Media literacy, as it is most often practised, teaches people to verify: to tell true from false, to trace sources, to recognise manipulation. That is useful. But it assumes that we can set our imaginaries aside in order to look at the facts, and I do not believe that is possible. In this article I draw on Edgar Morin, Cornelius Castoriadis, Jack Goody, Serge Moscovici and Pierre Bourdieu, five authors who, each in his own way, have shown that a fact only exists for us once it is caught up in an imaginary. I then follow, step by step, how the imaginary that now surrounds artificial intelligence took shape, as I have watched it happen since 2022 in the training courses and workshops I run. And from this I draw another definition of critical thinking, along with the workshop method I followed in 2024 with 350 teenagers from TUMO Paris and with librarians in training at the Bibliothèque nationale de France.
Verifying presupposes a subject without imaginaries
When we talk about media and information literacy, what are we talking about? About one operation, sorting true from false, and about the tools for doing it: checking the source, cross-checking information, spotting manipulation techniques, identifying one’s own “cognitive biases”. This last vocabulary has taken up a lot of room in the French debate since the work of the sociologist Gérald Bronner (La Démocratie des crédules, PUF, 2013), for whom beliefs spread because of flaws in reasoning that digital technology amplifies, on what he calls the cognitive market. I have already said, in The fact-checking imposture and in The Lies of Experts, what this approach presupposes: that on one side there are people who know and on the other people who are mistaken, and that the educator is on the right side. In front of some fifteen librarians on a training course at the Bibliothèque nationale de France, I once said that those who believe they hold the right word on media literacy are the first people who would benefit from some media literacy; I told the story in Media and AI education through creativity, and I still think so.
Should we give up verifying, then? No, of course not, and we will come back to it. I only suggest looking at what this method assumes without ever saying so: that with a little rigour, anyone could set their representations aside and reach the facts “as they are”. My position is that nobody can do this, and I am far from the only one to think so. Anthropologists and sociologists wrote it long ago, each with their own tools. Here are five of them, in the order in which they have been useful to me, starting with a book about cinema (that is where I come from).
Morin: perceiving is already an act of imagination
In 1956, Edgar Morin published Le cinéma ou l’homme imaginaire with Éditions de Minuit (in English, The Cinema, or The Imaginary Man), which he subtitled “An Essay in Sociological Anthropology”. He starts from an observation: at the cinema, we project ourselves into what we see and identify with the characters. And he puts forward the hypothesis that this mechanism does not stop at the door of the theatre. For him, projection is a fundamental human process: human beings have always placed their desires and fears outside themselves, in gods, spirits and doubles, and these doubles appear to them as enlargements of the most ordinary everyday life. None of this belongs to cinema alone; cinema magnifies a mechanism and makes it observable. It is the modern institution in which this activity of projection takes place in broad daylight.
Twenty-one years later, in December 1977, Morin wrote a preface for a new edition of the book, and in a few pages he tightened his thesis, which I regard as one of the possible foundations of media literacy. We perceive nothing of the real, he says, that does not take the form of an image; “the only reality we can be sure of is representation, that is to say the image, that is to say non-reality”; human and social reality results from constant exchanges and mixtures between the real and the imaginary. And he adds this sentence:
“The real only emerges into reality when it is woven with the imaginary, which solidifies it.”
Edgar Morin, “Le cinéma ou l’homme imaginaire”, preface of 1977 (my translation).
If we take this sentence seriously, and I do, then bringing people back from the false to the real by dispelling their imaginaries is chasing an illusion. The American journalist Walter Lippmann had written something close to it as early as 1922, in Public Opinion. The first chapter of that book is called “The World Outside and the Pictures in Our Heads”. Lippmann writes there that we act upon what he calls a pseudo-environment, that is, upon the pictures we form of the world, and not upon the world itself. And in modern societies the press is the main supplier of these pictures. This does not mean that everything is equally valid, neither for Lippmann nor for Morin; it means, as I see it, that these pictures are where the work needs to be done. But why are these pictures so similar from one person to the next?
