Does AI Threaten Our Cognitive Skills?

The growing use of artificial intelligence (AI) to solve complex problems could have unexpected consequences on our cognitive skills. A recent study conducted by researchers from Carnegie Mellon University, the University of Oxford, the Massachusetts Institute of Technology, and the University of California, Los Angeles, highlights the potentially negative effects of AI on our ability to perform tasks without technological assistance.

Using generative AI to solve a complex math problem or draft an email may seem like a quick and easy solution. However, this reliance could leave our critical thinking, creativity, and reasoning skills dormant. The study reveals that participants who relied on AI to practice a task for just ten minutes had more difficulty completing that task without it, compared to those who had never used it. The researchers emphasize that AI chatbots can assist in tasks requiring intensive reasoning, but depending on these tools could hinder our ability to accomplish them independently.

There is a significant difference between having an AI tool that solves problems for you and using it to obtain hints and clarifications—or not using it at all. The study raises ongoing concerns about how the use of AI affects our ability to perform the same tasks without the machine's help. It is one thing to have a tool like ChatGPT or Claude that provides guidance or answers questions leading us to a solution, but it is another to entrust it with the entirety of our tasks.

Dependence and Diminished Performance

The study involved 1,200 American participants on the Prolific research platform, divided into three distinct experiments. The tests consisted of solving fraction problems and answering reading comprehension questions in a SAT-style format. Some individuals received AI assistance and could use it as they wished, but it was removed after ten minutes. The results showed that participants who used AI for help were more likely to abandon problems and performed worse than those who did not use it.

Negative effects of AI usage were observed only among those who depended on it to solve problems, and not among those who did not use it during the study. Similar results were found in a MIT study last year that focused on using AI to write essays.

The researchers found that when participants who had used AI could no longer access it, they did not feel as confident in their ability to solve problems. They wrote that people not only become less competent in tasks but also stop trying. If such effects accumulate over months and years of AI use, we could end up creating a generation of learners who have lost the willingness to struggle productively without technological support.

Grace Liu, a PhD student at Carnegie Mellon University and co-author of the study, clarified that only short-term effects were measured. She emphasizes the need for further research to understand the long-term impact of this reliance on AI. "How significant this effect is on a large scale and over the long term requires more research," Liu stated. While AI can be used differently depending on the use case, the study does not evaluate how these tools are employed.

Comparison with Other Tools

The study also raises the question of whether the use of AI is similar to other methods or shortcuts for problem-solving, such as using a calculator to solve a math problem. One distinction is that generative AI can be used for almost anything, such as personal decisions, editing, and research with follow-up questions. "The two phenomena certainly share similarities as they allow people to delegate cognitive tasks," Liu said. "We believe it is particularly important to study cognitive delegation to AI because it can be used ubiquitously across many reasoning tasks, whereas previous tools were task-specific."

Towards Thoughtful Use of AI

The researchers indicated that the results raise questions about the effects on our persistence and reasoning when we use AI in our daily lives. "We warn that if such effects accumulate with sustained use of AI, current AI systems—optimized only for short-term assistance—risk eroding the human capabilities they are meant to support," the researchers added. Liu recommends exercising caution when using AI tools. "Our results suggest that we should be more intentional about how and when AI assistance is used and deployed—especially in learning contexts," she said. "This is not a reason to avoid AI, but it is a reason to design and use these tools carefully."