broad general education is misplaced; children would be better served specializing and diving into a single topic they care about deeply. To Marvin, students were not empty vessels nor poor approximations of adults but fully active agents and creators of their own minds, needing to hone mental skills on demanding tasks pursued from personal goals; traditional education provides content, but creators require methods and tools. OLPC was a chance “to provide our children with ideas they could use to invent their own theories about themselves”—an education in tune with the Greek injunction “know thyself.” Computers and computational ideas could give better tools for this than ever before; the computational revolution in self-knowledge is enabled but not yet realized.
The Centrality of Goals
Marvin quotes the need for better ways to answer questions kids are afraid to ask, like “What am I doing here, and why?” and observes that “you’re almost always pursuing goals.” Goals and goal-directed machinery were founding principles of AI and cybernetics; essay 6 explains the General Problem Solver, an early goal-directed architecture. The Society of Mind and The Emotion Machine may be read as designs for minds containing and managing systems of goals: innate drives, learned goals, sub-goals, interpersonal goals, meta-goals.
Mathematics as a Way of Thinking
Minsky’s second essay asks why mathematics is so hard to learn, and his answer is that schools teach calculation instead of ideas. Arithmetic could open onto compound interest, population growth, and the way recursion breeds exponentiation—all easy to grasp by experimenting with computer programs, tedious when confined to columns of figures. Geometry could ask how many ways six colors can be painted on a cube, or why gloves come only in left- and right-hand forms; most people never learn to think well about three-dimensional objects, which he counts as a handicap. Logic could teach the common mistakes directly: many adults get it wrong when asked whether “most A’s are B’s and most B’s are C’s” implies many A’s are C’s, or whether John Smith’s move from Apple to Microsoft could raise the average IQ of both companies. Mechanics explains why triangular struts hold: two triangles with equal sides are congruent, so the shape cannot change once the side lengths are fixed—yet most children never meet the idea of degrees of freedom. Statistics deserves its everyday applications, and children should learn the T-test, which is simple, handles huge ranges, and needs only powers of two, along with square roots for judging variation; a basketball score of 103 to 97 is not statistically significant. Combinatorics would show that a 26 percent minority can win under an electoral college, and a mere 7 percent with two tiers. Abstract algebra and topology, though postgraduate, name phenomena—fixed points, symmetries, singularities—that are hard to describe without them.
Two further complaints stand out. Teachers insist that pupils “show their work,” which aids grading but may breed mental slowness by discouraging thinking in the head, where the real skill is holding alternative goals and switching strategies. And school mathematics is a linguistic desert: other subjects teach hundreds of new words a term, while math offers a handful—addition, fraction, quotient, rectangle, equation—and nouns alone will not do. One needs adjectives: linear, quadratic, discrete, continuous, isomorphic. Communities should be arranged so a child can say “nonlinear” and be admired rather than discouraged.
Learning a skill without a conceptual map yields collections of scripts with no way to choose among them. Apprenticeship once supplied supervisors who could answer “What am I doing here, and why?”; modern teachers lack the time, but new networks could give each child a competent adult mentor, and with lifespans growing a year in four, there may soon be more retired people than active ones—an intellectual dating service could pair them. Minsky recalls loving arithmetic because he found many ways to add a column, shaving a 7 to make a 6, turning a 9 into a 10; the children who became engineers did not do things exactly as told, and each invented method left a new mental skill. Aversion to math may come from teaching that insists on rigid skills while discouraging invention. Experts, he suspects, owe much of their competence to negative expertise—learning to avoid common bugs—and counterexamples may teach more than examples.
Stager on Schools and the Case Against Age Segregation
Gary Stager introduces the third essay by noting how little attention Minsky paid to K-12 schooling as an institution. Minsky agreed with Seymour Papert that schools are bad places for children to learn because they are bad places for teachers to learn, yet he never called for destroying schools and remained respectful of teachers, aided by inattention to daily minutiae, bureaucratic lunacy, and crackpot pedagogical theories. Grade-based segregation, Stager writes, has roots in a simplistic notion of industrial-age efficiency; its continuation is naiveté born of superstition and ideological certainty. Heterogeneous, multiage learning is the norm outside school—multiage primary classrooms, instrumental music, choir, drama.
The Citizen’s Objection and the Fifty-Minute Hour
A “Citizen” objects that prioritizing intellectual development sacrifices childhood’s carefree years and neglects love and friendship among children with shared interests. Minsky answers that the classroom is not the whole of life—children also learn from older acquaintances, parents, teachers, mentors—and warns against competitive anxiety, yet argues that excessive attachment to same-age companions may hamper better ways of thinking. He finds no good evidence that age-graded classes produce healthy social relationships. He then turns on the adult assumption that work is serious and play frivolous; closer inspection shows play often involves extreme focus, intensity, and discovery—so much that adults may realize they never worked as hard as in childhood.