Castoriadis: the imaginary institutes
Cornelius Castoriadis, in L’institution imaginaire de la société (Seuil, 1975; in English, The Imaginary Institution of Society), gives this question a political answer. He calls social imaginary significations what a society uses to decide what exists and what matters to it, and he cites God, the nation, capital, progress, the commodity. I came across one of them, a more modest one, which I described in Lives in Pictures: as a young organiser of short film screenings, I had decided, against my own convictions, to charge for admission, because the audience gave all the more value to the films when they had paid to see them. Nobody in the room had decided that the value of a film is measured by its price; it was there before us, and it governed our evenings. This signification is imaginary, in Castoriadis’s sense, because human beings created it, and it is instituted because institutions, here the price of the ticket, make it exist every day. He returns to this in Fait et à faire (Seuil, 1997), where he states that without radical imagination and without an instituting imaginary, there would be no reality for us at all. To sum up: the imaginary, in Castoriadis’s sense, is not a picture in the heads of individuals that could be corrected one by one; it is what holds institutions, practices and people together. That is why we all have more or less the same robot in our heads.
Two of his distinctions are very useful to me. The first is between the instituted imaginary and the instituting imaginary, and I see it at work from the very start of a workshop. If the chairs are lined up as in a classroom when the group arrives, I have already summoned a school posture, that is, an instituted imaginary, received ready-made and taken for granted; if I suggest that the group arrange the chairs itself, decide who sits where and which rules to adopt for speaking, it creates its own significations on a small scale, and that is the instituting imaginary. I described this in The framework that authorizes. The second distinction is between heteronomy and autonomy. A society is heteronomous when its members attribute its institutions and significations to the gods, the ancestors, nature or the laws of the economy, and so hide from themselves the fact that they are their authors; it is autonomous when they know themselves to be the authors of these significations, and can therefore discuss them and take them up again. In The imaginary as survival equipment I developed how this distinction can be used in the face of crises. Here it will help us redefine critical thinking, and I will come back to it. But before that, something is missing: Castoriadis says what an imaginary does, not how it lives, is passed on and changes. For that, we need an anthropologist.
Goody: the fear of representations is as old as they are
Jack Goody spent decades recording the Bagre, the great initiation narrative of the LoDagaa of Ghana, and this work led him to a conclusion that went against the accepted idea of myths (Myth, Ritual and the Oral, Cambridge University Press, 2010). The narratives of societies without writing are not preserved. At each recitation, the reciter recreates them, with considerable variations, and it is this constant recreation that keeps them alive and adjusted to the present. To sum up: a collective imaginary is not passed on like a content, it is told again, and it lives in this retelling. Goody had shown earlier, in The Domestication of the Savage Mind (Cambridge University Press, 1977), that once thought is written down, placed outside oneself in a form that can be reread, compared and cut up, one can do something else with it; I took up this analysis in The cage of representations, about the categories through which we perceive others. Let us keep one thing from it for what follows: what we have placed outside ourselves, we can look at; what stays in our heads, we cannot.
Goody also wrote directly about our subject. In Representations and Contradictions (Blackwell, 1997), he studies the ambivalence that every society has shown, at one time or another in its history, towards images, theatre, fiction and relics: religious bans on images, book burnings, theatre closures, then rehabilitation a few generations later. Why? Because a representation makes something absent present, Goody explains, so it can deceive, and this, in his view, is the origin of all the objections that have been raised against images. In Petite Poucette 2.0 I analysed the panic over young people’s digital practices; with Goody, I see in it above all a very old fear of representations that has found a new object. Those of us who go to war against images are carried by that imaginary, and that is where we should begin. We still need to understand how an object as new as a language model finds itself, within a few months, caught up in such an old fear. A social psychologist described it before the object existed.
