Generative AI is supposed to make life easier. It drafts emails, summarizes documents, and even generates creative content, helping you offload some of that dreaded cognitive effort. But according to a new study from Carnegie Mellon University and Microsoft Research, it may also be making you dumber in the process.
The study, based on a survey of 319 knowledge workers, found that AI is substantially reducing people’s critical abilities. Even worse, it makes them feel more confident in their decisions, despite putting in less cognitive effort. The long-term result will be an overreliance on AI, weakened reasoning skills, and a growing inability to detect errors.
Generative AI can already produce all sorts of content. But it’s not precise. It can “hallucinate” facts, generate misleading information, or reinforce biases. Yet, many users treat AI-generated answers as trustworthy — even when they shouldn’t.
The study analyzed nearly 1,000 real-world examples of how professionals use AI in their daily work. The data was self-reported.
“In our survey, participants were asked to share three real examples of their GenAI tool use at work. To increase the variety of examples collected, participants were asked to think of three different examples, one for each task type: Creation, Information, and Advice. Then, participants were asked to share an example of each task type in detail,” the researchers explain in the study.
The researchers found that when users trusted AI, they were less likely to think critically about its responses. In other words, the more faith you have in AI, the less you actually think about the output.
AI is reshaping human cognition
We should emphasize that these are knowledge workers — people who, in various capacities, use their knowledge and expertise to create value. These are people who engage in active problem-solving — brainstorming, structuring ideas, and drawing conclusions. Now, their primary role has shifted from thinking to verifying. Instead of generating original ideas, they check AI’s output for obvious errors.
At first glance, this might seem like a good thing. AI does the heavy lifting, and humans simply verify the work. But in practice, if people skip the most intellectually demanding part of a task, they sacrifice their own critical thinking. Furthermore, if an AI-generated response seems correct, users are likely to accept it without deeper scrutiny. The result? A solution that looks good enough but erodes human reasoning over time.
“Used improperly, technologies can and do result in the deterioration of cognitive faculties that ought to be preserved,” the researchers note. “A key irony of automation is that by mechanising routine tasks and leaving exception-handling to the human user, you deprive the user of the routine opportunities to practice their judgement and strengthen their cognitive musculature, leaving them atrophied and unprepared when the exceptions do arise.”
This isn’t the first example of something like this happening. From GPS navigation reducing our sense of direction to spell checkers eroding our writing ability, there’s a long list of technologies that carry a hidden cognitive cost. But we’ve never had anything as powerful and pervasive as generative AI.
Overreliance on computing technology is not a novel phenomenon; however, GenAI tools can exacerbate the associated risks. Indeed, such reliance may be tolerable for low-stakes tasks, like grammar checking, but it can lead to significant negative outcomes in high-stakes contexts, like drafting legal documents,” the researchers note.
Don’t get dumbed down when using AI
The study did offer some hope. People were more likely to engage in critical thinking when they were more confident in themselves, had lower confidence in AI, or when the task had high stakes (ie when writing a legal or medical report).
But the more important thing is to start questioning AI’s input. Treat them like a smart intern — but an intern nonetheless. AI is not an expert. It’s not a fact-checker. Verify everything and don’t just check the response for blatant errors, cross-reference it with other sources. AI should be used to enhance and not replace your own reasoning.
But the study ends on a rather disturbing note. As AI gets more advanced, the temptation to rely on it will only increase. Sure, most people wouldn’t use it to draft a law today, but what about a few years from now? The way things are going, the risk that human intelligence could stagnate — or even decline — is very real.
If you use AI regularly, consider how it’s affecting your thinking. Using AI for idea generation, proofreading, or summarization is fine — as long as you don’t blindly trust the results. But if you find yourself becoming more passive, less engaged, or less likely to question information, it may be time to rethink your approach.
AI won’t make you dumb on its own. But you relationship with AI determines whether it helps you think — or stops you from thinking altogether.
It won’t be easy.
You can read the paper in its entirety here.
