Artificial intelligence chatbots have become a standard tool for quick answers, but new research shows that relying on AI for brief problem-solving tasks can quickly undermine human motivation. A study presented at the Conference on Language Modeling in San Francisco investigated how shifting mental work to an AI chatbot impacts performance when that help is suddenly taken away.
The findings indicate that just a few minutes of artificial assistance can significantly erode a person's perseverance and independent capability, raising broader questions about how cognitive offloading affects long-term learning.
Examining the Cost of Quick AI Answers
To understand the effects of AI on problem-solving, researchers recruited 1,222 participants for a series of randomized controlled experiments. The study looked at tasks lasting between 13 and 15 minutes, specifically focusing on arithmetic fraction problems and reading comprehension questions.
The first experiment assigned 354 participants into two groups: one with access to an AI chatbot and one without. Both groups tackled 12 fraction problems. While the AI-assisted group performed well initially, the researchers then removed the AI assistant without warning for a final set of three unassisted problems.
The results showed that participants who previously had access to AI experienced a sharp drop in performance. They skipped or failed more problems once the chatbot was gone, demonstrating a decline in task persistence compared to the control group.
Testing Reading Comprehension and Broader Reasoning
To see if the effect was limited to mathematics, the research team conducted additional trials. A second experiment with 667 participants incorporated a pretest phase and yielded similar results, confirming that direct AI solutions harm a user's drive to complete tasks independently.
A third experiment involving 201 participants examined reading comprehension, which relies on meaning-making skills and mental-model construction. When the AI was disabled after five comprehension questions, the performance of the AI-assisted group dropped again across the final three questions. According to study first author Grace Liu, a doctoral student in machine learning at Carnegie Mellon University, this suggests that reduced persistence is a general outcome of cognitive offloading rather than an artifact of math problems alone.
Why AI Assistance Reduces Human Persistence
The study suggests that people quickly grow accustomed to the immediacy of AI answers. Consequently, they begin to perceive any problem-solving method that takes longer as inefficient. However, cognitive scientists emphasize that "desirable difficulties"—the productive struggles involved in learning—are essential for building long-term skills and critical thinking abilities.
While calculators also allow people to offload computations, study authors note that AI differs because it allows users to offload reasoning across many different domains.
Designing AI Tutors for the Future
While researchers agree that more long-term studies are needed to track how small instances of cognitive offloading affect lasting competence, current findings suggest overreliance on AI erodes cognitive fitness.
To combat this, the researchers argue that future AI models and chatbots should be designed to act more like tutors. Instead of handing out direct solutions, systems should offer directions and instructions toward a solution.
"We need to work toward a future in which human abilities are, as much as possible, augmented rather than supplanted," said study co-author Brian Christian, a computer scientist and researcher at the Center for Human-Compatible AI at the University of California, Berkeley.
