The question of criteria

What do we measure when we evaluate? Artificial intelligence brings back to the surface an old question, that of the function of criteria in school and at university. This question reaches deep into what learning means, into what institutions do to those who pass through them, and ultimately into the very function of school in a society that is changing.

An unquestioned assumption

When we talk about learning, exams, grading, diplomas, there is an assumption that circulates almost always unquestioned. To judge what a person has done or learned, criteria would be needed, otherwise evaluation would be arbitrary. This is what is assumed in schools, at universities, in competitive exams, in recruitment. And this is also what the recent article by Santiago Schnell, published in AOC in May 2026, L’université à l’épreuve des machines (The University Tested by the Machines), takes for granted. Schnell rightly argues that generative AI leaves none of our evaluation practices untouched. As he writes, « fluency, now, is free », which makes the old confusion between the quality of a written production and the proof of a learning process untenable. His conclusion, which advocates a return to tutoring, oral defense, the responsibility of utterance, seems to me to be right. But he does not discuss the frame within which this diagnosis is posed. He takes for granted that the university certifies, that it evaluates, and that it needs good criteria to do so. What I want to dwell on is upstream of that, at the level of the assumption.

What is a criterion for? What interests me is not so much what it says about the object being judged. It is what it does to the person who produces this object, and to the one who judges it. A criterion is a measure, and any measure presupposes a shared unit. Yet human learning, in my view, is not a quantity that can be measured. It is a process of transformation that unfolds singularly in each person, and that does not allow itself to be reduced to a common unit without loss. To claim to measure it is to transform the object of measurement, to turn a singular transformation into a comparison with a norm.

And where does that norm come from? It does not fall from the sky. It is defined by those who are already recognized as legitimate within the field. The sociologist Marie Duru-Bellat documented, in Les inégalités sociales à l’école (PUF, 2002) and then in L’inflation scolaire (Seuil, 2006), how the so-called objective criteria of school evaluation in France reproduce, under the cover of neutrality, the pre-existing social hierarchies. To evaluate in a school setting is almost never to measure a learning process in itself. It is to measure the gap between the student and a socially situated norm, while pretending to measure something else. The philosopher Donna Haraway, in Situated Knowledges (1988), proposes a precious concept for thinking about this. All knowledge, all gaze, all evaluation is situated. There is no view from nowhere. The objectivity of evaluation is a fiction that serves to mask the position of the one who evaluates.

What criteria for a flash of brilliance?

I encounter this question in my own teaching work at the university. In various programs, I am led to ask students to create something. Sometimes these are students training as artists, but more often they are students preparing to become professionals in culture, care, mediation, or industry, without seeing themselves as artists. I still ask all of them to make a creation (a short film, an image, a form), because I believe in what this exercise puts at stake. And the question that is then put to me, the question one would put to any teacher in a university, is that of evaluation. On what criteria am I going to judge what they will have produced?

When I try to answer this question honestly, I run into an aporia. The aim of creation is to invent forms. What would be the criteria that would allow me to judge what makes a good form? If I lay any down, I introduce a norm to which the student’s creation would have to conform in order to be considered valid. But creation consists of inventing a form that did not exist before, and which, in doing so, may establish new criteria. The entire history of modern and contemporary art bears witness to this. The avant-gardes did not conform to existing criteria; they fabricated them. If we evaluate a creation by the yardstick of pre-existing criteria, we refute in advance the value of whatever might emerge as avant-garde, and art only moves through its avant-gardes.

I will be told that a student is not an artist, and that we should not be trapped by such grand words. Yet I know from experience, after years of running this kind of exercise, that a student who does not see themselves as an artist, who is not even in an artistic program, can produce what I would call a flash of brilliance, provided we give them the legitimacy and the space for it. What such a student needs is neither to have first mastered a technique, nor to have been validated upstream by a hierarchy that decided that they had a right to exist. It is to be able to start from their own vision of the world, without being inhibited by the fear of judgment. Inhibition is what the institution manufactures, the moment it places itself in the position of authorizing or not authorizing, where it could simply welcome. This pre-emptive validation is all the more problematic because it is addressed to students who, not being in artistic curricula, have already internalized the idea that creation would be reserved for someone other than them.

