For centuries we have been creating and using tools to save ourselves effort. The wheel lightened our load, the calculator took care of complicated operations, and GPS saved us from memorizing impossible maps. Technology has always expanded our capabilities.
However, the arrival of generative artificial intelligence is something else entirely, because it’s no longer just about delegating certain cognitive tasks; we’re beginning to outsource the very process of thinking. And that already has a name: cognitive surrender.
When the answer comes too quickly and easily
We like to think of ourselves as rational beings who enjoy reflecting, analyzing arguments, and evaluating evidence. However, reality is far from that image. Thinking consumes energy, demands constant attention, and, on top of that, forces us to tolerate uncertainty.
Our brains, for purely evolutionary reasons, are always looking for shortcuts to consume less energy. That’s why it’s so tempting to use artificial intelligence and get an immediate, well-written, seemingly logical answer presented as if it were the absolute truth.
The problem is that this convenience comes at a price.
When we receive an answer, our brain skips all the work it would normally do to arrive at it. We stop comparing, questioning, or exploring alternatives. Thus, little by little, we begin to accept conclusions instead of constructing them.
The new oracle
Although it may be difficult to admit, the truth is that, as humanity, we have spent thousands of years relying on figures who convey (or appear to convey) knowledge and security. This applies to priests, but also to doctors, teachers, political leaders, and the “experts” who used to appear on television and now flood social media. They all share something in common: they project authority.
Artificial intelligence has inherited (or perhaps more accurately, has been designed to simulate) that ability. When we read a well-structured and fluently written response, our brain automatically interprets that there is a certain level of competence.
Ultimately, we don’t evaluate an answer solely on its content; we also consider its form, albeit often unconsciously. This means we’re more likely to accept a well-constructed, fluent, and coherent sentence than an awkward one, even if both are correct (or incorrect).
Artificial intelligence is programmed to appear expert even when it’s wrong (something that happens quite often and has been dubbed “AI hallucinations“) because it doesn’t hesitate or waver. It doesn’t correct itself mid-sentence or say “I’m not sure,” as a person would. And in a world saturated with uncertainty, having that certainty readily available is extraordinarily tempting.
A study that should make us reflect.
A recent experiment at the University of Pennsylvania analyzed how we interact with responses generated by artificial intelligence. Its results were, to say the least, worrying.
When the AI provided incorrect answers, over 70% of participants still accepted them. But the most worrying aspect wasn’t the error itself, but the trust they placed in it. In other words, they weren’t just wrong, they were convinced they were right.
In three experiments involving more than 1,300 participants, researchers observed that they consulted AI to solve many of the tasks, and its accuracy was virtually identical to that of the artificial intelligence itself. Therefore, when the AI was right, they were right, but when the machine was wrong, they were wrong too.
And the worst part is that, in many cases, they didn’t even try to verify the answer. They simply assumed it to be true, like someone copying an answer from a classmate without bothering to think about it for a second.
What exactly is cognitive surrender?
Cognitive surrender is a new term coined by these researchers to refer to the phenomenon that occurs when we stop critically evaluating information and begin to automatically accept the answers provided by an external source due to the trust it inspires in us.
The result is a progressive externalization of reasoning. Obviously, the problem isn’t that we seek information externally (something we’ve always done to test our hypotheses or obtain more data), but rather that we begin to delegate the thought process. We all consult external sources constantly: books, articles, and teachers are a natural part of learning.
Cognitive surrender occurs when we outsource our ability to structure an argument, weigh alternatives, detect inconsistencies, or draw conclusions—in other words, everything that defines critical thinking.
Where does the difference lie? When we ask AI to help us write an email or correct grammatical or syntactic errors in an article, we’re using it as a tool. However, when we ask it what we should think about a political issue, a moral dilemma, or an important life decision and simply accept its answer, we’re delegating our thinking.
The invisible risk of outsourcing thinking
We’re not using AI to think better (as we’d like to convince ourselves), but to avoid having to think . And not because we don’t know how, but because it’s faster and more convenient.
The point is that critical thinking works like a muscle, so if we stop using it, it weakens. Imagine someone who uses a calculator for absolutely everything. After a few years, they’ll still get correct results, but they’ll probably have lost much of their mental dexterity for calculation.
Something similar happens with reasoning. When we let artificial intelligence constantly construct arguments, synthesize information, or generate explanations, we run the risk of practicing those skills less and less.
Obviously, it’s not something that happens overnight. It’s not like you wake up one morning unable to think; it’s a gradual process in which we make small concessions, until one day we struggle to follow a logical thread, connect the dots, and draw our own conclusions.
How to use AI without handing over the keys to your brain?
At this point, it’s quite difficult to reject artificial intelligence, but we still have time to change how we use it. For example, instead of asking it, “What do you think about this?”, we could ask it to give us arguments for and against, so we can draw our own conclusions. And instead of accepting the first answer, we should reflect on it and compare the data.
Obviously, this requires more cognitive effort, but it’s a small price to pay for continuing to think freely. Ultimately, the real risk isn’t that machines will become more intelligent, but that we will cognitively capitulate.
Source:
Shaw, S. D. & Nave, G. (2026) Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender. The Wharton School Research Paper; 10.31234.
Leave a Reply