AI Makes You Give Up Faster? New Study Says Yes.

A new study has found that using artificial intelligence (AI) chatbots can make people less willing to keep trying when the help is taken away. The research, presented at the 2026 Conference on Language Modeling in San Francisco, suggests that getting instant answers may weaken our drive to work through problems alone. The study involved 1,222 participants across three randomized controlled experiments. Researchers tested how digital support affected two areas: solving math problems with fractions and reading comprehension. Each test lasted between 13 and 15 minutes, so the findings reflect short-term behavior, not long-term effects. In the first experiment, 354 people were split into two groups. One group could use a chatbot to solve 12 math exercises, either by asking for direct answers or for hints. The other group worked without the tool. Then researchers removed the AI and gave everyone three more problems. The group that had used the chatbot completed more tasks at first, but their performance dropped when working alone. They left more questions blank. The team repeated the test with 667 more participants and added a preliminary phase to exclude those who could not solve basic exercises. The results were the same. The drop was most noticeable among users who asked for direct answers rather than guided hints. A third experiment tested whether this effect appears outside of math. In this case, 201 people took a reading comprehension test. They answered five questions with or without AI help, then solved three more after the tool was disabled. The results showed the same pattern, suggesting that the loss of perseverance can appear in tasks that require understanding texts, not just numbers. One possible explanation offered in the study is that getting used to instant answers makes people see slower methods as less efficient. However, the experiments do not prove that AI causes widespread apathy or permanent cognitive decline. The tests were brief, and the authors acknowledge that more research is needed on how repeated use of these tools affects long-term learning. In response to these findings, researchers propose developing chatbots that act more like tutors. Instead of giving a direct solution, these tools could offer hints, explanations, and guidance so users complete the reasoning themselves. Brian Christian, co-author of the study and a researcher at the Center for Human-Compatible AI at the University of California, Berkeley, said in a statement that these technologies should enhance human abilities rather than replace them. New studies that follow the same people over longer periods are needed to test this approach.