I propose a concept, cognitive air conditioning: artificial intelligence as an aid to thought and action, just as coolness puts a body back in working order. An aid is neither a triumph to celebrate nor an abdication to deplore; it is judged by its technology, its construction and its place in the shared environment. I draw on Ivan Illich, Bruno Latour, Olivier Hamant and the TUMO school, and from them I derive four questions anyone can ask in order to build their own cognitive air conditioner.
Too hot to think
When I am too hot, I cannot do anything. My body suffers, it wears down, and my whole attention shrinks to looking for shade. Coolness does not think in my place, it gives me back my means. Many of us know, at work, the cognitive equivalent of that heat: shrunken teams drowning in emails, summaries impossible to produce, reporting obligations that devour the time for doing. In that state, we no longer think, we hold on. Artificial intelligence takes over part of that mental expenditure, the first drafts, the summaries, the rewordings, just as the air conditioner takes over the heat, and I call cognitive air conditioning this use of the machine as an aid that puts people back in a condition to think and to act. I put this concept forward because the public debate remains stuck between those who celebrate the tool without looking at its costs, and those who deplore it and hold any aid to be an abdication of the mind. Yet an aid is neither good nor bad in itself. Penicillin is an aid, glasses are one, so is the air conditioner. The useful questions lie elsewhere: what technology is this aid made of, how is it built and by whom, and what place does it occupy in the shared environment, the one where everyone lives and thinks.
The school in Gando
The formula attributed to Lavoisier, nothing is lost, nothing is created, everything is transformed, is true at the scale of the universe, and it says nothing about a situation until the system has been studied. To get from one point to another, I can take a bicycle or a semi-trailer; the journey will be the same, the expenditure beyond comparison. The compression air conditioner, the one bolted to façades, does indeed reject outdoors the heat it removes, plus the heat of its own compressor, but that is true of that particular machine, not of coolness in general. Step off a concrete slab in the sun onto the shaded earth under trees, twenty metres away, and you enter another thermal world, more than twenty degrees of difference at ground level, several degrees in the air, without consuming or rejecting anything. The Burkinabè architect Diébédo Francis Kéré built his first building on this, the primary school in Gando, completed in 2001, which I visited in 2015. Walls of compressed earth bricks, made and laid by the village itself, buffer the heat of the day; a large tin roof, raised above the ceiling, shields it from the sun and lets the wind carry the heat away; classes can be held there when it is forty degrees outside, without a single appliance. For the secondary school, incoming air travels through buried pipes, because deep down the ground keeps a roughly constant temperature. This work earned Kéré the Aga Khan Award in 2004, then in 2022 the Pritzker Prize, awarded for the first time to an African architect. The school in Gando provides the model: an aid designed with the materials of the place, built by the people who use it, attuned to the wind and the earth, and which enriches its environment, in coolness, in skills and in pride, instead of dumping its heat on the neighbours. The technology of an aid is never given in advance, it is designed.
The convivial tool
Ivan Illich, too often reduced to his critical side, spent an entire book defending this idea. In Tools for Conviviality (1973), he does not condemn the tool, he draws a dividing line between two ways for the tool to exist among humans:
« Such a society, in which modern technologies serve politically interrelated individuals rather than managers, I will call “convivial”. »
Ivan Illich, « Tools for Conviviality », 1973.
On the other side of the line, Illich describes the radical monopoly, the tool that no longer merely drives out its competitors but the non-technical answers to the need it covers; the car did not just replace the tram, cities were built around it where walking became impossible. The need that artificial intelligence covers, knowing, formulating, deciding, has its non-technical answers, asking a colleague, searching, letting a question ripen, and they deserve to be maintained the way one maintains the trees of a city that also has air conditioners. Illich’s line serves me as a compass: cognitive air conditioning remains convivial as long as the tool serves people integrated in their community, as long as everyone can use it or do without it, and it tips into radical monopoly when it becomes the only answer available. Keeping the tool on the right side of that line is work, not fate.
The end of means
Bruno Latour brings the concept its second support. In “Morality and Technology: The End of the Means” (Réseaux, no. 100, 2000), he argues that techniques are never mere means in the service of ends that would pre-exist them: they are mediations, they transform our ends along with our means, they make us want things we did not want before them. That is why an aid is not judged once and for all; we come back to it again and again. The rebound effect, which the economist William Stanley Jevons describes as early as 1865 and which Olivier Hamant recounts through the story of refrigerators (Antidote au culte de la performance. La robustesse du vivant, Gallimard, “Tracts”, 2023), is the exact illustration of this: efficiency gains lower the cost of use, use spreads, and the means has displaced the ends. One can read this effect as a condemnation, and that would be the slope of “things were better before”, which I distrust, because it never says when that before was, nor for whom it was better. The Suez Canal gives the measure of what this nostalgia forgets. From 1859 onwards it was dug by hand under the corvée, a system of forced labour that mobilised four hundred thousand Egyptian fellahs and killed several tens of thousands of them, the figure of one hundred and twenty thousand deaths put forward later by Nasser remaining disputed for lack of archives; after the abolition of the corvée in 1864, steam dredgers took over, and nobody misses the shovels. Penicillin, another aid, changed the fate of millions of patients. The advance of knowledge and techniques is not an evil, and the rebound effect is not a moral fatality: all it tells us is that the gains of an aid do not find their destination by themselves, and that this destination is ours to decide. That is Latour’s lesson: since the means displaces the ends, we must keep sorting out what we hold dear, regularly, together.
