A 2026 interview study of 45 daily ChatGPT users found something the researcher didn’t set out looking for: participants describing their own minds in the past tense. Not “ChatGPT helps me think.” More like — I used to be able to do this. One student put it plainly: “I used to take notes and actively review materials, but now I just ask ChatGPT whenever I need something. The problem is, I don’t retain much. I feel like my memory has become weaker because I rely on ChatGPT instead of trying to recall things on my own.” Immanuel Kant read this exact pattern in 1784, in an essay that never mentions a machine. He just called it by its real name: immaturity. Self-incurred. Chosen, not imposed. And, he insisted, remarkably comfortable to live inside.

The Quick Verdict

Question	What we found
Who ran the study?	Zeineb Farhat, assistant professor at ISC Paris
Where and when?	Applied Cognitive Psychology (Wiley), published online September 2, 2026, DOI 10.1002/acp.70205
How big was it?	45-60 minute semi-structured interviews with 45 daily ChatGPT users, ages 18-25, in France
What was measured?	Thematic analysis of interview transcripts — no biometric or test-score data, self-reported experience only
What themes emerged?	Cognitive offloading, eroded autonomy and self-trust, memory decline through reduced active recall, social isolation from reduced peer interaction
What did Kant already argue?	In An Answer to the Question: What Is Enlightenment? (1784), immaturity (Unmündigkeit) is self-incurred through laziness and cowardice — not a lack of intelligence, a lack of nerve

The Quick Version

A 2026 study interviewing 45 daily ChatGPT users found a recurring pattern: participants outsource memory and decisions to the chatbot, then report weaker recall, thinner independent thought, and a growing preference for the AI over people. Kant described the identical mechanism in 1784 — immaturity as something people choose because thinking for yourself is effortful and paying someone (or something) else to do it is so easy. He named the laziness. The study just measured what it does to you.

What the 2026 Study Actually Found

Farhat wasn’t trying to prove ChatGPT is bad for you. She was asking a narrower question: what happens, psychologically and socially, to people who use it every day? So she sat down with 45 of them — daily users, 18 to 25, in France — for 45-to-60-minute conversations about how the tool had worked its way into their studying, their decisions, their friendships. Then she ran a thematic analysis on the transcripts, the standard qualitative method for finding what actually repeats across dozens of separate conversations rather than what one person happened to say.

Four themes kept surfacing. Cognitive offloading — using the chatbot to hold information so participants didn’t have to. Eroded autonomy — a diminishing trust in their own judgment on decisions ChatGPT hadn’t even weighed in on. Memory decline — the specific, repeated experience of noticing recall getting worse, not hypothetically but as something they’d caught happening to them. And social isolation — a documented drift toward preferring the chatbot’s company over a friend’s, not because the chatbot understands them better, but because it’s easier.

None of this is Farhat claiming ChatGPT rewires the brain. It’s qualitative — 45 interviews, one country, one narrow age band, all self-report. What it captures well is something a brain scan can’t: how offloading feels from the inside, while it’s happening, described by the people doing it. And what it describes is a group of young adults who know, while they’re telling you, that something has changed. “I feel like my memory has become weaker,” the student said — not “I forgot something once.” A pattern, noticed by the person living inside it.

Kant Already Diagnosed This, Two Centuries Before the Chatbot

Kant’s essay An Answer to the Question: What Is Enlightenment? opens with a definition that still lands: “Enlightenment is man’s release from his self-incurred tutelage.” Immaturity, Unmündigkeit, isn’t stupidity. It’s the inability to use your own understanding without someone else’s guidance — and Kant is explicit that when this immaturity persists in an adult with a functioning mind, the fault isn’t a lack of capacity. It’s “lack of resolve and courage.” That’s the whole argument in one clause. You could think for yourself. You’ve arranged not to.

He doesn’t stop at diagnosis. He describes exactly how the arrangement gets made, and the passage reads less like eighteenth-century moral philosophy and more like a description of a subscription: “If I have a book that understands for me, a spiritual advisor who has a conscience for me, a doctor who decides upon a regimen for me, and so forth, I need not trouble myself at all. I need not think, if only I can pay; others will readily undertake the irksome business for me.” Book, pastor, physician — three professions, three different domains of judgment, all doing the same job: absorbing the cognitive labor a person would otherwise have to do themselves. Kant’s list is period-specific. The mechanism it describes is not. Swap in “chatbot” and the sentence doesn’t need editing, just updating.

What Kant adds that the 2026 study can’t, because it isn’t a philosophical text, is the motive. He’s not describing people who lack access to better tools of thought. He’s describing people who have them, and who choose the easier path anyway, because — his word — it’s comfortable. “It is so easy to be immature,” he writes. Laziness and cowardice, not incapacity, are what keep “so great a proportion of men… in lifelong immaturity” long after nature and circumstance would let them think for themselves. The 2026 participants aren’t stupider than people were a decade ago. They’ve made the same trade Kant described his contemporaries making — the effort of thinking, exchanged for the comfort of not having to.

What Did Kant Actually Mean by Self-Incurred Immaturity?

1. It’s not about intelligence. Immaturity, for Kant, describes an inability to use your own understanding without guidance — a failure of nerve, not of capacity. The book, the pastor, and the doctor aren’t smarter than you. They’re just doing the thinking you’ve agreed to stop doing.
2. It’s chosen, which is what makes it your responsibility. “Self-incurred” (selbstverschuldet) is the load-bearing word in the whole essay. If the cause were external — poverty, oppression, a genuinely unequal starting point — Kant wouldn’t call it a fault. He calls it self-incurred specifically because “the cause of it is not lack of understanding, but lack of resolve and courage.”
3. It persists because it’s comfortable, not because it’s forced. Kant is blunt that guardians don’t need to trap anyone. People “gladly remain” dependent because independent thought is effortful and dependence is not. Nobody has to coerce you into the easy version.
4. The way out is a habit, not an event. Kant’s motto — Sapere aude, dare to know — isn’t a one-time decision. It’s a standing instruction to keep using your own understanding, case by case, rather than defaulting to whoever’s willing to think for you this time.

