A Rather Stupid Turtle
The turtle could not do very much. It could move forward, backward, and turn. To get it to draw a square, you had to tell it how far to move, when to stop, how many degrees to turn, and how many times to repeat the operation. If the instructions were wrong, the square simply did not appear.
In the late 1960s, when computers were still enormous machines, remote from the lives of the vast majority of people, Seymour Papert was determined to put them within children’s reach. He did not want to teach them computer science or prepare them for a future labor market. He wanted to find out what happened when a child had to explain something to a machine that obeyed meticulously but understood nothing.
The answer was Logo, the programming language developed beginning in 1967 by a team that included Papert, Wallace Feurzeig, and other researchers. Its most famous character was that turtle: first, a small robot that moved across the floor and, later, a cursor on a screen.
Imagine that we want to make it draw a square. For us, “square” already contains a great deal of information. We recognize the shape without having to think about how it is constructed. The turtle, by contrast, does not know what a square is. You have to explain it: move forward, turn 90 degrees, move forward again, turn again. And then discover that those instructions have to be repeated four times.
Suddenly, geometry is no longer about recognizing a shape drawn in a book. You have to reconstruct it from the inside.
Papert even encouraged children to stand up and become turtles themselves. To walk around the room, stop, turn their bodies, calculate the angle. He called it playing turtle. The body became a way of thinking through a mathematical problem and then translating it into the language the computer understood.
Papert had spent several years working with Jean Piaget in Geneva and was deeply influenced by his idea that children construct knowledge by acting on the world. He later moved to MIT and came into contact with another world: that of researchers trying to build intelligent machines.
From that encounter between Piaget and the first computers emerged an excellent question: what would happen if, instead of using machines to teach children, we let children teach machines? Because teaching something to that turtle required understanding it in a way that school rarely demanded.
The Right to Be Wrong
Sometimes the turtle got it wrong. Or, more accurately, it did exactly what it had been told to do, which did not always coincide with what the child wanted it to do. A turn of 80 degrees instead of 90 could ruin a square. A misplaced instruction sent the turtle off in an unexpected direction. The drawing on the screen made it possible to retrace the path back to the mistake. You had to look, formulate a hypothesis, change an instruction, and try again.
In programming, this had a name: debugging. Papert saw in this practice something that went far beyond learning to code. Error ceased to function as the endpoint of an exercise. A program that does not work yet is an open problem: it forces you to ask what happened and what modification might bring you closer to the desired result. In Mindstorms, Papert explicitly contrasted this logic with the school culture of right and wrong answers.
The difference matters. A calculation crossed out in red tells the student that something is wrong. A turtle that ends up facing the wrong way shows the consequences of the student’s reasoning and allows them to intervene. Correcting requires understanding.
This was where constructionism began to take shape, the theory of learning Papert would develop from Piaget’s constructivism. Children construct knowledge, but that process becomes especially powerful when they also construct something that exists outside their heads: a drawing, a program, a machine, a story, an object they can look at, show to others, discuss, take apart, and make again.
Mathematics was particularly fertile ground for experimenting with this idea. Papert recalled that, as a child, he had been fascinated by gears and that those objects had helped him think about mathematical relationships. He dreamed that computers might become something similar: objects flexible enough for every child to find in them a gateway to ideas that might otherwise seem abstract.
Hence the title Mindstorms: Children, Computers, and Powerful Ideas. The “powerful ideas” were precisely those capable of changing the way a person thinks about a problem. But there was more. Programming required making thought explicit. To get the turtle anywhere, the child had to break an intention down into steps, anticipate what those steps would produce, and explain each one precisely.
The machine did not think for the child. It returned the child’s own thinking in the form of action.
Papert was interested in the computer because it could change who asked the questions, who made decisions, and who had permission to experiment.
Who Programs Whom?
While Papert was experimenting with Logo, other researchers were also imagining the future of computers in classrooms. One possibility seemed obvious: use their capabilities to teach better. The machine could present an explanation, ask a question, check the answer, and decide which exercise should come next. The teacher would have a more efficient tool, and each student could progress at their own pace.
But Papert was deeply suspicious of this path. In Mindstorms, he framed the difference through an inversion that would ultimately encapsulate much of his thinking: rather than using a computer to “program the child,” he proposed that the child should program the computer.
It was not a play on words. In the first model, technology improved an educational relationship that already existed. The computer could replace the worksheet with a screen, correct exercises more quickly, or adapt them, but the student essentially remained in the same position: receiving problems designed by others and expected to find the prescribed answers.