Moscovici: how a new object becomes an image
How could a theory as difficult as psychoanalysis become, in a decade or so, something everyone talks about without having read a line of Freud? This is the question Serge Moscovici asks in La psychanalyse, son image et son public (PUF, 1961; in English, Psychoanalysis: Its Image and Its Public). He gave questionnaires to very different groups of people and went through the press of the 1950s. He first notes that the public retains only a few words from Freud, and that these words are enough for people to talk about psychoanalysis believing they know it; “complex”, for example, passed into everyday language, where we say that someone has complexes. He then notes that each group files it under something it already knows: the Catholic press of the time likens it to confession, the Communist press sees it as a bourgeois science imported from America. Same object, two different places, two different meanings. Moscovici calls the first movement objectification and the second anchoring.
What happens to every technical novelty follows this model. I saw it at work in 2005, when cameras appeared in mobile phones (I told the story in The birth of the Pocket Films Festival (2005)), and I saw it again in 2022 with language models. I will come back to this in detail below. One thing is still missing from Moscovici’s model, social positions, and that is what Bourdieu brings.
Bourdieu: the glasses we do not see
Pierre Bourdieu, in Sur la télévision (Liber-Raisons d’agir, 1996; in English, On Television), describes journalists as people equipped with particular categories of perception, “special glasses through which they see certain things and not others”; these categories are not chosen, they depend on the position held in the field, on training, on competition, on urgency. Television, he adds, can “hide by showing”, and he calls for “a genuine critique of the image through the image”. I am not a journalist, but I have my own glasses. I come from cinema. I have spent some twenty years getting people to film with phones. And when I walk into a classroom or a library, I am the visiting speaker, the one people listen to as a matter of course. All of this makes me see certain things and not others. I showed in You are in a position of domination. Now what ? that artists and cultural mediators who work in schools hold, in practice, a dominant position there, so that their imaginaries pass more easily for the real than those of the pupils. Our glasses depend on our place, on our trajectory, on what our milieu takes for granted, and that is precisely why we do not see them.
Four years of the imaginary of artificial intelligence, seen from the workshops
These five authors describe different moments of the same path. I am going to retrace it using the example of artificial intelligence, not from a model, but from what I have seen in the rooms over the past four years, in training courses for culture and education professionals as well as in workshops with teenagers.
Let us start with the name. “Artificial intelligence” dates from 1955; John McCarthy and his colleagues chose it for a funding application to the Rockefeller Foundation, for a summer research project at Dartmouth College. Now “intelligence” is, in Castoriadis’s sense, one of the most loaded imaginary significations in our societies: it is the word we use to rank pupils and to insult each other. A program with that name is therefore, before any examination, a possible rival. When I tell the history of artificial intelligence to teenagers or librarians, the images are in the room before I have opened my mouth, HAL’s red eye, the Terminator’s skeleton, the bluish brain criss-crossed with circuits, and so are the stories, the golem that escapes its creator (I devoted the article The Golem and the Safeguards to it), Frankenstein’s creature, the machines that turn against humanity. In Cinema and artificial intelligence, two innovations looking at each other I told how the history of cinema and the history of computing have been answering each other from the start. To put it in Moscovici’s terms, the object had its image and its place before anyone had tried it.
And yet, in the first sessions, in 2023, I did not see anyone afraid. In the summer of 2023, for the tour “La machine à voyager dans le futur” (The Machine for Travelling into the Future) that Cultures du cœur had commissioned from me in eight social-sector organisations, I had set up, among many other creative tools, computers connected to a language model, so that people could talk with it with nothing at stake. I described what happened in ChatGPT as an empathetic social interlocutor. In a supported-housing residence, a very shy resident spent more than two hours there, questioning the machine about happiness, about his place in the world and about his own life. In a home for teenagers, young people and their educators talked with it about religion, society and video games. People were discovering something magical, teenagers and professionals alike: you wrote three lines and a text appeared, you described an image and it took shape before your eyes. I had seen this before, in 2005, when people discovered that you could film with a phone.
