We hear two discourses about artificial intelligence: it supposedly saves us time, or it supposedly makes us lazy. Both speak of intellectual effort as a single block. I posit that some cognitive efforts transform us and others no longer teach us anything, and that it is now up to us to choose where to put ours.
Three million years of tools
We are completely bound up with our techniques: tools for eating, drinking and healing, work techniques, social techniques, etc. Since when? Since before we even existed! In 2015, the team of archaeologist Sonia Harmand published in Nature the discovery, at Lomekwi in Kenya, of knapped stone tools about 3.3 million years old, older than the earliest known members of the genus Homo. Bernard Stiegler already argued, in the first volume of Technics and Time, published in 1994, that humans invented themselves with their tools as much as they invented them. And I wrote about André Leroi-Gourhan, in Being a subject with technologies, or being their object, that human thought took shape with the hand and the tool. Technical objects are not external to us, since we are the ones who make them, and it is by making them that we became human. Our humanity is therefore at stake in technology, and in the way we use it.
So what changes with artificial intelligence? We had already entrusted tools with our physical strength (the ox, the mill, the steam engine, etc.) and with part of our memory (writing, the book, the hard drive). On 30 November 2022, OpenAI put ChatGPT online, and millions of people began having it write emails, summarize reports, translate, program, etc. I myself began using these tools in 2023 in my cultural and therapeutic mediation work, as I recount in Artificial Intelligence, a Tool for Emancipation in Mediation. For the first time, part of intellectual effort itself is entrusted to a machine, and on this scale. I posit that this is a major anthropological change.
Should we rejoice or worry? Anne Alombert, who has devoted several books to the thought of Bernard Stiegler, published De la bêtise artificielle (Allia) in 2025, in which she speaks of the “proletarianization” of our expressions: by entrusting writing to language models, we would lose the ability to write. Others see nothing in it but “time saved”. In my view, both positions lack a distinction between the efforts themselves.
I have proposed seeing artificial intelligence as a form of cognitive air conditioning, an aid that puts us back in a state to think and act, provided we choose its technology, the way it is built and its place. I have also presented, in Not Letting Go of the Controller, the document published in July 2026 by the TUMO centre, for which authorship and connection become scarce when capability becomes abundant. But how can we know what we can entrust to an aid, and what we had better keep for ourselves?
The effort to spare oneself effort
The Spanish philosopher José Ortega y Gasset gave a summer course in Santander in 1933, later published as Meditación de la técnica. In it, he defines technology as the effort human beings make to spare themselves effort. He is struck by this paradox, then asks the question that follows from it: where does the effort thus spared go? That is exactly my question! And I posit that the spared effort does not disappear: we put it somewhere else.
In 1958, Hannah Arendt published The Human Condition. In the prologue, she evokes the automation that will soon empty the factories, and she writes: “It is a society of laborers which is about to be liberated from the fetters of labor.” She adds that this society no longer knows of the higher activities for the sake of which this freedom would deserve to be won.
People often tell me that these tools save time. It looks like time saved, but I believe it is above all effort saved, and that is why I speak of effort.
Effort that enriches and effort that wears us out
I therefore propose to distinguish two kinds of cognitive effort. I call effort that enriches the effort that transforms us: in making it, we learn, we understand something, we become a little different. And I call effort that wears us out the effort that demands attention, memory and vigilance, but no longer changes us, because we already know how to do it.
I see it in a specific case. This autumn I built, for my website, a module that automatically imports my articles into SPIP. Putting an article online used to require a sequence of always identical operations: copying the French and English texts into the right fields, choosing the section, adding a dozen keywords, linking the article to others, etc. I know these operations by heart, and I learn absolutely nothing by repeating them. That is effort that wears us out. Writing the article, on the other hand, transforms me every time, and that is effort that enriches.
Is this a new distinction? Not quite. Researchers in education have words for it. The Australian psychologist John Sweller has been studying what he calls “cognitive load” since 1988. He showed that part of the effort asked of a pupil serves no purpose for learning, for example when the text of a geometry exercise is printed far from the diagram and one has to keep searching for which angle is meant. He calls this “extraneous” load. With Jeroen van Merriënboer and Fred Paas, in a 1998 article, he distinguishes it from “germane” load, the load that builds knowledge. My import module does nothing other than remove an extraneous load!
Robert Bjork, a psychologist at the University of California, Los Angeles, coined the expression “desirable difficulties” in 1994. Testing oneself on a lesson takes more effort than rereading it, and yet it is what makes us learn; rereading one’s notes only gives the impression of having learned. Bjork thus shows that some efforts are worth keeping precisely because they are costly.
I connect this distinction with the way I define work in Work and Mastery (2): a process of transformation of people through doing. Effort that enriches is work in this sense. Effort that wears us out no longer is, even if it tires us.
So what should we make of the proletarianization objection? I believe it is right, but for only one of the two kinds of effort. In June 2025, Nataliya Kosmyna and her team at the MIT Media Lab put online a study, still a preprint, in which 54 people had to write essays, some with ChatGPT, others with a search engine, others with nothing. Those who wrote with ChatGPT showed the lowest brain activity, and they had trouble quoting what they had just written. Writing an essay is precisely an effort that enriches, and entrusting it to a machine deprives the writer of it. Copying keywords into a form trains no one.
