Over-Reliance on Artificial Intelligence Is Weakening Complex Problem-Solving: University of California Study

Almost every morning, coffee in hand, many of us now open our laptops and first type a question to an artificial intelligence (AI) chatbot. Complex calculations, draft reports, even computer code — this technology has become a helper in all of it. Work is easier, time is saved, answers arrive instantly. But does this convenience carry a hidden cost? A recent University of California study suggests the answer is probably yes. The study indicates that those who rely heavily on AI for complex problem-solving may do well at first but fall behind later. Once the helper is taken away, their performance drops noticeably. So what evidence sits behind this warning? Broadcast in Spanish, the warning comes from a study at the University of California. It involved one thousand two hundred and twenty-two online participants. They were split across three separate experiments — three hundred and fifty-four in the first, six hundred and sixty-seven in the second, and two hundred and one in the third. That design matters, because a consistent result across three separate trials raises confidence in the conclusion far more than a single small sample would. At the center of the study was one question: when people use AI, what changes in their concentration, their error rate, and their persistence? The researchers wanted to see how human performance behaves when an AI helper sits alongside during a complex task, and what happens when that helper is removed. Brian Christian, affiliated with the University of California, Berkeley, was involved in the work. After the results emerged, discussion began. It appeared that those who used AI during the task performed well early on — solving complex problems quickly and accurately. But when the helper was removed mid-experiment, their performance began to slide. According to one statistic from the study, this group solved only fifty-seven percent of the complex problems correctly, whereas those who had not used AI reached seventy-three percent. The gap is not small — a sixteen-point spread shows how little one's own skill holds up once the helper disappears. Another observation surfaced: after roughly twelve exercises, participants' performance began to decline. In other words, when AI hands over the answer at every step, the user's own analytical skill gradually erodes. This is not a sudden collapse but a slow, accumulating effect. A larger question emerges here. What is AI actually doing — expanding human thinking, or lifting the act of thinking onto its own shoulders and leaving people passive? The findings suggest the answer depends on the type of use. If someone treats AI as a tool for asking questions, verifying claims, and building their own understanding, benefit follows. But if someone simply asks AI for the answer and treats it as final, the chance to build their own skill is lost. This study should not be read one-sidedly, however. The researchers stated clearly that the damage caused by AI use is not permanent. The paper notes that the result is not a final verdict; it is a caution. The key distinction is whether AI is being used as a learning aid, or whether it is doing the entire act of thinking for someone. The first builds skill; the second builds dependence. That subtle distinction is the most important part. The study does not claim technology is harmful. The real question is what the relationship between people and technology will look like. If someone uses AI like a coach — trying themselves, making mistakes, then verifying — their skill grows. But if someone uses AI purely as an answer machine, their problem-solving muscle slowly weakens. This university study is therefore not only for students; it is equally relevant to professionals, journalists, programmers, and researchers who use AI daily. If someone leaves the entire job of writing reports to AI every day, then after a few months their own ability to produce a report weakens. The risk is not immediate; it accumulates — much like a muscle weakening after physical exertion stops. For education systems, the finding carries particular weight. If a student uses AI to solve everything before an exam, the grade may look good — but the exam hall will have no AI. Then they are left with only their own preparation. The study is therefore a signal to teachers: the center of learning must be the habit of thinking, with AI as an aid to building that habit, not a substitute for it. It is worth noting that although the study is disseminated by the University of California, many reports carry no direct citation or link. Reaching a final verdict on the basis of a short warning alone, without reading the full study, would be unwise. Even so, one lesson stands out clearly from what is known — the relationship with AI should be conscious, not habitual. On the other hand, one reality must be accepted: the study is not a call to stop using AI. Technology is now an inseparable part of work, and denying that is not realistic. Rather, using AI wisely — understanding where to stop — is the real lesson. Knowing when to question the helper and when to use one's own head is the decisive balance. Interestingly, this warning is not a new anxiety. When the calculator, the internet, and the search engine arrived, the same question was asked — will people forget how to calculate, or learn new skills? History suggests the answer depends on how people use the technology. With AI, the same question stands before us, only larger in scale. What to watch in the coming days is how universities and institutions design their rules for AI use. If the habit of thinking stays at the center of learning, with AI as only an aid, the benefits of technology can be captured and the harm avoided. But if the aid becomes the center, the question will remain — are we actually learning, or merely receiving answers?