A short stretch of AI use may be enough to change how people approach difficult work on their own. A study found that just 10 to 15 minutes with AI assistance can reduce persistence when users later have to solve problems without digital help.
That finding matters because chatbots are now widely used to answer questions, complete tasks, and deliver instant solutions. Researchers warn that the convenience of AI may make some users give up faster when the tool is suddenly no longer available.
The study, titled AI Assistance Reduces Persistence and Hurts Independent Performance, was conducted by researchers from Carnegie Mellon University, MIT, the University of Oxford, and UCLA. Across three large-scale experiments, the team worked with 1,222 participants and tested them through math and reading comprehension tasks on an online platform.
Some participants were allowed to use an AI assistant that could solve questions directly. Others had to complete the work without any AI support. In the early stage, the group with AI looked stronger because they answered faster and recorded more correct responses while help remained available.
The pattern changed once the chatbot was removed from the task. Participants who had relied on AI earlier performed worse when they were forced to work independently.
Performance dropped after AI was taken away
In the math experiment, participants who had never used AI from the beginning solved about 73 percent of the questions correctly during the independent testing phase. By contrast, those who had previously depended on AI reached only 57 percent correct answers.
A similar gap appeared in the reading comprehension task. The group that had used AI scored 76 percent, while participants who kept working without AI reached 89 percent.
The researchers said the central issue was not only lower scores. More concerning was the decline in persistence, meaning the willingness to keep trying when a question became difficult. Participants who used AI were far more likely to skip hard questions or stop trying altogether.
That behavior suggests a shift in how people respond to challenge after being used to instant answers. The study argues that repeated exposure to easy solutions may leave users less prepared when they must think without digital support.
Not all AI use had the same effect
The findings did not treat every form of AI assistance equally. The strongest negative impact appeared among participants who asked AI for direct answers.
By contrast, participants who used AI mainly for hints, explanations, or clarification did not show the same level of decline. That distinction points to a key issue in how AI is used, not just whether it is used.
The researchers suggest that AI works better as a coach or guide than as a shortcut to the final answer. In that setup, the tool supports understanding instead of replacing the mental effort needed to learn and solve problems independently.
The study also places the issue in the context of everyday chatbot use, which has become normal in both work and study. AI is often the fastest route to a solution, but the research shows that speed does not always translate into lasting learning.
The paper describes the effect as a potential “boiling frog” problem. One interaction with AI may seem harmless, but repeated dependence can slowly weaken the habit of deep thinking and independent problem-solving.
In that sense, the difference between asking for a hint and asking for the full answer may matter more than many users realize. The study indicates that this choice can shape whether AI remains a supportive tool or becomes a shortcut that erodes persistence.
Source: www.indiatoday.in
