Newsletters
The new peer-reviewed study shows that adults who often use generative AI tools for daily work tasks report feeling less confident in their own reasoning and less ownership of their ideas. This adds new evidence to the ongoing debate about how AI changes the way we think.
Published on April 16 in Technology, Mind, and Behavior, the study surveyed 1,923 working adults in the US and Canada aged 25 to 57.
Participants completed 10 simulated workplace tasks, such as drafting a salary negotiation plan and interpreting unclear data, using chatbots like ChatGPT, Claude, and Gemini.
Overall, 58% of participants said that "AI did most of the thinking," especially during tasks that involved planning or several steps.
People who accepted AI-generated answers without much change felt less confident in their own reasoning. Those who edited, questioned, or rejected the AI's output felt more confident and had a stronger sense of authorship.
Sarah Baldeo, the study's author and a PhD candidate in AI and neuroscience at Middlesex University in London, said that how people use AI shapes its impact on thinking, not the tools themselves.
"Generative AI can lead to cognitive decline or cognitive evolution. It depends on your interaction style," Baldeo told TIME magazine.
"When we look at brain activity based on how people use the tool, we see increases or decreases. It really doesn't have to do with the tool itself."
The research found that greater reliance on AI was strongly associated with lower self-reported confidence. In contrast, people who changed or rejected AI output tended to feel more confident.
Senior workers gave more prompts and overrode AI suggestions more often than entry-level workers. They also reported higher confidence, suggesting that experience may help protect against excessive reliance on AI.
These findings are similar to another, not-yet-peer-reviewed paper from MIT and Carnegie Mellon University. That study found that when people lost access to AI in the middle of a task, their reasoning ability and willingness to solve problems independently dropped quickly.
Together, the two studies suggest a common pattern: letting AI do all the thinking is linked to weaker independent reasoning, but using AI as a supplement may help maintain strong reasoning skills.
Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania who studies AI and work, said this pattern matches how people react to any tool that saves effort.
"If the AI solves a problem for you, you don't think, and you don't learn," Mollick told TIME. "But if you use AI like a tutor and let it challenge people, you get better results."
Baldeo suggests a method she calls "cognitive scaffolding." This means building a basic understanding of a task before letting AI help, and actively questioning the AI instead of just accepting its first answer.
She recommends going back and forth with the AI two or three times, asking for more details or disagreeing if needed.
She also suggests starting your questions with a note to avoid AI flattery: "Respond based on third-party verifiable information. Do not attempt to flatter me or build an emotional bond."
The study authors emphasized that their findings are descriptive and do not prove cause and effect.
However, they believe that as generative AI becomes part of daily work, the key factor will be whether people continue to think for themselves while using it.
Set as preferred source
Set as preferred source
1 min read
1 min read