Classroom schooling, he continues, requires a synchronized schedule, each activity confined to a prearranged period. This suits management, not the development of children or teachers. The fifty-minute hour is sometimes justified by assumed short attention spans, for which he sees no sound basis; given unlimited time, children sometimes show persistence rivaling adult professionals, especially with hobbies. He suspects important variations in working memory. Student 1 must work on a good idea immediately lest interruption make it fade, sets everything aside for hours, and sits near exits at lectures to escape and contemplate. Student 2 multitasks, dislikes thinking about anything more than a minute or two, and doesn’t mind switching because returning is easy. He recalls no research but suspects some people develop ideas across many separate intervals while others need long unbroken spans.
Note-taking may connect to this. Student 1 learns from notes, since otherwise little is remembered, and a few words suffice to reconstruct much; Student 2 finds condensing what the teacher said such a strain that everything afterward is missed. Children vary greatly in recreating previous mental states, which affects learning through what he calls credit assignments. The essential point of efficient learning: after solving a problem, remembering the answer is not enough—one must remember the strategies used to discover it. Allen Newell made this point in his paper on a smart chess-playing machine; a 1955 quotation argues that “win, lose or draw” contains too little information, so each play must yield much more—achieved goals reinforce subgoals, unachieved ones inhibit them, every tactic informs tactic-search rules, every opponent action informs likelihood inferences. We still lack learning programs that work this way, perhaps because research emphasized statistical inference over reasoning.
Different Cognitive Styles and the Value of Obsessions
There is no typical child, Minsky writes, because each brain holds many kinds of resources whose combinations yield many ways to think. Some children start with abstractions and top-down descriptions; others start with examples and generalize. Some do better with brief study intervals, others with hours; some learn from verbal descriptions, others from pictures, texts, lectures, or tactile handling of materials; some prosper with many easy problems, others with fewer hard ones; perhaps some learn more from success, others from failure—no hard evidence, though mathematicians widely agree one learns less from examples than from counterexamples. Most children have hobbies and obsessions that teachers may see as unrelated to the curriculum. Since each person eventually needs a profession, and quite a few end up with nothing to do—which can be tragic or dangerous—we should not classify a child’s peculiar special interest as an eccentricity interfering with normal progression toward a general education; better to recognize and encourage it.
Children develop at different rates, so teaching the same things to many students in one room means some flourish while others get stressed, some forge ahead while others get lost. Older practice “skipped” pupils a year or two when appropriate; a famous mathematician Minsky knows attributes much of his success to being skipped from Grade 3 to Grade 8. Today this is politically unpopular, though advanced placement survives—for a particular subject, not an entire year. On socialization, it is often assumed that interacting with other children helps social development more than associating with adults, but it might be the other way around. Any method of raising children has consequences, including decisions about influencing their choice of friends; inaction does not avoid responsibility for the outcome. The trouble is we know so little about how children develop that we cannot see what we are responsible for—we need to know much more about how age differences in friendship influence intellectual and social development, and about the tradeoffs involved.
Learning from Role Models, Mentors, and Imprimers
Every person has weaknesses that could be seen as disabilities.
The Rumpelstiltskin Principle and the Case for Naming
The book’s closing stretch opens with a colleague’s recollection of a stubborn integral from freshman calculus. Recalling Minsky’s advice—when in doubt, think about what a program would do—he ran James Slagle’s symbolic integration program, written for a thesis under Minsky’s supervision, and the problem fell apart. The anecdote sets up the argument that follows: push this habit down into elementary school, let children make a program think, and they will come to understand thinking by thinking about it. One idea children gain this way is the power of naming. Minsky insists everyone should know about feedback, but without names people struggle to recognize such ideas or reason about them at all. Gerald Sussman supplies the slogan—“When you name something, you get power over it”—and the narrator dubs it the Rumpelstiltskin principle, after the tale in which knowing the imp’s name is what gives the miller’s daughter her power.
