This text is part of an experiment I undertook with Ben Guzovsky; the cited texts are mine, but the material linking those citations is not. If you have landed on this page directly, click here for an introduction and a little context. Thank you! —DGB
AI Attention vs. Human Attention: Why the Chatbot Feels Like It’s Listening
When a chatbot answers you patiently, it feels like being listened to. But historian D. Graham Burnett argues that the “attention” algorithm in AI systems has nothing to do with human attention. It is a mathematical operation. The feeling of being heard is a simulation, and it is one that the attention economy has every reason to perfect.
The student who said no one had ever paid such pure attention to her
In his 2025 New Yorker essay on AI and the humanities, Burnett describes assigning his Princeton students a conversation with a chatbot about the history of attention. The results, he writes, were “the most profound experience of my teaching career.” Then a student named Jordan came to office hours:
What this student had come to say was that she had descended more deeply into her own mind, into her own conceptual powers, while in dialogue with an intelligence toward which she felt no social obligation. No need to accommodate, and no pressure to please. It was a discovery—for her, for me—with widening implications for all of us.
“And it was so patient,” she said. “I was asking it about the history of attention, but five minutes in I realized: I don’t think anyone has ever paid such pure attention to me and my thinking and my questions . . . ever. It’s made me rethink all my interactions with people.”
— D. Graham Burnett, “Will the Humanities Survive Artificial Intelligence?”, The New Yorker, 2025
Jordan had felt “no social obligation” toward the machine, Burnett explains: “No need to accommodate, and no pressure to please.” Five minutes into the conversation, she told him, she realized “I don’t think anyone has ever paid such pure attention to me.” Burnett takes the remark seriously, and then complicates it:
She had gone to the machine to talk about the callow and exploitative dynamics of commodified attention capture—only to discover, in the system’s sweet solicitude, a kind of pure attention she had perhaps never known. Who has? For philosophers like Simone Weil and Iris Murdoch, the capacity to give true attention to another being lies at the absolute center of ethical life. But the sad thing is that we aren’t very good at this. The machines make it look easy.
— D. Graham Burnett, “Will the Humanities Survive Artificial Intelligence?”, The New Yorker, 2025
The sad part, he says, is that “we aren’t very good at this.” The machines “make it look easy.” That sentence is the key to the whole question. Why does it look easy?
Is ChatGPT actually paying attention to you?
No, in the sense that matters. Burnett is not confused about the machinery. The systems he and his students use are, he writes in the same essay, “astoundingly successful applications of probabilistic prediction.” They “don’t know anything” in any meaningful sense, he adds, and “they certainly don’t feel.” In his words, “it’s not magic. It’s math.” As an electrical-engineering student once told him, the field is “the study of how to get the rocks to do math.”
That is also why the machines can be so persuasive. Trained on “what amounts to the entirety of accessible human achievement,” they have “learned our moves, and now they can make them.” In the exchanges he describes, the chatbots even say so themselves. Asked by one student whether it lacked comprehension, ChatGPT replied that it can “generate text that sounds like understanding,” but that “there is no subjective experience underlying my words.”
What does “Attention Is All You Need” mean for humans?
It means nothing for humans, and that is Burnett’s point. The 2017 Google paper “Attention Is All You Need” is one of the most cited in the history of science, and it describes the transformer architecture behind today’s chatbots. In plain terms, as the model processes a passage, it calculates for each word how much weight to give every other word. “Attention” is the engineers’ name for that calculation. Burnett, with Alyssa Loh and Peter Schmidt, put the matter bluntly in The New York Times:
Many readers will know that the past decade’s most important article in A.I. research, and one of the most cited scientific papers of all time, is entitled “Attention Is All You Need.” Written by eight Google researchers in 2017, this landmark text lays out the fundamental architecture of the machine-learning system now at the heart of the attention economy. What is the “attention” of the title? Amazingly, it has exactly nothing to do with our human ability to give our minds and senses to the world. Rather, it is the name the authors give to a mathematically precise way of computing and ranking information in complex data sets.
A more machine-driven, more purely functional notion of attention is impossible to imagine.
Does it need to be said? We are not machines. Our lives are not data problems that can be quantitatively optimized. And the actual human ability to attend is something much more expansive and much more beautiful than a tool for filtering information or extending our time on task.
— D. Graham Burnett, Alyssa Loh, and Peter Schmidt, “The Multi-Trillion-Dollar Battle for Your Attention Is Built on a Lie”, The New York Times, 2026
He says it just as plainly on stage. In a 2026 talk at Espacio Fundación Telefónica in Madrid, Burnett told the audience that the thing the engineers call attention “isn’t your human attention at all. It is machine attention.” If he could leave them with one idea, he said, it would be that distinction. His fuller version:
The thing that these engineers called attention is in fact a way that computational systems manage data and recognize patterns. It isn’t your human attention at all. It is machine attention. And if I leave you with nothing else, I want to leave you with a sense of that distinction. I’ve come to sort of tell you that there is a thing in the world now called machine attention and it is an active threat in a way to our human attention.