From proteins to archives
Cognition become abundant makes it possible to do things we could not do, and this must be said as plainly as the criticisms are stated. The AlphaFold program has predicted the structure of more than two hundred million proteins, placed at the free disposal of researchers the world over, a task that decades of crystallography would not have sufficed to accomplish, and the 2024 Nobel Prize in chemistry rewarded work built on these machines, while the physics prize honoured the discoveries that made them possible. Heritage conservators I work with tell me that artificial intelligence will change the way history is written, because it makes immense bodies of archives explorable that no researcher’s lifetime would suffice to read, and because a cognition different from ours will find connections there that we would not have made; in the article Sa bêtise, c’est sa force I set out why this difference of cognition is the most precious thing these machines have to offer us. And the same aid also serves to churn out papers: the scientific paper mills, now equipped with generative models, account for 1 to 3% of publications according to the low estimates discussed at the CNRS in late 2025, and world output, growing fast, overwhelms the capacities of peer review. The air conditioner question arises here in the same terms: does this improvement enrich the shared environment, like the protein structures offered to all, or does it enrich some while impoverishing the environment where the others do their searching, like the papers manufactured to be counted? The same tool, depending on its place, does one or the other.
Coolness given back to connection
I tested the concept before formulating it. In 2025 and 2026, with ALCA Nouvelle-Aquitaine, I led a collaborative assessment of screen literacy education across the region, which I recounted in the article Collaborative assessment of screen literacy education in Nouvelle-Aquitaine. Artificial intelligence took over heavy cognitive tasks, the first drafts of the summaries of interviews and workshops, which we then revised and validated. The gain did not show in the documents, it showed on the people: less fatigue, more availability, and all the human energy carried over to what cannot be delegated, the quality of the bond, the listening, the trust that makes contributions possible. Without this aid, those summaries would have had to be paid for in salaries, in travel, in evenings, and we would have arrived at the meetings tired, busy, less open; the contributions would have been fewer and poorer. Ordinarily, this kind of assessment produces tables, how many actions, how many sessions, figures useful to the funding authorities and which change nothing in the work of those in the field, something I analysed in the article Media Literacy : Should We Judge or Guide ?; here, we enriched one another, and we extended the process with a day of collective intelligence. This is also the conclusion reached by the authors of the TUMO school, whose strategy paper I presented in the article The School of What Cannot Be Delegated: when intellectual capability becomes abundant, what becomes scarce is authorship and connection, and a well-placed aid serves to cultivate them, not to replace them.
« When capability becomes a commodity, school does not become less necessary. It grows in importance, but becomes something else. »
TUMO Center for Creative Technologies, « Making, thinking and learning when capability is no longer scarce », 2026.
The body, first user of aids
Living things have used aids for as long as they have existed, and they regulate their use better than we do. Olivier Hamant describes fever as performance under special dispensation: our cruising temperature, thirty-seven degrees, leaves a margin in reserve, and the organism draws on it when a pathogen demands it, for a time, because a permanent fever would destroy the very proteins it defends. And illness acquaints everyone with a second gesture. When we are ill, even without fever, the body withdraws from us the energy of relating, speaking, acting, digesting cost too much, and it puts us to bed; it stops spending itself outward in order to devote all its energy to repairing itself. Medicine assists this work, antibiotics, vaccines, operations sometimes tip the balance between recovery and death, but they are aids in the strict sense: in the end, it is always the body that heals itself, and the best doctor in the world is the one who helps it do so best. I also know this through a practice. I have been fasting for eight years, following the so-called hygienist method, on water alone; the body comes to a standstill as in illness, no more energy for digestion, no more energy for communication, and I can do nothing, even when I would like to, the body decides. It then puts all its energy into inner maintenance, lives on its reserves and recycles its damaged components, what biologists call autophagy, whose mechanisms Yoshinori Ohsumi, Nobel Prize in medicine 2016, elucidated; fasting is found, moreover, in almost every religion, and our language keeps the trace of it, the French déjeuner coming from the Latin disjejunare, to break the fast. From this I take the rule of use that the body applies to its own aids: they assist a work they do not do in our place, they switch on when the situation demands it, and they step aside when the work is internal. Cognitive air conditioning is best set the same way, mobilised to cope, switched off to let things ripen and to protect that time when nothing is owed, which the authors at TUMO defend.
Four questions to build your air conditioner
This concept is only of interest if it translates into action. For every personal use and every deployment in an organisation, I propose that we ask ourselves four questions, the ones we would put to an air conditioner:
- What is its technology? A model proportionate to the task, as one takes the bicycle to fetch the bread; in the article The arte povera of artificial intelligence I defended the creative use that can be made of modest models, and there are small, specialised models that can be installed on one’s own machine.
- How is it built, and by whom? The school in Gando was built by the village, which kept the skills; the ownership of machines and models is a political question, which I put forward in the article The Land We Cultivate.
- What is its place in the shared environment? Olivier Hamant’s common health test serves here as a gauge: the use must feed the health of individuals, social health and the health of ecosystems, without exception; a use that relieves one person while burdening the environment where the others do their searching fails the test.
- What do we do with the coolness? The time and energy that the aid frees up do not find their destination by themselves; it is decided, and the best one I know is that of the Nouvelle-Aquitaine assessment: the bond, the listening, the quality of the contributions, everything that cannot be delegated.
An air conditioner is not something one endures, it is something one designs, and this designing is within our reach: it begins at the next meeting, when someone asks what the time the machine has just given back will be used for.
