A short spell of AI assistance may be enough to change how people approach difficult tasks once that help disappears. New findings suggest that the effect is not limited to faster answers or higher accuracy while the tool is available, but can also reshape how long someone keeps trying when facing a problem alone.
The study examined whether AI support changes not just performance, but also persistence. Researchers found that users who relied on AI tended to do better while the tool was active, yet they became less resilient when asked to continue without it.
AI support can improve results quickly
The research, titled AI Assistance Reduces Persistence and Hurts Independent Performance, was carried out by a team from the United States and the United Kingdom. It focused on reasoning-heavy tasks, including mathematics, reading comprehension, and information processing.
When AI help was available, participants completed tasks more quickly and with better accuracy than those working without assistance. The tool made the work feel easier, and the gains were visible while the support remained in place.
The pattern changed once the AI was removed in the middle of the task. Participants who had relied on the assistant showed a measurable drop in performance when they had to return to independent work.
A change appeared after just 10 minutes
One of the most striking findings was how quickly the shift emerged. The researchers said that around 10 minutes of AI use in problem-solving was enough to trigger weaker performance once participants were left on their own again.
That detail matters because the decline was not tied to long-term dependence. Even a brief period of assistance seemed sufficient to make independent reasoning more fragile when the tool was no longer available.
The study also suggests that the effect did not depend on a single kind of task. Across several controlled experiments, the same general pattern kept appearing.
Persistence also weakened
The researchers looked beyond correctness and speed. They also measured persistence, or the willingness to keep working through a difficult question instead of giving up early.
When AI assistance was taken away, some participants became quicker to stop trying. In certain cases, they were also more likely to skip questions rather than attempt an answer.
Rachit Dubey, one of the study’s authors, said the issue was not only that people made more mistakes without AI. He noted that users also appeared less inclined to think through a solution on their own.
That distinction is important because it points to a broader risk. If a person relies too heavily on chatbot-based help, the problem may not be limited to weaker answers at a given moment. The habit itself may make independent thinking feel less necessary.
Repeated tests showed a similar pattern
One experiment involved around 350 participants who were asked to solve fraction-based math problems. Some had access to an AI assistant that could provide instant answers, while others completed the task without help.
The AI group performed better during the assisted phase, but their performance dropped clearly once the assistance was withdrawn. A second experiment with nearly 670 participants produced the same general outcome.
A third test used a different type of challenge: reading comprehension. Even though the task format changed, the decline still appeared after AI support was removed.
Those repeated results strengthened the researchers’ concern that the impact is not limited to one setting. The effect showed up across different tasks and in multiple controlled trials.
What the researchers worry about
The team warned that frequent use of AI could gradually weaken deeper thinking habits. They also suggested that the tendency to stay focused and keep trying without help may slowly erode.
Dubey cautioned that overreliance on chatbots may make independent thinking seem like unnecessary effort. If that pattern grows, the possible impact may reach beyond lower performance and into confidence and creativity.
The study has not yet undergone peer review, so the findings still need further examination. Even so, the early results raise a clear question about the trade-off between convenience and mental resilience as AI becomes a more common part of study and work.
Source: www.indiatoday.in