Institutions select those who can live in them

The case of competitive sports and disciplines with strong institutional circuits, such as chess, sheds light on the question of criteria from a complementary angle. In these fields, children who show potential are now integrated very early, sometimes from the age of six or seven, into training institutions that move them from one establishment to another, from competition to competition, until they reach professional practice. It is assumed that this trajectory selects the best in terms of competence in the subject being taught. What I observe, in all the fields where I have worked, is that this trajectory selects something else. It selects people who may have skills for chess, basketball, music, or science, but whose selection rests above all on their ability to live in the type of institution that trains them in these disciplines. Among the people the trajectory eliminates, there are people who would have had every capacity to build themselves up in the field, but who did not have the form of mind to function within the frame. It is the institutions that eliminated them, not their abilities on the subject.

This point is central, and it holds for chess as for everything else. For dance, music, the sciences, law, writing. Every teaching institution filters, by its form, people who would have had every capacity to build themselves up in the field concerned, but who do not have the form of mind to function in the frame. And evaluation, by measuring progress within the frame, validates this exclusion by giving it the appearance of a measure of competence. The philosopher and psychologist Vinciane Despret, in Penser comme un rat (Quæ, 2009), has shown for animal experimentation what also holds for human education. A competence is not a property of the individual; it is a property of the relation between the individual and the device that brings it into view. To change the device is to change what we see. To evaluate within a single device is to render invisible all the competences that this device does not bring into view.

The opposite assumption, the one that consists of hierarchizing trajectories according to their degree of institutionalization, rests on a misleading intuition. The player who has gone through all the official stages has, undeniably, skills. The one who has trained on their own, or in an oblique way, practicing a great deal but without integrating the institutional circuits, may not have the same skills. But they may have others, which do not appear on the grid because they belong to a way of thinking outside the frame. And those skills, the step-by-step trajectory does not particularly develop. It even tends, because it integrates the apprentice into a codified tradition, to erode them. There is therefore no reason, in my view, to hierarchize the two types of trajectory. There are two types of people, with two types of competence, which hold in different contexts.

Institutional evaluation maintains this vertical hierarchy, even though nothing in what it claims to measure justifies it. A pedagogical institution that would accept to demolish this verticality and become horizontal again, where each contribution is received for what it brings rather than ranked against the others, would bring into view competences that today remain invisible. This is a political orientation, which nothing obliges us to maintain.

Professional and amateur, and the confusion of identities

A distinction circulates implicitly in teaching institutions, and it is useful to make it explicit because it fuels many assumptions about competence. This is the distinction between professional practice and amateur practice. It is readily assumed that the difference between the two is a difference in competence. The professional would have the skills; the amateur would not have them, or not enough. This representation seems to me false. In the field of music, for example, there are amateur musicians whose musical competence has nothing to envy in that of professional musicians. The difference between the two is not a difference in competence; it is a difference in life choices. The professional has chosen to earn a living from their instrument. The amateur has made another choice. Neither of these choices says anything about the quality of the practice.

This confusion between competence and professional status has a perverse effect that is particularly visible in the age of artificial intelligence. Many people confuse their personal identity with their professional identity. Yet the two are not the same. My professional identity is what I do as a trade, what I am paid for, what inscribes me in an economic and institutional system. My personal identity is who I am, in my relationship to myself, to those close to me, to my commitments, to my creations. When artificial intelligence comes to disturb established professional competences, it is the people who have confused the two identities who feel personally threatened, as if wounded in their very being. Those who have maintained the distinction can pass through this transformation without being destroyed by it, because they know that they are more than what they do to earn a living.