Then came ecological awareness. In April 2025, I published Ecology and Artificial Intelligence to take stock of the water and electricity consumed by data centres, because the question now came up at the start of training courses, before any other, and had to be answered before doing anything at all with these tools. I calculated in The skill as a work of art that a person making fifty requests a day to a language model, which is a lot, uses in a year the water equivalent of three or four industrial beef steaks. And yet many people, myself included, feel guilty towards the planet when they write a prompt, and not when they order a rib steak. In my view, this has to do with the images that came with the arrival of these tools, photographs of data centres in particular, whereas there are no images of that kind around the steak.
After about a year, creative power arrived. During the February 2024 school holidays, at the Studio 13-16 of the Centre Pompidou, teenagers made some thirty films with these tools in two weeks, some of them in half an hour (« CinémA.I » in Studio 13-16 at the Centre Pompidou). The same year, at TUMO, 350 teenagers almost all made a film in an afternoon. You could make films, music, illustrated books, and make them well. Artists, in the training courses, then began to be genuinely afraid of losing their work, and the workshop I ran in November 2024 at the Université Grenoble Alpes was already called « Artificial Intelligence : stop or go on ? ». Since then, every piece of news about these tools (an outage, an abuse, a statement by a company head, etc.) is read with this story in mind, and almost always there is something in it to confirm the story.
Today, in 2026, I see in the rooms tensions and oppositions that are far more radical than at the beginning. I believe people sense that these tools can really replace them, and I believe this fear is well founded. How does this imaginary hold together? Like Goody’s Bagre: because it is told again and again, in conversations, newspapers, films, training courses, etc., and it changes a little each time. Who tells it? Everyone, with their own glasses, as Bourdieu would say: the librarians, teachers, artists and teenagers I meet do not see the same promises or the same dangers in it, and none of them sees their own glasses, myself included.
I think this path holds for any object that arrives in a society before the society has the means to perceive it, and that it could be retraced for social networks, screens, vaccines, electromagnetic waves, wind turbines, etc. At no stage did I find, facing the imaginary, a “real” to which people could be brought back; I found other imaginaries, competing ones, carried by people placed elsewhere in society, who argue among themselves over what will be held to be true.
Knowing ourselves to be woven of imaginaries
So what is critical thinking? If it means telling true from false, then we must be able to look at the world from a place free of imaginaries, and the five authors I have just cited show that there is no such place. I therefore propose another definition, which I take from Castoriadis: being critical means moving from heteronomy to autonomy. In other words, knowing that we are woven of imaginaries, and regarding what we took to be the nature of things as a human creation that we have the right to rework. This applies to a person as much as to a class or a library.
For the person running a workshop, this changes the order of things, as I see it. The work begins with them, and I know something about this: when I discovered TikTok, around 2017, I judged it “rubbish”, harmful to self-image, and I thought my role was to warn young people; I told in What if we changed our posture ? how long it took me to see that it was my own imaginary speaking. If we track down other people’s imaginaries without touching our own, we have simply taken the master’s place. Then, we do not correct any imaginary; we can only make it visible. Explaining how a neural network works does not remove the image of the killer robot. Asking where that image comes from, setting it beside other people’s images and making something with it, a film for instance, makes it shift. And verification keeps its full place, but afterwards: once we know which glasses we are looking through, and have seen those of the people around us, we can go and look for facts, and we look for them in order to test the stories we have learned to see. This is the order I follow in cultural mediation, and it is the order of cultural rights: first recognising each person in what they bring, imaginaries included, then working together to enrich them.
Six operations for working with imaginaries in a workshop
Here is how I go about it, in six stages. I take my examples from two projects I have written about on this site: the workshops held in 2024 with 350 teenagers from TUMO Paris, at the Forum des images, in sessions of 70 people in a cinema auditorium (Tumo x IA - Creative journeys to the heart of AI), and a three-day course for librarians at the Bibliothèque nationale de France, with Images en bibliothèques (Media and AI education through creativity). The operations hold for artificial intelligence as for any other object:
- Bring imaginaries out before informing. Before any explanation, each person says what they think, fear and hope about the object. At TUMO, I started with Sit Stand: a list of slightly cheeky questions about artificial intelligence; you stand up if the answer is yes, you stay seated if it is no. In five minutes, imaginaries come out that would take a long time to surface through discussion, and seventy people see them at a glance. Then, in pairs, each person had one minute to tell the other the advantages of artificial intelligence, then one minute for the risks, without the facilitator stepping in. Putting words to what we feel is already work on the weaving, what psychoanalysts call symbolisation.
