There was a time when not knowing something wasn’t seen as a problem to eliminate as quickly as possible. A difficult math question, a confusing paragraph in a book, or a puzzle that refused to make sense often became an invitation to think a little deeper. Those moments demanded patience, persistence, and sometimes frustration, but they also strengthened something far more valuable than simply arriving at the correct answer. They trained the mind to stay curious, to tolerate uncertainty, and to discover solutions through effort rather than convenience. Today, however, that experience is quietly disappearing. With artificial intelligence available in our pockets, many of us no longer wrestle with difficult questions. We ask, we receive, and we move on. The process feels wonderfully efficient, but a growing number of researchers are beginning to wonder whether something essential is being left behind.

A new study from researchers at Carnegie Mellon University, MIT, Oxford University, and UCLA has added fresh weight to that concern. Their findings suggest that relying on AI for even ten minutes during simple cognitive tasks may reduce a person’s willingness to persist once that digital assistance is removed. The issue is not that artificial intelligence instantly makes people less intelligent. Rather, it appears to influence the habits that intelligence depends on, including perseverance, independent reasoning, and the confidence to struggle through uncertainty. Combined with earlier research on what scientists call “cognitive debt,” the study invites a deeper conversation about our relationship with technology. Artificial intelligence may become one of humanity’s greatest tools, but the way we choose to use it could determine whether it expands our minds or gradually encourages us to stop exercising them.

The Experiment That Revealed an Unexpected Pattern

Researchers designed the study around tasks that most people would consider straightforward. Around 1,200 participants were asked to complete basic fraction problems and reading comprehension exercises through an online platform. Some participants worked independently, relying only on their own thinking, while another group had access to an artificial intelligence assistant capable of producing answers almost instantly. At first, the results looked exactly as many people would expect. Those using AI completed more questions correctly and often finished more quickly than those solving the tasks on their own.

The researchers, however, were far less interested in immediate performance than in what happened after the technology disappeared. Once the AI assistant was removed, participants had to continue solving similar problems without any digital support. That simple change exposed a striking difference between the two groups. People who had depended on AI were significantly more likely to make mistakes, skip questions altogether, or abandon difficult problems before reaching a solution.

Perhaps the most revealing observation was not about accuracy at all. It was about persistence. Participants who had become accustomed to receiving immediate answers appeared less willing to remain engaged when those answers were no longer available. The researchers argued that this willingness to struggle is one of the most important ingredients in genuine learning because new skills are rarely developed through effortless success. They grow through repeated attempts, mistakes, and gradual improvement.

The report captured that concern with a warning that reaches far beyond the classroom. The researchers wrote, “If such effects accumulate over months and years of AI use, we may end up creating a generation of learners who have lost the disposition to struggle productively without technological support.” It is a statement that shifts the discussion away from whether AI is useful and toward a more profound question about what happens when convenience quietly replaces effort.

Why Struggle Has Always Been Part of Learning

Modern life often teaches us to avoid difficulty whenever possible. Faster internet, quicker deliveries, and instant communication have conditioned us to see waiting as an inconvenience rather than a natural part of progress. Artificial intelligence fits comfortably into that pattern by removing many of the obstacles that once slowed our thinking. Yet educational psychologists have long argued that those obstacles serve an important purpose. They create what researchers call productive struggle, a process in which the effort required to solve a problem actually strengthens learning instead of preventing it.

Think back to a moment when you finally understood something that had challenged you for days or even weeks. The memory probably remained with you far longer than information that came easily. That is because the brain forms stronger connections when it actively works through confusion instead of passively receiving answers. Every failed attempt becomes part of the learning process, gradually building confidence alongside knowledge. Remove that struggle completely, and the lesson may never become deeply rooted.

Artificial intelligence changes that equation in subtle ways. Rather than encouraging someone to wrestle with uncertainty, it often presents a polished solution within seconds. The answer arrives before the mind has fully explored the problem itself. While this improves efficiency, it may also reduce the opportunity for deeper understanding. Over time, the habit of turning to AI first could make independent thinking feel unusually difficult, not because people have become less capable, but because they have had fewer opportunities to strengthen those mental muscles.