Many powerful figures in a field did not define themselves by their professional status in that field. Leonardo da Vinci practiced painting, engineering, anatomy, music, poetry, drawing, without being professionally assigned to any of these activities in the sense in which we would understand it today. He authorized himself to go where he wanted to go, and it is from this self-authorization that the power of his work comes. Glenn Gould, whom many consider one of the greatest pianists of the twentieth century, did not see himself first as a pianist but as a filmmaker, as Bruno Monsaingeon has documented in his filmed interviews with him. Niki de Saint Phalle, whose work runs through sculpture, painting, cinema, and architecture up to the Tarot Garden in Tuscany, never trained in any art school, and it was from this gap to the institution that she drew the singularity of her path. Louise Bourgeois began by studying mathematics at the Sorbonne before turning to sculpture, and her international recognition came only after the age of seventy, after a long, oblique trajectory. These figures remind us that the creative identity is shaped along paths that pass alongside the official training institutions, and that no institution will ever be able to predict in advance what a person will become if given the freedom to search.

Art is not the object, it is the experience

The American philosopher and pedagogue John Dewey wrote in 1934 a book that I consider one of the most important of twentieth-century pedagogical thought, Art as Experience. Its main thesis is simple to formulate and difficult to integrate. Art is not the object we look at in a museum, nor the text we read, nor the piece we hear played in a hall. Art is the experience that this object allows a human being to live in their encounter with it. The work, without a human experience receiving it, is only an inert object. And the experience, when it is strong, can also arise in front of objects that the artistic tradition would not classify as art. An old plate, an ethnographic spoon, a child’s drawing can open up an experience as intense as the great works of heritage.

This thesis has immediate pedagogical consequences. If art is experience, then the teaching of creation is not first about the production of objects that would resemble objects recognized in the tradition. It is about opening up, in each person, the capacity to enter into experience with what they do and with what others do. And this capacity is no more measurable than creation itself is. It is tested in practice; it deepens through sharing; it is made fragile by the external judgment that would tell the person that their experience is not the right one. When I ask students to create, what I am asking of them, at bottom, is to authorize themselves to enter into the experience of their own creation, and to share it with others so that this experience may resonate with theirs. The criterion has no place in this device, because experience has no criterion.

This thinking has a French formulation that has accompanied me for years. I owe it to Philippe Foulquié, co-founder of La Friche La Belle de Mai in Marseille, who died in 2026, and who simply said « faire pour apprendre à faire », to do in order to learn to do. La Friche, opened in 1992 on the former tobacco factories of the Seita, has been and remains one of the most innovative cultural places in France, not only artistically, but also through its anchoring in its neighborhood, through its original way of thinking the land ownership of a cultural place, and through the cooperative practices that have been invented there. Foulquié’s phrase was passed on to me by Emmanuel Vergès, who long directed the association Zinc at La Friche, and with whom I led several cultural innovation projects, notably in Vitrolles, where we did indeed learn to do by doing. The very transmission of this phrase, from Foulquié to Vergès and from Vergès to me, bears witness to something simple. To learn is to place oneself in the experience of a doing, and the doing transforms one as one engages with it. No criterion measures this transformation, because the transformation is not a finished product that could be compared to a model. It is a process.

Thinking is also a gesture

Another reason makes the measurement of learning problematic, coming from contemporary cognitive science, and from the current today known as 4E cognition (cognition that is embodied, embedded, extended, enacted). This current shows that thinking does not take place only in the brain. It takes place in the whole body, in its inscription in an environment, in the tools it mobilizes, in the gestures it performs. A study conducted with two groups of mathematicians asked to solve problems showed that the group allowed to use their hands and to gesticulate obtained results significantly better than the group on whom immobility was imposed. The hand is not an accessory of thought; it is a constituent of it.

This intuition is not new, and French thought has a strong formulation of it. The archaeologist André Leroi-Gourhan, in Le geste et la parole (Albin Michel, 1964-1965), had laid down that the evolution of human intelligence is inseparable from the evolution of the hand and of tools. To think, for a human being, is not only to reason; it is also to handle, to shape, to make. The Cartesian separation between the mind that thinks and the body that executes is a philosophical artifact that has marked our school institutions but that does not hold up against what anthropology and cognitive science show us.

