The age of cognitive surrender

Are we letting AI do our thinking for us?

The term “cognitive surrender” has been quietly circulating since January, but it took an Ars Technica piece in early April to bring it into wider conversation. The phrase was coined by Wharton Business School marketing researchers Steven Shaw and Gideon Nave, and once you encounter it, it’s difficult to shake.

The concept is straightforward but unsettling: when people consult AI, they tend to accept whatever it tells them—even when it’s wrong.

What the research found

As explained here, Shaw and Nave tested 1,372 participants using an adapted version of the Cognitive Reflection Test (CRT), a classic tool designed to measure whether people rely on quick intuition or careful reasoning. The questions are deliberately tricky. Consider this example:

If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets?

The answer is 5 minutes — but your gut instinct probably said 100. Getting it right requires what psychologist Daniel Kahneman famously called “slow thinking”: deliberate, analytical reasoning, as opposed to the fast, instinctive kind.

The twist in Shaw and Nave’s study was that participants had access to an AI chatbot during the test—one that was sometimes intentionally wrong. The results were striking:

• Participants consulted the chatbot about half the time
• When it gave correct answers, they accepted them 93% of the time
• When it gave wrong answers, they still accepted them 80% of the time
• Those who used the AI reported 11.7% higher confidence than those who didn’t—despite performing worse

In other words, people weren’t just being lazy. They were confidently wrong.

A new kind of thinking—or the end of it?

Drawing on Kahneman’s framework of System 1 (fast, intuitive thinking) and System 2 (slow, analytical thinking), the researchers propose that AI represents an emerging System 3: an external cognitive layer that people increasingly outsource their reasoning to.

The authors describe it this way: people are incorporating AI outputs into their decision-making “with minimal friction or skepticism,” allowing an outside system to substitute for their own internal reasoning.

That said, Shaw and Nave don’t treat cognitive surrender as purely catastrophic. There are genuine benefits to having a powerful reasoning tool at your fingertips—faster decisions, reduced mental effort, and access to vast information. The danger lies in using it uncritically.

This isn’t entirely new

Before you despair for humanity, it’s worth noting that blind deference to perceived authority is hardly a modern invention.

The theologian Peter Berger used the phrase “cognitive surrender” back in the 1990s, in a religious context—referring to the act of surrendering one’s doubts to faith. An example is what happens in a classic American sit-com like Home Improvement, where the dynamic is in a more comic form: every week, Tim “The Toolman” Taylor would take his problems to his wise neighbor Wilson, absorb some ancient piece of wisdom, and then repeat it so mangled that it was clear he’d processed none of it. Tim wasn’t reflecting—he was outsourcing.

The difference, of course, is scale. Wilson could only reach Tim. AI can reach everyone, all at once.

A note of caution

Before treating these findings as settled science, it’s worth keeping the replication crisis in mind. Psychological research over the past decade has repeatedly shown that experimental results—especially those involving human behavior—often fail to hold up when tested again by independent researchers. That doesn’t mean Shaw and Nave’s findings are wrong, but it does mean they deserve scrutiny rather than the same uncritical acceptance the study’s participants gave to their AI chatbot.

There’s something almost poetic about that irony.

The takeaway

AI is an extraordinary tool. But tools don’t think for you, and when we let them, the confidence we feel may be the most dangerous part. The question isn’t whether to use AI, but whether we’re using it with our eyes open.

Otherwise, we’re all just Tim Taylor, confidently mangling Wilson’s advice and wondering why the fence fell down.

What few people recognize is that cognitive delegation did not begin with artificial intelligence. The process started much earlier, the moment we stopped memorizing information and began searching for it online instead. AI, however, has accelerated and deepened this trend in a way that is qualitative as much as quantitative: we no longer delegate only theoretical knowledge but also practical reasoning and the capacity for judgment.

Those who grew up in the 1980s remember what it was like to navigate the world without a constantly available information network—making do, making mistakes, learning. That generation knows the difference between cognitive effort and technological convenience because they lived through both. Generations born in the internet age, and even more so those raised alongside AI, risk never having developed that mental muscle in the first place. The danger is not just dependency, but the inability to recognize it: not knowing what it means to figure things out on your own, because you never truly had to.

The real risk is a standardization of thought—minds shaped not by experience and independent reasoning, but by the patterns and outputs of an algorithm. Preventing this starts with education. If schools do not return to placing critical thinking, independent reasoning, and tolerance for uncertainty at their core, we risk producing generations that are technically connected but cognitively passive—not thinkers, but consumers of other people’s thoughts.