The researchers believe this shift deserves careful attention precisely because it develops so gradually. One conversation with an AI assistant is unlikely to reshape the human mind. Thousands of interactions spread across months or years, however, may slowly influence how comfortable people become with uncertainty, challenge, and independent reasoning. Like physical fitness, intellectual resilience grows through regular exercise, not through avoiding effort altogether.

The Rise of Cognitive Debt

This new research builds upon another influential study conducted earlier this year by researchers at the Massachusetts Institute of Technology. That investigation focused on people using OpenAI’s ChatGPT to write essays. While AI users often completed their assignments more quickly, researchers noticed something unexpected afterward. Many participants struggled to remember what they had written or recognize passages from their own work. Brain activity associated with deeper cognitive engagement also appeared lower compared with participants who completed the writing independently.

The researchers described this phenomenon as cognitive debt, a phrase that captures how mental shortcuts can accumulate hidden costs over time. Much like financial debt, cognitive debt rarely creates immediate consequences. Borrowing effort today may feel harmless because the benefits appear instantly. The payment arrives later, often in the form of weaker memory, reduced understanding, or diminished confidence when facing similar challenges without assistance.

Viewed together, the two studies paint a remarkably consistent picture. One suggests that AI may reduce the persistence required for learning. The other indicates that it can weaken long-term engagement with the material itself. Neither study argues that artificial intelligence is inherently harmful, but both suggest that the design of these tools matters enormously. Systems built only to provide answers may produce different long-term outcomes than systems designed to teach people how to discover those answers themselves.

Michiel Bakker, an assistant professor at MIT and one of the researchers involved in the latest study, believes this distinction should shape the future of AI development. “AI can clearly help people perform better in the moment, and that can be valuable,” he explained. “But we should be more careful about what kind of help AI provides, and when.”

The “Boiling Frog” Effect of AI Dependence

One of the most compelling ideas to emerge from the research is something the authors describe as the “boiling frog” effect. The metaphor suggests that gradual change often goes unnoticed because each individual step feels too small to matter. A frog placed in slowly warming water supposedly fails to recognize the growing danger until it is too late. While the metaphor itself has been debated scientifically, its message remains powerful. The researchers believe our relationship with artificial intelligence may evolve in much the same way.

No one loses their critical thinking after asking ChatGPT to explain a difficult concept or summarize an article. The first interaction saves time, and that feels harmless. The second feels even more natural. Before long, AI begins drafting emails, organizing ideas, solving equations, and answering questions that once required careful thought. None of these individual moments seem significant enough to cause concern. Yet together they may slowly reshape what we expect from our own minds.

The study argues that AI changes more than the way we complete tasks. It changes our expectations about effort itself. When answers consistently arrive within seconds, anything requiring sustained concentration can begin to feel unusually difficult. Instead of interpreting struggle as a normal part of learning, people may start viewing it as a sign that they should simply ask the machine again. Over time, the habit of seeking immediate solutions becomes stronger than the habit of thinking independently.

That possibility deserves attention because intellectual resilience is rarely lost all at once. It fades through countless small decisions that seem perfectly reasonable in isolation. The convenience of AI is real, but so is the possibility that convenience can slowly replace the patience required for deep understanding. The researchers warn that the cumulative effect, rather than any single interaction, is what society should be watching most closely.

AI Should Challenge Us, Not Replace Us

Despite the study’s concerning findings, the researchers are careful not to frame artificial intelligence as something that should be feared or rejected. Their message is surprisingly balanced. AI has enormous potential to improve education, increase productivity, and make knowledge more accessible than ever before. The challenge lies in deciding how that assistance should be delivered. Technology that solves every problem immediately may produce very different outcomes than technology designed to help people develop their own abilities.