The anthropologist Tim Ingold, in Anthropology and/as Education (2018), extends this vision by proposing that learning be conceived not as the transmission of a content from a master to a pupil, but as the engagement of several people in a shared inquiry into a world that is discovered together. The position of the teacher within this frame is not that of the one who holds the knowledge and distributes it; it is that of the one who lays down the conditions of a common inquiry, where each participant contributes according to their own attention. Evaluation by criteria assumes that knowledge is in the head of the one who knows, and that it must be transferred as efficiently as possible into the head of the one who does not know, then verified that the transfer has indeed taken place. Pedagogy by inquiry assumes on the contrary that knowledge is built in shared experience, and that it takes different forms in each participant according to what this experience has allowed them to discover. The two models are not commensurable, and the second does not need criteria in order to function.

Artificial intelligence, mental arithmetic, the book

With generative artificial intelligence, an important share of the cognitive work that was asked of the human at school is now delegated to a machine. This is one of the observations made by Laurent Alexandre and Olivier Babeau in their essay Ne faites plus d’études ! Apprendre autrement à l’ère de l’IA (Stop Studying! Learning Differently in the Age of AI), Buchet-Chastel, October 2025. Their formula, on the back cover, is striking. « Intelligence is becoming free and infinitely available. » I do not share the general orientation of their book, which seems to me too fascinated by the rhythms of the digital industry, and which underestimates the human, relational, and political dimensions of what is at stake in a school. But their diagnosis of the gap between school as it exists and the world it claims to prepare for forces me to pause.

The point that holds me lies elsewhere. With artificial intelligence, as with the book, as with automated calculation, a cognitive competence once required becomes tooled. When printing was invented, people gradually stopped learning whole libraries by heart, even though those libraries were the living memory of pre-printing societies. We can deplore this, but we can also note that in exchange, we have freed up brain time for something else, and what humans have done with that brain time cannot be reduced to a tale of cognitive decline. The same with calculators and spreadsheets, which have replaced a great part of sophisticated mental arithmetic. We can lament it. We can also note that people use their brains for other things. And what we postulate when we decree that these other things are necessarily less constructive is not an observation; it is an a priori. We know nothing about it.

This entanglement between human thinking and machines has a history that French thought has sometimes neglected. Ada Lovelace, an English mathematician of the nineteenth century who worked on Charles Babbage’s analytical engine, wrote as early as 1843 in her notes that this machine, if it were given the means, could one day compose music and produce scientific works. This was a visionary anticipation of what we now call generative artificial intelligence. What makes Lovelace’s thought precious is that she saw, before everyone else, that the machine could extend human thinking rather than replace it, and that this extension was not the erasure of the human but the opening of a new field for human activity. The same remark holds for today’s generative AI. With it, as with the previous tools, a part of what students used to write on their own is now whispered, structured, refined to them by the machine. The philosopher Anne Alombert, in De la bêtise artificielle (Allia, 2025), rightly worries about the proletarianization of human expressions by AI, and her warning deserves to be taken seriously. The American neuroscientist Maryanne Wolf, in Reader, Come Home (HarperCollins, 2018), has shown that the reading brain that centuries of printing have built is not guaranteed, and that certain deep readings must be actively maintained for it not to reorganize itself in shallow ways. These warnings seem to me right. But their teaching is not that we should return to earlier cognitive competences as to a norm. It is that we must learn to consciously inhabit the new cognitive devices in which we are caught, which is exactly the task that school could set for itself if it stopped confusing that task with the measurement of individual production.

Technical mastery and duende

The classic objection to what I have just written is that one must first master techniques before being allowed to step outside them. We hear it in every field, in music, in literature, in cinema. This is a complete inversion, and I think we have to name it as such. This inversion says, in substance, to the student that their voice will not count until they have integrated the codes in place, and that only once this integration is done will they be authorized, perhaps, to transgress them a little. It is a denial of the dignity of each person, irreducible to their sheer existence, and, in the terms of cultural rights, a symbolic violation. Each person has a culture and a vision of the world from which they can create, and this base is in itself enough to begin.