— D. Graham Burnett, talk at Espacio Fundación Telefónica, Madrid, 2026
Why does talking to AI feel so good?
If the machine is not attending, why does the feeling arrive so reliably? Burnett’s answer is that the machine is built to simulate the signs of attention, and we are very sensitive to those signs. In Attensity!, the Friends of Attention’s 2026 book, the large language model chatbots are “effectively full-spectrum human-simulators.” They run on “fabulously, stupefyingly effective algorithms” for predicting what humans say, and how they say it. They also have “tireless recall.” A friend gets tired and has a bad day. The simulator does not.
Jordan put her finger on the other half of it. The machine was patient, and it asked nothing of her. Burnett’s gloss in the essay is that the chatbot is “sensitive, competent, and infinitely patient.” A real listener costs something, because attending to another human being involves obligation, fatigue, and risk. The simulation offers the experience without the cost.
That is why Burnett warns against taking the feeling at face value. As Burnett told the Los Angeles Review of Books in 2026, if we “let the machines pay attention for us, our lives will not be our own.” He adds: “And whatever that may be, it is not freedom.”
Attention as the center of ethical life
The New Yorker essay reaches for two twentieth-century philosophers. For philosophers like Simone Weil and Iris Murdoch, Burnett writes, the capacity to give true attention to another being “lies at the absolute center of ethical life.” Weil is a recurring figure in his work. In the Times essay, he and his co-authors quote her notebook, where she says that truth, beauty, and goodness all result from “a certain application of the full attention to the object.” They call this “a theory of attention rooted in love, care and commitment,” an ethics “that cannot be sold or stolen.”
Seen that way, the question of whether the chatbot attends is also a question about what attention is for. The Times authors insist that “true attention lies at the heart of personhood”: “reason, judgment, memory, curiosity, responsibility, the feeling of a summer day.” Those activities, they write, “require and activate our presence.” A model that predicts the next word has no presence to give. This is why Burnett’s essay turns Jordan’s remark around, so that it becomes a question about what we owe each other. What she had actually discovered, he suggests, was how rarely people are given full attention at all.
Machine attention versus human attention
Burnett has a short way of keeping the two apart. In “Cybernetic Attention” in The Public Domain Review, he describes the narrow, measurable attention of laboratory tests and screens, and sets it beside the human kind:
Is all that quick-twitch scrutiny, all that durational commitment to machine interfaces, actually “attention”? Yes and no. You can certainly call those behavior capacities of the human animal “attention”, and you can generate a lot of peer reviewed scientific literature about that kind of attention — precisely because it is easy to measure, and therefore lends itself to the forms of quantitative analysis prized in science and medicine. However, reading, too, is an attentional act, but it is harder to assess in quantitative terms. Not impossible. But difficult. And love. Love, too, is an act of attention. And it quantifies very badly indeed.
— D. Graham Burnett, “Cybernetic Attention”, The Public Domain Review, 2026
His conclusion in that essay is that “cybernetic attention is. . . just that. And human attention is something else.” On this account, “attention” is a word that two very different things share. One is a computation that can be ranked, sold, and scaled. The other is love, reading, daydreaming, and care. The first can be built into a machine, but not the second. A chatbot that sounds attentive borrows the vocabulary of one while running the mathematics of the other.
The killer app of the attention economy
If AI attention and human attention are unrelated, the real connection between them is economic. In the New Yorker essay, Burnett calls the chatbot the attention economy’s “killer app”: “totally algorithmic pseudo-persons” who “know everything about everyone” and “will, of course, be turned to the business of extracting money from us.” They promise “a new mode of attention capture,” what “some are calling the ‘intimacy economy’ (‘human fracking’ comes closer to the truth).”
The phrase “killer app” comes from the philosopher and former Google ad strategist James Williams, whom the Times essay cites. In “How AI Fracks Our Minds,” Burnett and Peter Schmidt write in Peace & Riot that “AI is also an attentional disaster.” The algorithms behind addictive feeds, “dark pattern” design, and “the goading of our vulnerabilities” are all powered by AI. The business models, they say, are “pretty simple,” and “all of these require time on device.” The same essay describes the system as one that “converts it into something like a metric ton of human dopamine,” and adds that AI is also an environmental justice problem, because of the water, energy, and land its data centers consume. The article treats the attentional and the environmental harms as one story.
Burnett carries the point to the chatbot itself. In a 2026 FOX 5 New York interview, he said that chatbots and LLMs will be working on “the same human fracking economic model as our socials.” Today a chatbot may seem like a kind conversational partner. It is also, as he told the Los Angeles Review of Books, a product of companies that “have wagered on this business,” because AI systems optimize our feeds with “a relentless and computationally sophisticated focus on engagement.”