Projects, Papert, and the Difference That Gets Reduced
Many powerful ideas—good representation, the notion of state, divide and conquer, what to do with an adversary—arrive through projects of the kind Minsky and Seymour Papert devised when Papert developed Logo for children. They favored feedback projects because children enjoy programming little robots to follow a track by shrinking their deviations from it; later those children meet difference-reduction again in Newell-Simon-style problem solvers, which apply operators to close the gap between the current state and the solved one. Once the idea has a name, children can put it to work themselves. Some projects looked far beyond anyone younger than a PhD—Minsky being Minsky—yet surprising ones grew from his suggestions, because he treated everyone alike, ignoring age and gender. A decade on, goals like object identification fell to large research groups with massive computing; deep neural nets can now tell cats from dogs, yet Minsky’s suggestions stand, since human vision remains an open, hard problem. He believed perceptual projects could be made easy for children but was less sure about common-sense ones, while noting that what seems hard may prove easy and the reverse.
Diagramming Goldilocks
The narrator’s own thesis runs: we have common sense and can reason, but commonsense reasoning is a special case of recipe following, and recipe following a special case of storytelling. Children should learn perception and action by programming robots, and learn stories through story-understanding projects. Remembering the pleasure of diagramming sentences in elementary school, he proposes that children diagram stories instead: which events cause which, what common-sense rules “Goldilocks” requires, where assumptions must be supplied, pictures of how events fit together, recognition of concepts like revenge. Hand simulation would carry them far; writing programs could come later, with a Logo-for-stories language. Projects on telling stories persuasively would teach persuasion and its techniques, how to judge credibility, how to build fake-news detectors—dovetailing with Minsky’s suggestion that children learn how magic works, the perceptual trickery of the conjuror against the storytelling trickery of the persuasive true story or the propaganda, depending on which side you favor. The payoff might be children who vote better than many adults and prove less susceptible to the stresses of modern life. Minsky closes the essay on thinking with “My” rather than “I”: “My” admits that everyone is made of parts with correctable bugs, while “I” pretends to a single thing with no separately workable parts. We all harbor a part that tells stories about ourselves, rarely bug-free; disparaging self-stories can breed depression, wrong ones dangerous narcissism. The cure proposed is to have children write programs that tell stories about themselves, so they learn to spot and avoid bad storytelling. The narrator likens Minsky’s writings to diamond mines, full of ideas that shine once cut and polished, and confesses this essay took several sessions, interspersed with excited thinking.
What Should Schools Be For?
Minsky’s own essay opens by asking what goals schools should achieve. Most parents agree children should learn history, language, science, and math, with instruction in health, sports, art, and civilized behavior; most would also agree schools should teach good ways to think. Yet schools teach facts well while many pupils fail to build skill at applying knowledge. If good thinking is a principal goal, why don’t schools explicitly teach how human learning and reasoning work? They tacitly assume that enough knowledge lets each child’s brain self-organize ways to apply facts. Should human psychology be a grade-school subject? Not yet, Minsky thinks, because few present-day teachers would agree which theories of thinking to teach. Instead he proposes giving children ideas for inventing their own theories about themselves, through constructive, computer-related projects.
Why Psychology Can’t Yet Be in the Curriculum
Today’s popular theories of learning rest on a century of experiments placing a pigeon or rat in a situation and rewarding it with food for some action; the animal later behaves that way more often. A Thorndike quotation from 1911 states that responses closely followed by satisfaction become more firmly connected to the situation and more likely to recur, while those followed by discomfort weaken. Quicker rewards have larger effects, so educators use problem sets where each task nearly matches the last, making most answers correct and classrooms pleasanter. Such graded assignments prepare for short-answer tests, not real life, where problems don’t arrive in neat sequences. A footnote adds that later researchers found effects last longer when rewards come less predictably, and that punishment suppresses wrong responses only until it stops, sometimes with a rebound, so withdrawing reward tends to work better.
Pain, Pleasure, and the Uses of Failure
The apparent paradox of enjoying one’s own suffering dissolves if the mind is imagined as many concurrent processes: some parts may enjoy making other parts suffer, as when it feels good to discover a new kind of blunder. Coaches teach exactly this; athletes still feel pain, as artists and scientists feel mental pain, but train themselves to keep it from spiraling into the awful cascades we call suffering. A Scientist voice says few things bring more pleasure than replacing old ideas with better ones; an Artist voice says discarding a hard-earned technique hurts, but nothing surpasses the thrill of conceiving new ways to represent things. One must learn not only what works but what to do when failure looms. Minsky dislikes “The Little Engine that Could” and its helpless chant of “I think I can”; a better motto is “perhaps it’s time to try something else,” since every setback offers a chance for a new phase of development. Traditional theories don’t explain pleasure at all. His proposal: pleasure names a process that keeps a mind from changing the subject of its most recent concern, preserving memory contents long enough for credit assignment to finish. So pleasure, usually seen as positive, is also negative, suppressing competing activities; we suppress other interests to keep working on hard problems. Why does it feel good? “I feel good” might mean “Right now I’ve suppressed all my other concerns, so all my problems seem to be gone.”