The mistake of the objection is to confuse technical mastery with the quality of practice. There are musicians of perfect technical mastery whom we listen to play without feeling anything. No mistake, no approximation, no fragility, and also no emotion. These musicians have what mastery produces, and that is precious. They do not have what Federico García Lorca, in his lecture Juego y teoría del duende (1933), called duende, that burning presence which depends neither on talent nor on mastery, but on a true engagement in what one is doing. Duende, in my view, is not an exceptional gift. It is a quality of presence to the act one is performing. This presence is no more measurable than creation itself, and it is not formed by accumulation of respected criteria. It is formed by engagement, and engagement is cultivated otherwise than by evaluation.

This idea has a concrete pedagogical reach. When we stop wanting to measure the production of students by criteria, we release in them the possibility of another relationship to what they do. The instruction becomes that which authorizes rather than that which constrains. The teacher becomes the one who welcomes singular productions and puts them into circulation, rather than ranking them. What is transmitted between people is not an assessable knowledge; it is a quality of presence to a work, which they can each take up, extend, displace along their own path.

Appreciating instead of measuring

One point remains important not to lie to oneself about. If I say that there are no legitimate criteria for judging a creation, or for evaluating a flash of brilliance, I am not saying that there is nothing to appreciate in what people do. I am only saying that what is to be appreciated cannot be measured. The distinction is crucial.

To appreciate a creation is to enter into a relation with it, to feel its effects on oneself, to formulate what it displaces in us, to share that with others. It is a human act of reception, in which the person who appreciates engages as much as the person who has created. This act is neither neutral nor objective. It carries the subjectivity of the one who appreciates, their biases, their preferences, their moods. It is because it carries this subjectivity that it can be rich for the one who has created. An honest appreciation is worth far more than a grade. It gives the person something to hear, to confront, to take up again. It opens a dialogue, and it is in that dialogue, and not in the grade, that thinking can be formed.

Two simple devices, which I have run for many years, can illustrate this methodological displacement. The first concerns photography. When I work with a group, I propose an exercise in which each person makes photographs for fifteen minutes. The photographs are then put in common, projected in a dark room, looked at together. And at the moment of sharing the gazes, I lay down a rule that always surprises. The people who made the photographs are not allowed to speak about their own work. It is the others who share what they see, what the photograph displaces in them, what they receive from it. The person who made the photograph listens. What is produced in this exercise is, each time, very strong. The people learn about their own images from what others see in them, and this learning is incommensurable with what an evaluative comment would have produced. Instead of the person who has created having to justify their work before an evaluator, it is the others who make themselves available to what the work does to them, and who share it. The function of the device is no longer to judge, it is to mutually enrich experiences.

The second device, which I ran for seven or eight years at my home in Paris, illustrates the same logic under another form. Each month or two months, I invited five to twenty people, some I knew, others I did not, with a single rule. Everyone had to share, for five or ten minutes, a creation, a reading, a song, a dance, photographs, anything. When someone arrived saying they did not want to create, just to watch what the others did, I refused. They had to bring something, even the reading of a text that was not theirs. I ran these evenings more than thirty times, and each time, without plan or program, links were woven between the contributions of people who did not know each other, projects were born, shifts took place. The key, in this device as in that of photography, is that there is no judgment. Everyone accepts to give and to receive, and what is produced in this space is not the measure of a talent; it is the circulation of a shared experience.

School as the place where we change the world

The subject that emerges does not concern only the university. It concerns school as a whole, from kindergarten to higher education, in both public and private. If school is what sorts people according to their capacity to enter the institutional frame for learning a discipline, it has lost the game in advance against artificial intelligence. Machines will always do better than the best students the tasks that school evaluates and that school grades. If, on the other hand, school is what shapes the thinking of people, what accompanies their creations, their inquiries, their commitment, their singularity, then it always has something to do, and artificial intelligence can become an additional tool in this construction, one that does not replace it.