In the second, the computer changed hands. The child could decide what they wanted to build and use the machine to make it happen. Perhaps a geometric figure. Maybe a game. Perhaps an animation. Knowledge then emerged in connection with the problems that arose during the project. Something had to be learned because it was needed in order to continue.
Papert did not dismiss teaching, nor did he imagine that children should discover the entirety of accumulated human knowledge on their own. His criticism was directed at a school organized around content divided into subjects, sequences established in advance, and answers that somebody already knows.
In his view, the computer opened up the possibility of altering that organization. It also changed the traditional distribution of authority. A child might know more Logo than their teacher. Two classmates might find completely different solutions to the same problem. The teacher was no longer necessarily the person who knew the path in advance and could instead become someone who helped investigate.
That is why Papert was much more than a pioneer of educational technology. He was interested in the computer because it could change who asked the questions, who made decisions, and who had permission to experiment.
The Revolution School Swallowed
Computers eventually arrived. The revolution, however, proved far more elusive. During the 1980s, Logo spread with the arrival of personal computers. That turtle, born in research laboratories, appeared in schools in numerous countries. For a moment, it seemed possible that learning to program might become a way of transforming children’s relationship with mathematics, science, and knowledge.
But something happened that Papert would observe with frustration: schools had an extraordinary capacity to absorb a new technology without changing very much. The computer could become just another subject. Logo could be taught as another set of instructions to memorize. And a machine conceived as a laboratory for experimentation could end up being used to complete exercises, look up information, or reproduce on a screen activities very similar to those previously done on paper.
Papert came to believe that personal computers had changed society much faster than they had changed schools.
Some of his ideas, however, found other paths. In 1985, he began a long collaboration with LEGO. The combination of building pieces, motors, sensors, and programming brought the logic of Logo into the physical world: build something, observe how it behaves, discover why it does not do what was expected, and modify it. The LEGO Mindstorms robotics line, launched in 1998, took its name directly from Papert’s book.
Later, Scratch would inherit much of that legacy. Developed at the MIT Media Lab under the direction of Mitchel Resnick, a former student and collaborator of Papert’s, it replaced written instructions with blocks that could be combined to create stories, animations, and games. The goal remained recognizable: for children to use code to create and express themselves, rather than simply to learn programming. Educational robotics, maker labs, and much of today’s culture of creative learning also retain something of that genealogy.
But perhaps Papert’s greatest legacy is not any of those objects. It is a way of looking at technology. Whenever a new machine enters a school, we can ask what it does, how much it costs, what content it offers, or how much time it saves. Papert would propose an earlier question: what does it allow a child to do that they could not do before? The difference remains enormous.
Now That Machines Can Think
The Logo turtle had a quality that seems almost exotic today: it was desperately stupid. It could not guess what the child wanted. It did not complete an incomplete instruction, suggest a better solution, write the program, or explain where the error was. To get anything out of it, you had to think.
Half a century later, we are building exactly the opposite. A student can ask an artificial intelligence to solve an equation, write an essay, summarize a text, generate an image, program an application, or explain a concept they do not understand. The machine interprets imprecise instructions, suggests paths, and even corrects its own mistakes. Each new generation of models attempts to require less effort to produce better results.
The reversal is fascinating when viewed through Papert. For decades, we wanted computers to understand people better. Now that they are beginning to do so, an educational question arises that barely existed in front of that turtle: how much thinking do we want the machine to do for us?
The easy answer is to frame the problem in terms of use: allow artificial intelligence or ban it, bring it into classrooms or keep it out. Papert’s work suggests a far more demanding question. What kind of intellectual relationship are we building with these machines?
AI can be used to avoid a problem or to make it more interesting. It can deliver a finished text or allow us to discuss why that text does not work. Also, it can solve an exercise or generate models that the student must compare, take apart, and improve. It can reduce the student’s activity to formulating a request or become material for experimentation.
Papert died in 2016 and never lived to see ChatGPT or the explosion of generative artificial intelligence. It would be risky to imagine what he might have thought of it. But there is no need to do so to recognize the continuing relevance of the question that ran through his work.
In the 1960s, when computers were inaccessible machines that filled entire rooms, Papert imagined children playing with them. When they arrived in schools, he warned that having computers did not necessarily mean learning differently. And as machines became increasingly powerful, he continued to argue that their value depended on what they enabled those who used them to build.
Perhaps that is why that old turtle deserves another look. It was slow, limited, and needed instructions for everything. It could not write an essay or solve a problem on its own. But it forced a child to stop in front of a square and ask how one is constructed.
Now that computers can answer almost any question, Papert gives us back a much harder one: what do we still want children to have to think about?