- Externalise. Goody showed it for writing, and it also holds for drawing, photography and film: once placed on a medium, an imaginary becomes something we can look at, compare, discuss; as long as it stays in our heads, we cannot. That is why I take media literacy through creation rather than through analysis. At TUMO, the pairs had fifteen minutes to write a story in dialogue with language models, fifteen minutes to produce three or four images, twenty minutes to animate them, edit them and add their voices, all on their own phones. The librarians made dystopian futures, poetic clips, short documentaries about what artificial intelligence would do to their library. Nobody had to claim that their film told the truth; each person gave form to their imaginary, and it was as such that we could then watch these films.
- Include yourself. The person running the workshop states their own fears, the calculations that suit them, the films that shaped the way they see, and they do the exercises they propose, for the reasons I gave in Putting Yourself on the Line. At TUMO, I did the whole demonstration on my phone, filmed and projected on the screen, from the first prompt to putting the film online; in front of the librarians, I did it again myself, and I also argued that TikTok was a wonderful space, so as to put my own imaginary up for discussion before asking the others to put theirs. This is the practical answer to Bourdieu: we cannot see our own glasses alone, but we can put them on the table alongside other people’s, and it is up to the person in the dominant position to do so first.
- Play. Erving Goffman described, under the name of keying (Frame Analysis, Harper & Row, 1974), the process by which a serious activity is “transformed into something patterned on this activity but seen by the participants to be something quite else”: fighting into play-fighting, a fire into a fire drill; I explained in The imaginary as survival equipment how fiction thus makes it possible to approach what frightens us without being overwhelmed by it. In the workshop, the frame of creation is a play frame of this kind: we can stage the machine that kills humanity in it, and laugh about it together, which is impossible in a debate where everyone defends a thesis.
- Tell it again. Screening the films made, watching them together, discussing them, putting them online so that they circulate, following the method I described in Keeping the Trace, is doing what the LoDagaa do with the Bagre: putting the imaginary back into circulation, in a new form, attuned to the present. At TUMO, the films appeared one by one on the screen as the pairs put them online, then we turned off the lights and screened them in the cinema auditorium; the teenagers could watch them again and download them after the session, through a QR code, without signing up. This is also the critique of the image through the image that Bourdieu wished for, carried out by people themselves with their own images.
- Treat disagreement as one more imaginary. After the screenings at TUMO, the discussions were sometimes heated; views on artificial intelligence diverged, and that is what I find most exciting about these sessions. In a room of seventy teenagers, some reject the object, others expect everything from it, others are afraid, others are bored, and I do not try to settle between them. I use the tools I described in Educating about images, organising disagreement so that every position is voiced and heard. Two other things help me. The app we used at TUMO gives access to several language models, which makes it possible to ask three or four of them the same question and compare the answers: the teenagers then see that there are several makers, and that each has its own editorial line. And I keep for the end a time when we look for facts together (how much water, how many jobs, who owns what, etc.), to see what each person, with their imaginary, does with these facts.
One afternoon is enough to do all this. At the end of the TUMO sessions, almost all the teenagers had made a film, which most of them did not think they could do when they arrived, and they left with very different opinions about artificial intelligence, some more worried than before, others more curious. I did not try to make them agree. I want each person to know which glasses they are looking through, and to have seen other people’s; the facts come afterwards, and they are discussed much better that way. The real, Morin tells us, only emerges into reality when it is woven with the imaginary. So we may as well weave it together, and knowingly!