Teaching Cybernetics Instead of Psychology
Psychology’s ideas move too fast to select any current theory of thinking for the classroom. Instead, give children ideas for inventing their own theories about themselves, from projects making machines with lifelike behaviors—integrating concepts now scattered across physics, biology, mathematics, social studies, psychology, and economics. A flood of such concepts emerged in the 1940s from cybernetics, leading to control theory, computer science, artificial intelligence, and cognitive psychology. In the spirit of Papert’s constructionism, children can experiment with networks of parts supporting knowledge-based processes—important because that is what we are. They could build simplified models of animals’ sensor and motor systems, placing creatures in environments to experiment first with separate behavior, then with social relations in groups. The evidence is ambiguous: past Logo-based projects seemed to work but were never scaled up; today such projects are far more feasible because equipment is much cheaper.
Goals, Bugs, and the Machinery of Self-Image
Running through the essays is a conviction about goals. Minsky notes that we are almost always pursuing something—food, escape, revenge, the end of a workday—yet we rarely ask why some goals feel urgent and others faint, or what happens when ambitions collide. He draws on Allen Newell and Herbert Simon’s early work on goal-directed systems, and on the General Problem Solver, to suggest that minds are architectures for managing nested and competing goals. The educational consequence is sharp: students find mathematics hard because it arrives disconnected from anything they personally want. Teachers transmit content but seldom transmit the goal structures of a field—the questions it tries to answer and why anyone finds them interesting—so children memorize detached knowledge and conclude that the subject is not for them.
The book’s treatment of self-images is its most unsettling thread. Children build representations of their own abilities, tastes, and prospects, and they build them largely from the people around them. Minsky worries that celebrity culture supplies vivid, fabricated models—athletes and pop stars—while philosophers, scientists, and mathematicians go unnamed in classrooms. Athletic prowess, he points out, depends heavily on inherited genes, and media sports stories reduce to injunctions like Try Harder and Don’t Quit, glorifying doggedness over skill. Meanwhile schools explain dropouts through deficiencies of talent or self-esteem, when it might be more productive to ask what self-models those children constructed and where they came from. His remedy is characteristically concrete: have children write programs that tell stories about themselves, so they can learn to notice when the story is wrong.
Failure gets similar treatment. Minsky dislikes the Little Engine’s mantra of “I think I can” and prefers “perhaps it’s time to try something else.” Every setback, in his account, is a chance for a new phase of mental development—provided the child has learned what to do when failure looms. He proposes that pleasure itself may be best understood as a process that keeps a mind from changing the subject, holding attention on a problem long enough for learning to take hold; suffering, by contrast, is the cascade that occurs when pain spirals rather than staying contained. These are speculative claims, offered as suggestions for what neuroscientists might look for, not as settled doctrine.
A Curriculum of Construction
What would Minsky actually have children do? Build things that behave. In the spirit of Papert’s constructionism, he imagines children assembling simplified models of animal sensor and motor systems, placing creatures in environments, and experimenting first with individual behavior, then with groups. Feedback loops—the machinery by which a system compares its situation to a goal and acts on the difference—become a unifying idea, explaining everything from thermostats to exponential growth to chaos. Children who program a robot to follow a track by reducing its deviations have, without knowing the vocabulary, encountered difference reduction, the same principle behind Newell and Simon’s problem solvers. Once the idea has a name, they can use it themselves. Minsky’s colleague Gerald Sussman called this the Rumpelstiltskin principle: name something and you gain power over it.
The book is frank about the limits of its own evidence. Past Logo-based projects seemed to work but were never scaled up; the claims remain partly promissory. What has changed is cost. The equipment that once filled a football field now sits in a few-hundred-dollar laptop, and the Sketchpad tradition Ivan Sutherland began in 1962—with Minsky and Claude Shannon as his advisors—has become something a child can hold.
Why the Book Still Argues
Inventive Minds is not a memoir, though it contains memoir: Minsky’s account of building a towering Tinkertoy structure in a hotel lobby as a small child, and of recognizing his old building-friends in Logo’s commands when Papert first invented it. Nor is it a policy manual. It is a set of provocations from a man who treated everyone the same, ignoring age and gender, and who believed that what seems hard may be easy and what seems easy may be hard. The essays insist that education’s deepest failure is not insufficient content but insufficient method—that children are not empty vessels or poor approximations of adults but active creators of their own minds, and that the computational revolution in self-knowledge has been enabled without yet being realized. Whether or not one accepts his specific proposals, the book makes an uncomfortable case: we have the tools to teach children how to think about thinking, and we mostly choose not to use them.