This Site Has Covered AI and Isolation Before — Here’s the Difference

An earlier post here on Levinas and another on Buber’s I-Thou both asked why AI companionship deepens loneliness rather than curing it. Their answer was relational: comfort-optimized AI can’t produce the friction — the resistance of another mind that might disagree, disappoint, or surprise you — that real encounter requires. Those posts are about what AI does to your capacity for connection.

This one is about something upstream of connection: what happens to your capacity to think, before you’ve even gotten to another person. The 2026 study’s social isolation theme overlaps with what Levinas and Buber already diagnosed — participants preferring the chatbot’s company to a peer’s. But its cognitive-offloading and memory-decline themes are a different problem entirely, one this site hadn’t named yet. Kant’s framework is built for exactly that gap. He’s not asking whether a machine can meet you as a full subject. He’s asking whether you’ve quietly handed over the work of judging, deciding, and remembering — and whether you noticed doing it.

Socrates ran a related argument from the opposite direction — that real understanding has to be drawn out of you through your own answers, not deposited into you by someone else’s. Kant would agree with the diagnosis and add the missing piece: the reason people accept answers instead of doing the drawing-out themselves usually isn’t confusion. It’s that drawing out your own answer is slower and harder than being handed one.

Where This Doesn’t Hold Up

None of this makes Kant a cognitive scientist, and the honest version of this piece says so directly.

Farhat’s study is qualitative — 45 people, one country, one age band, self-report only. There’s no control group, no pre/post memory testing, no way to rule out that heavy ChatGPT users were already more prone to skip active recall before the chatbot existed. “I feel like my memory has become weaker” is a genuine, repeated experience worth taking seriously. It is not the same claim as “ChatGPT causes measurable memory decline,” and the study doesn’t claim the stronger version either. It’s describing a lived pattern, not establishing a mechanism.

Kant’s claim is a different kind of statement altogether. He wasn’t describing a psychological process with a measurable before-and-after. He was making a moral argument — that people who could reason for themselves and don’t are responsible for the gap, whatever the cause of the gap turns out to be. That argument doesn’t need cognitive offloading to exist as a phenomenon to be true; it would hold in a world with no AI at all, applied to pastors and physicians instead. The 2026 data doesn’t prove Kant’s account of moral responsibility. It’s compatible with the shape of his diagnosis in a way that’s genuinely striking — which is a smaller, more honest claim than saying an eighteenth-century essay predicted a psychology study.

It’s also worth saying plainly what this isn’t: an argument that ChatGPT or any AI tool is inherently corrosive. Farhat’s own findings include participants using it constructively — as a scaffold for creativity and reflection rather than a replacement for thinking. Kant’s framework doesn’t condemn the tool either. A book doesn’t make you immature; using a book so you never have to reason again does. The line he draws is about the choice to stop trying, not about the existence of something that makes stopping easier.

A Practice Worth Trying

Neither Farhat nor Kant wrote this as a five-step plan, but the shared shape of what they found points somewhere specific.

1. Try before you ask. Before opening the chat window, spend sixty seconds attempting the answer, the plan, or the recollection yourself. Kant’s test isn’t whether you eventually get help — it’s whether you used your own understanding first. Prosoche, the Stoic practice of sustained attention, is built on the same premise: the muscle you don’t use starts to go quiet.
2. Notice which decisions you’ve quietly outsourced. Not just facts — judgment calls. If you catch yourself asking ChatGPT what you should do about something only you have the context to weigh, that’s the autonomy theme from the 2026 study showing up in real time, not a hypothetical.
3. Keep something that requires unaided recall. Daily journaling as a philosophical practice works partly because it forces retrieval without a lookup tool. The 2026 participants who lost the note-taking habit lost the recall along with it — the two were never separate.
4. Let the tool answer the easy part and keep the hard part. Kant’s target was never assistance — it was substitution. Using ChatGPT to draft a first pass, check a fact, or handle a mechanical task isn’t the immaturity he’s describing. Handing over the reasoning you’re capable of doing yourself is.

The Choosing Was the Point

Farhat wasn’t testing Kant. She was asking a narrower, more clinical question — what happens to memory, autonomy, and social connection in people who use ChatGPT every day — and she found something sharper than “AI makes you lazy”: a specific, self-reported pattern of offloaded memory, thinned-out independent judgment, and a growing preference for the chatbot’s company, described by the people it was happening to, while it was happening. Kant wasn’t running interviews. He was working out, from a single insight about human nature, why enlightenment is so hard to sustain even once it’s available — and he landed on laziness and cowardice, not ignorance, as the actual obstacle. Neither needed the other to be right. It’s just hard to ignore that a 1784 essay about pastors and physicians and a 2026 study about a chatbot are describing the same unglamorous choice: it is so easy to let something else do the thinking. Kant’s whole point was that ease was never a good enough reason.

---

This post discusses a 2026 qualitative psychology study on AI use alongside a classical philosophical text. It isn’t a diagnosis of anyone’s relationship with AI tools. If reliance on any technology is affecting your work, memory, or relationships in ways that concern you, a therapist or counselor can help in ways a philosophy essay can’t.