Michiel Bakker believes future AI systems could function more like experienced teachers than answer machines. A good teacher does not immediately reveal the solution whenever a student becomes confused. Instead, they ask questions, offer hints, encourage reflection, and allow the learner to experience the satisfaction of reaching an answer through their own effort. That process takes longer, but it also builds confidence that remains long after the lesson ends.

As Bakker explained, “Systems that give direct answers may have very different long-term effects from systems that scaffold, coach, or challenge the user.” That distinction may sound subtle, but it represents two very different visions for the future of artificial intelligence. One treats human thinking as something to replace. The other treats it as something worth strengthening.

Finding that balance will not be easy. People naturally prefer tools that are fast and effortless, while educational experiences often require patience and discomfort. Designing AI that sometimes chooses not to provide the easiest answer may feel frustrating in the moment. Yet it could prove far more valuable over the course of a lifetime. The goal should not be to compete with human intelligence, but to cultivate it.

What This Means for the Way We Learn

The implications of this research extend far beyond classrooms or homework assignments. Every profession increasingly relies on digital tools capable of analyzing information, generating ideas, and producing polished work within moments. Writers, programmers, marketers, designers, and students all face the same temptation to outsource not only repetitive tasks but also the thinking that gives those tasks meaning. The question is no longer whether AI belongs in our daily lives. It already does. The more important question is whether we are using it intentionally.

History offers a useful perspective. Calculators transformed mathematics, yet schools still teach arithmetic because understanding numbers remains important even when machines perform calculations more quickly. GPS changed navigation, but people still benefit from developing a sense of direction. Search engines made information easier to access, yet education continues to value critical thinking over simple memorization. Artificial intelligence represents another step in that evolution, but one with broader implications because it increasingly imitates reasoning itself.

Rather than viewing AI as either a miracle or a threat, perhaps the healthiest approach is to treat it as a partner whose value depends on how we engage with it. There are moments when instant assistance is genuinely helpful, especially when learning something new or overcoming a temporary obstacle. There are also moments when resisting the shortcut allows understanding to grow in ways that no machine can replicate. Knowing the difference may become one of the defining skills of the AI era.

Some practical habits can help preserve that balance:

• Attempt the problem first before asking AI for assistance.
• Use AI to explain concepts, not simply generate finished answers.
• Question AI responses instead of accepting them automatically.
• Reflect on what you’ve learned after using AI, rather than moving immediately to the next task.
• Make room for productive struggle, recognizing that frustration is often part of genuine growth.

These habits do not reject technology. They simply ensure that technology continues to serve human development instead of quietly replacing it.

The Future of Intelligence Is Still a Human Choice

Artificial intelligence is advancing at a remarkable pace, and there is little doubt that it will continue transforming the way we live and work. The question raised by this latest research is not whether these tools should exist, but what kind of relationship we want to build with them. Every generation inherits technologies that reshape daily life. Few, however, have encountered tools capable of performing so many cognitive tasks once considered uniquely human.

The findings from Carnegie Mellon, MIT, Oxford, and UCLA do not suggest that our brains begin to deteriorate after ten minutes with AI. What they do suggest is far more subtle. Our habits shape our abilities, and the habits encouraged by technology matter. If we allow every challenge to be solved for us, we may gradually lose confidence in solving challenges ourselves. If, instead, we use AI as a guide that expands curiosity rather than replacing it, the technology could become one of the greatest educational companions humanity has ever created.

Perhaps the real opportunity lies in remembering that intelligence has never been measured solely by how quickly someone finds an answer. It is revealed in the willingness to ask better questions, remain curious through uncertainty, and keep thinking long after the easiest solution has appeared. AI may provide answers in seconds, but the deeper work of becoming wiser still belongs to us.

Sources:

1. Liu, G., Christian, B., Dumbalska, T., Bakker, M. A., & Dubey, R. (2026). AI assistance reduces persistence and hurts independent performance. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2604.04721
2. Kos’myna, N. (n.d.). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task – MIT Media Lab. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/?_bhlid=0fb4df91e32c50d58a8a7b29aab1525665676640

Loading...