Cognitive surrender begins as convenience and becomes reliance: We stop using AI to support human judgment and start letting it stand in.Kiichiro Sato/The Associated Press
Steven Shaw is a Canadian academic, assistant professor at King’s College London, and the lead author of "Thinking – Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender."
Cognitive surrender (noun): adopting AI outputs without scrutiny; deferring thinking to AI.
After my research on the risks of outsourcing thought to AI went viral, a good friend asked me: “Did ‘cognitive surrender’ come from you, or from AI?”
I hesitated as I tried to trace the origins of my thoughts. The boundary between human and AI had blurred.
I remember the day: sitting at my work-from-home computer, surrounded by vinyl records and even a Y2K Furby. As I was listening to Boards of Canada (who are in fact Scottish, but Stompin’ Tom was on deck), it came to me: That’s the idea, that’s the term. But as memories go, the scene fades at the edges. I questioned myself … I almost always have ChatGPT open while I work, while I think. And so, I wondered: Could it have come from the machine?
This kind of source uncertainty reflects a larger cultural shift. In the comments of social-media posts and in casual conversation, we have become obsessed with asking whether a piece of text, an image, or a video “is AI,” as if we are still standing at the border, capable of patrolling it – even though the question itself indicates that AI can successfully mimic human content. To me, the more interesting question is: What happens when we accept artificiality into our minds?
Cognitive surrender begins as convenience and becomes reliance: We stop using AI to support human judgment and start letting it stand in. It may not be intentional or even noticed, but this reliance shifts agency from the user to the system. In doing so, we outsource not only a task, but the ambiguity, effort, doubt and intellectual friction from which judgment is formed.
And the rise of cognitive surrender is well under way. Students use AI to generate interpretations before they have wrestled with a text. Professionals draft memos before they have clarified what they believe. The danger is not merely that AI can be wrong. It’s that it can be right often enough, and useful often enough, to make deference feel rational, even inevitable.
Opinion: Universities should stop treating AI as a cheating problem
The central issue becomes not whether AI is “good” or “bad,” or whether using it is cheating, or whether it sometimes outperforms us. Of course it does. The seduction of AI comes from its ease and competence; the option to surrender is always available.
But capability is not wisdom. Fluency is not judgment. And intelligence, however defined, is not benevolence.
Every meaningful form of learning contains friction. We struggle with a sentence before it becomes our own. We sit in uncertainty, percolating, before an idea forms. In making mistakes, we build a mind capable of good judgment. Such difficulties are not ornamental, but essential to the development of critical thinking. When AI removes or pre-empts cognitive processes, it does more than save labour; it can deskill thought.
In my favourite childhood movie, Willy Wonka & the Chocolate Factory, scientists attempt to use a giant computer to predict the location of the next golden ticket. When asked, the machine whirrs and replies: “I won’t tell, that would be cheating!” The machine is not only intelligent enough to know the answer; it is principled enough to withhold it. It understands that having the answer is not the same as deserving to give it, and that the value of knowledge depends on how it is used.
Many want to imagine AI as an oracle, an adviser, a companion, a neutral supermind that will improve our lives if only we let it. But unlike the computer in Willy Wonka, the systems entering daily life are not wise machines with moral instincts. They are prediction engines built by companies, optimized for efficiency, engagement and scale. They may produce brilliant outputs. They may save time. But they do not care whether dependence is making you shallower, duller, or less free.
Editorial Board: Keeping at bay the dangers of AI
We have always offloaded elements of cognition. Calculators did not destroy mathematics. Search engines did not destroy memory. We routinely defer to teachers, experts and textbooks. But AI is different: it does not merely provide answers, but simulates reasoning with unmatched fluency and access. It can produce answers that feel like thinking, enabling users to adopt unthought thoughts as their own, without noticing what has been quietly surrendered.
And cognitive surrender is more than just a psychological risk; it’s a political and cultural risk too. To hand over judgment to AI is, by extension, to hand it to the institutions that design, fine-tune and deploy these systems. Their assumptions, biases, incentives and blind spots do not stay inside the machine. As we adopt and follow AI outputs, they travel outward, into the choices we make, the beliefs we adopt, the reasons we repeat.
AI will no doubt become more capable and more embedded into our lives. The decisive question in the age of AI will not be whether a sentence, a diagnosis, a lesson plan, or a strategy memo was produced with AI. The decisive questions will be about whose values speak through the technology, whether outsourcing was deliberate, and what is lost to those who surrender.
Let me return to my work-from-home desk, with the Y2K Furby. After some time and reflection, I did remember whether it was me or the machine that came up with “cognitive surrender.” But, like the computer in Willy Wonka, I’ll say only this: I won’t tell – that would be cheating! This obsession with spotting AI is a useless distraction.
AI is ubiquitous. Without a clear understanding of how these technologies are affecting our psychology, cognitive surrender may lead not only to job loss and deskilling, but to the erosion of the deliberative thoughts that make judgment, creativity and autonomy possible.
The most important thing we may lose to artificial intelligence is not our jobs, or even our originality. It may be the habit of mind that makes either possible. To me, it’s hard to think that AI could ever create a song like Stompin’ Tom’s Bud the Spud.