A new freedom
Until now, necessity usually chose for us. We made an effort because there was no other way to obtain the result. We had to copy, sort, format and check, and cognitive effort, whether it enriched us or not, was part of the price to pay. School, moreover, taught us to value intellectual effort in itself. Today, a machine can do a good part of these tasks. We must therefore ask ourselves which effort we want to make ourselves.
This is what I call choosing one’s efforts: consciously distributing one’s energy between these two kinds of effort, keeping those that enrich us and entrusting to tools those that wear us out. We are not used to it, and it is in this sense that I speak of an anthropological change.
Choosing one’s efforts also makes it possible to do new things. In 2026 I am working on the Culture 2027 platform, which cross-references the texts of political programmes ahead of the 2027 French presidential election. It is artificial intelligence that writes the comparisons, the questions and the dialectical confrontations, from a very large and very precise corpus and from objectives we have defined in great detail. It is a work of cognitive techniques, which makes it possible to discover common ground, oppositions, blind spots, etc. There are two of us running it. Without generative artificial intelligence, it would perhaps have taken thirty people to do it in the same time, and it would probably even have been impossible: we would have had to hire, organize and establish protocols. The people who visit the platform receive food for thought and places to contribute, and the work it required is less and less their concern. The effort of production thus becomes more democratic: we can offer more to others, and bring our intention closer to its realization.
The same task, two natures
We must nevertheless be subtle. A task wears us out or enriches us depending on our relationship with it, and on who decides about it.
Housework is a good example. Hannah Arendt places it, in The Human Condition, on the side of what she calls labor, the activity that starts again every day to sustain life and leaves nothing behind it. If I do the housework at home in place of others, who do not make the effort themselves, I become in a way the others’ “skivvy”, and this effort amounts to alienation. But if I do it for myself, it can become a form of meditation: by cleaning and tidying, I reclaim my own space, and I have a material grip on reality. I observed something of this kind with the storing of equipment in my workshops, which I describe in The invisible art of material organization. The same goes for what we choose to eat, for the way we choose to move, etc.
Housework can thus belong either to what André Gorz called “heteronomous” work, the work we do because society requires it, or to “autonomous” activities, those we choose for their own sake (Critique of Economic Reason, published in French as Métamorphoses du travail, Galilée, 1988). And Harry Braverman, who had worked in metalworking and shipbuilding before becoming a publisher, recounted in Labor and Monopoly Capital, published in 1974, how Taylorism had separated the conception of work from its execution, reserving it for the planning departments. When it is management that decides, on behalf of others, what will be entrusted to machines, this is no longer choosing one’s efforts.
Two people close to me showed me the other side. A friend deeply committed to ecology, a vegan, whom I would have thought refused artificial intelligence, uses it for his administrative paperwork, which he finds unbearable, and he builds his life better thanks to it. A woman who works in an institution, deeply committed to feminism, uses it to put her texts into gender-inclusive writing, according to her own rules and her own decisions. For her, applying these rules word by word is effort that wears her out, because she already knows what she wants. For someone discovering inclusive writing, the same effort might be enriching. I would therefore say that effort that wears us out is a relationship between a person and a task, at a given moment in that person’s life.
Depending on one’s tools, and knowing it
Choosing one’s efforts means knowing which tools one depends on. What, then, distinguishes dependence on a tool from being connected to reality through it? I believe they are two sides of the same coin. Computing renders us enormous services, and we depend on it: without it, many things can no longer be done. The difference lies, in my view, in our awareness of this dependence.
It is like the body. Very often we have no awareness of our muscles, of the way we hold ourselves, of what moves when we walk. In the method developed by Moshé Feldenkrais (1904-1984), whose words I quoted in Body Awareness, we make very slow and very small movements, without unnecessary effort, and suddenly we become aware that certain muscles exist, that we can distinguish parts of our body we thought were all of a piece. By distinguishing these parts, we come to inhabit them, and we gain the power to act on ourselves.
It is the same with tools. When artificial intelligence enters our lives very forcefully, when everyone uses it and it becomes naturalized, we risk no longer perceiving it. This is neither good nor bad in itself. But knowing that these are tools, telling one from another, knowing why we choose this one rather than that one, is already being their subject and not their object, in the sense I give to the subject’s presence. Gilbert Simondon wrote in 1958, in the introduction to On the Mode of Existence of Technical Objects, that the misunderstanding of the machine was the most powerful cause of alienation in his time. I believe so too. This awareness also makes it possible to break the reflex of entrusting everything to a single tool, which I have called mononormative delegation.
My import module, for example, I built with the help of Claude, an artificial intelligence model to which I subscribe. I depend on it, and I know it. I also know that I use techniques made by others, and that in turn I make some for others. It is in this sense that I speak of a connection to reality rather than of dependence: our human reality is made of techniques, and everything depends on our awareness of them.
Choosing one’s efforts, in practice
Before entrusting an effort to a machine, I now ask myself a few questions, which I suggest to you:
- Does this effort still teach me anything? For putting my articles online, the answer had been no for a long time. For writing the articles, it is yes.
- Am I the one who decides to entrust it to a machine, or is someone else deciding for me?
- Am I making this effort for myself, or in place of others? Housework does not have the same meaning in the two cases.
- What will I do with the effort thus freed? This is the question Ortega y Gasset asked in 1933, and that Hannah Arendt took up again in 1958.
- Do I know which tool I depend on, and why I chose it rather than another?
I do not always have the answers. But simply asking myself these questions is already changing the way I work!