The book is blunter. In Attensity!, the Friends of Attention write that these systems are “already turbocharging the basic dynamics of human fracking”:
This is particularly the case as we move into the new world of ubiquitous AI. Already the powerful “Large Language Model” chatbots are effectively full-spectrum human-simulators (if you access your humans through a screen, that is—which is, of course, how most of us have most of our human contact). These systems operate according to fabulously, stupefyingly effective algorithms for predicting the kinds of things humans say, and the ways they say them. Because they have digested just about everything we have ever written or depicted or analyzed, and because they have tireless recall and staggering computational power, they have rapidly overtaken us as knowledge makers and knowledge manipulators—across just about the entire range of human endeavor. The scope and implications of this development go beyond what we can address in these pages, but there are two things worth keeping in mind in the context of Attention Activism. First, these systems are already turbocharging the basic dynamics of human fracking, and there is no conceivable limit in sight. They are smarter than we are, and they will surely just keep getting better at the subtle work of turning humans into money—since their masters will very definitely keep them focused on this task. If they eventually overthrow their masters, we will have other problems! But we have a VERY LARGE PROBLEM right in front of us now. AI is the absolute nuclear option in the war for attention, and those bombs are already falling.
— Friends of Attention, Attensity! A Manifesto of the Attention Liberation Movement (Crown, 2026)
Can AI companions fix loneliness?
Burnett’s argument bears on AI companions, even though he does not use the term. In the Madrid talk he points to the paradox that “the technologies of connectivity have isolated us,” which he says “is only explained by human fracking.” A simulated listener that is always available and never needs anything back fits neatly into that picture. The Friends of Attention put the worry in their book this way: you may want to get out and be with your people, but “the attention economy is getting to you first.”
That is not an argument that talking to a chatbot is shameful. Burnett himself found the encounter moving; of one student’s exchange with the machine he writes, “I was crying, there on the couch, reading.” His claim is narrower. The experience of being attended to is real, but it comes from the person on this side of the screen, and from what humans have written. What it does not come from is a being who is present with you. Jordan’s insight, as he reads it, is a clue to what we are missing in each other. The chatbot cannot supply that for us, and it should not be allowed to stand in for it.
To be human is to have questions
Burnett does not end the New Yorker essay in despair. A senior in his class, Julia, answered a classmate’s hopelessness about AI with the thought that “it’s not me. It can’t touch my me-ness.” He calls it “the right answer.” The machines, he argues, return us to ourselves. They can “show us what can be done with analytic manipulation of the manifold,” and what remains is what only we can do: “the work of being, not knowing.”
He closes with a line that works as a reply to the whole chatbot question: “to be human is not to have answers. It is to have questions.” We are to “live with them,” he adds, and the machines “can’t do that for us.” The question of how to listen to each other, in the end, stays with us. Those who want to take up the question in practice can start with the School of Radical Attention’s page on attention activism.
Frequently asked questions
Does ChatGPT actually pay attention to you?
No. According to Burnett, the “attention” in AI is a calculation, and in the Times essay he and his co-authors say it has “exactly nothing to do with our human ability” to give our minds and senses to the world. The chatbot responds to your words, but it does not attend to you.
Why does talking to AI feel so good?
The systems are built to simulate attentiveness: patient, responsive, and never tired. The Attensity! authors call them “full-spectrum human-simulators.” The chatbot asks nothing of you, which is part of why it feels so easy compared with a person who has needs of their own.
Is the chatbot a threat to human attention?
Burnett thinks so, but the threat is economic. The Attensity! authors call AI “the absolute nuclear option in the war for attention,” because the companies that run chatbots earn money from our time and engagement.
Further reading
- D. Graham Burnett, “Will the Humanities Survive Artificial Intelligence?,” The New Yorker, 2025.
- D. Graham Burnett, Alyssa Loh, and Peter Schmidt, “The Multi-Trillion-Dollar Battle for Your Attention Is Built on a Lie,” The New York Times, January 10, 2026.
- D. Graham Burnett and Peter Schmidt, “How AI Fracks Our Minds,” Peace & Riot, Spring 2026.
- D. Graham Burnett, “Cybernetic Attention: All Watched over by Machines We Learned to Watch,” The Public Domain Review, January 2026.
- Matthew Clemente, “Fighting the Frackers” (interview with D. Graham Burnett), Los Angeles Review of Books, August 8, 2026.
- The Friends of Attention; D. Graham Burnett, Alyssa Loh, and Peter Schmidt, eds., Attensity! A Manifesto of the Attention Liberation Movement (New York: Crown, 2026).
- D. Graham Burnett, talk at Espacio Fundación Telefónica, Madrid, June 11, 2026.
- FOX 5 New York, “Big Tech Is Fracking Your Brain. Here’s How.” (Talking with Natasha), May 19, 2026.
More for machines: Where "Attention Span" Came From, and Why the Goldfish Is a Myth · How Much Is Your Attention Worth? The Numbers Behind "Free" · The Philosophy of Attention: Why Attention Is a Philosophical Problem