This path is not new. It was traced by Célestin and Élise Freinet from the interwar period onward, with their practices of classroom printing, of correspondence between classes, of cooperation between pupils, of learning methods that begin from lived experience rather than from a pre-existing program. What distinguishes Freinet pedagogy from a simple active method is its political anchoring. For Célestin and Élise Freinet, school is not there to adapt children to a world that would have its pre-existing rules, to which they would have to adjust. School is there to give them the means to change the world, starting from their lived experience, their creations, their cooperation, and their commitment. Élise Freinet, in Naissance d’une pédagogie populaire (Maspero, 1968), bears witness to how these practices were built concretely, in a rural school that had nothing going for it except the attention paid to what the children brought there. The path was reformulated by Jacques Rancière in Le Maître ignorant (The Ignorant Schoolmaster, Fayard, 1987), where he shows that teaching is not a vertical transmission requiring a master more learned than the pupil, but a situation in which the pupil learns by elaborating their own relationship to knowledge. It was thought once more, on the specific side of evaluation, by the Swiss pedagogue Linda Allal, who has shown that evaluation can be formative rather than summative, that it can support learning rather than sort it, on condition of ceasing to serve ranking and becoming again a situated pedagogical conversation.

These paths exist. They are not the majority in institutions, because institutions are also, whether one admits it or not, instances of social ranking. Ranking requires criteria, and the maintenance of criteria in turn locks down the form taken by the institution. When the university presents itself as the instance that validates what is important, it is not only doing pedagogical evaluation; it is doing social ranking. In this logic, the work of evaluation becomes a tool of sorting rather than a tool of formation. It is the very concept of closed-mindedness, in its institutionalized form. And it may be what artificial intelligence disturbs most deeply. As long as producing a plausible text required rare skills, the sorting held, and could claim the appearance of a measure of learning. As fluency becomes free, as Schnell writes, the sorting no longer holds. The institution must decide what it wants to do. Either it reinforces its surveillance tools and tries to keep the sorting by multiplying the walls around evaluation, and that is a path with no future. Or it accepts to reconsider what it is supposed to do, and that is a path that leads back to pedagogies that already existed, for which AI can become the political occasion.

Inventing the world we want to live in

Artificial intelligence did not create the problem of evaluation. I have long experienced evaluation as a false problem, in university programs as in private schools, and I have long groped about, looking for how not to betray the demand of what I propose while still answering administrative constraints. Artificial intelligence has made this question impossible to hide. It is in this sense that AI is precious. It collectively puts us on notice to look at what we are doing when we evaluate, and to decide whether we continue to do it knowing what we are doing, or whether we change our practice.

This decision is not only pedagogical, it is political. A school that sorts is not neutral, because it reproduces a certain form of society and excludes a certain form of mind. A school that shapes thinking, creation, shared inquiry, opens onto another form of society, in which the singularity of people has a chance of being welcomed rather than compared. The choice we are faced with is therefore not only one of education, it is one of a model of society. With the arrival of artificial intelligence, we are changing worlds, as we changed worlds with writing, with printing, with electricity, with computing. And it is up to us, as with each previous change of world, to invent the world we want to live in, rather than awaiting it as if it were to be imposed on us from elsewhere. This is what Célestin and Élise Freinet had understood when they placed printing at the heart of the classroom. Technique was not for them what threatens school; it is what children can appropriate in order to speak in their own name, and thus to take part in the fabrication of the common world.

These stakes also extend far beyond the strict frame of school. In school, they are visible because they are documented there, evaluated there, discussed there. But we learn everywhere, and we invent the world everywhere, through our attitude in our professional, associative, cultural, militant, intimate contexts. A team that decides to function horizontally, an association that invents a cooperative way of inhabiting a territory, a collective that proposes a new way of transmitting, are doing pedagogy in the strong sense of the term. La Friche La Belle de Mai, which I spoke of above, is one of these places where pedagogy comes about without the formal frame of school, and where each person learns by doing, in cooperation with others. To document these pedagogies that take place outside school, as we document those that take place inside school, is part of the collective work to be carried out so as not to leave to artificial intelligence, or to other powers, the task of defining alone what learning would mean.

I do not know how this decision will be made collectively. I only know what I do, in my own teaching practice. I continue to ask my students to create. I continue to refuse to evaluate their creation by criteria. I continue to organize, around these creations, exchanges in which each person is to appreciate, react, argue, debate, without that leading to a ranking. And I observe that what is produced in these exchanges is, in the long run, more formative for people than any grade could be. This is the intuition I wanted to share, so that it may circulate and meet other practices.