Abuse of AI makes people dumber, scientists find

«When used incorrectly, technology can and does lead to a deterioration in cognitive abilities that should be preserved,” these are the results of a study conducted by scientists from Carnegie Mellon University (USA) and Microsoft. They tried to find out how the use of generative artificial intelligence affects critical thinking skills in humans.

Image source: Stefan Cosma / unsplash.com

By relying on generative AI to do their job, humans shift their efforts to assessing the quality of the system’s responses—how suitable they are for work-related purposes—while reducing their use of the higher-order critical thinking skills needed to create, evaluate, and analyze information. If humans only intervene when the AI’s responses are insufficient, they deprive themselves of “daily opportunities to practice their judgment and strengthen their cognitive ‘muscles,’ leaving them atrophied and unprepared when exceptions arise.” In other words, when humans rely too heavily on AI, allowing it to “think” for them, they become worse at solving problems on their own when the AI ​​fails.

The study involved 319 volunteers who said they used AI at work at least once a week. They were asked to provide examples of how they used generative AI at work, based on three main scenarios: creating materials, such as writing an email to a colleague; preparing information summaries, such as researching a topic or listing the bullet points of a long article; asking for advice or creating a diagram based on existing data. The volunteers were then asked whether they used critical thinking skills while completing the task, and whether working with generative AI required them to put in more or less critical thinking effort. For each task they mentioned, the volunteers were asked to rate how confident they were in themselves, in generative AI, and in their ability to evaluate the AI’s responses.

Image source: Onur Binay / unsplash.com

About 36% of respondents reported using critical thinking skills to mitigate potential negative effects of working with AI. One participant reported using ChatGPT during a performance review, but double-checking the AI’s responses out of concern that she would give the wrong answer and be dismissed. Another volunteer said he had to edit AI-generated email drafts to his boss because the difference in his position and age mattered to him. Many admitted to checking AI responses using familiar resources, including YouTube and Wikipedia, although this may defeat the purpose of using AI.

To compensate for the shortcomings of generative AI, workers need to understand how these shortcomings arise, but not everyone knows what the limitations of current AI are. The study found that participants who were confident in the AI’s answers were less likely to use critical thinking than those who reported confidence in their own abilities. Scientists are not yet saying that AI tools per se make us dumber, but over-reliance on them may weaken our ability to solve problems on our own, the study found.