AI Is Making Students Lazy and That's the Least of It

The Devil You Summon

The number is almost too neat to be believed. 68.9 percent. That is the proportion of laziness among university students that researchers say can be attributed to artificial intelligence. Not a rounding error. Not a vague correlation. More than two thirds of the inertia, the procrastination, the outsourcing of thinking that now characterizes how young people move through their education.

Here is the strange thing. When Sayed Fayaz Ahmad, Heesup Han, Muhammad Mansoor Alam, and Mohd. Khairul Rehmat published their study in 2023, they were not trying to make a sensational claim. They were trying to measure something that everyone in education already sensed but nobody had put a number on. They surveyed 285 students across universities in Pakistan and China, ran the data through PLS Smart statistical modeling, and found that AI does not just enable laziness. It manufactures it.

And that is the least troubling part of what they discovered.

What 285 Students Revealed About Thinking Less

Ahmad et al. (2023) used what is called a purposive sampling technique. That means they deliberately chose students who had regular exposure to AI tools in their academic work. These were not Luddites or technophobes. These were the students who use ChatGPT to draft essays, who let Grammarly rewrite their sentences, who ask AI to summarize papers they do not want to read. The researchers then measured three things: loss of decision making ability, laziness, and concerns about privacy and security.

The results were stark. AI accounted for 68.9 percent of the variance in laziness scores. That is a statistical way of saying that if you want to predict whether a student has become lazier over the course of their studies, the single best predictor is how much they rely on AI. Not their personality. Not their workload. Not their sleep habits. The tool itself.

But here is where the story gets more interesting. The same data showed that AI explained 68.6 percent of privacy and security concerns. That is nearly identical to the laziness figure. Students know they are being watched. They know their queries are being logged, their drafts analyzed, their thought processes turned into training data. And they mostly do not care enough to stop using the tools.

The smallest number in the study was the one that should worry us most. Only 27.7 percent of the loss in decision making ability could be attributed to AI. That sounds like good news. Until you think about what that 27.7 percent actually represents.

The Muscle You Stop Using

Here is the thing about decision making. It is a practice. Every time you choose between two sources, evaluate an argument, or decide which piece of evidence matters most, you are strengthening a cognitive muscle. When AI hands you the answer, that muscle atrophies.

Ahmad et al. (2023) found that the loss of decision making ability was the smallest of the three effects they measured. But that might be because it is the hardest to detect in the short term. Laziness is obvious. You can see a student not doing their work. Loss of decision making is invisible. It looks like a student who gets the same grade but took a different path to get there. A path that did not require them to think.

The researchers put it bluntly in their paper. Accepting AI without addressing these human concerns, they wrote, would be like summoning the devils. That is strong language for an academic journal. It suggests that the authors believe the stakes are higher than most of their peers are willing to admit.

How the Study Was Actually Done

To understand why this paper matters, you need to understand how the evidence was gathered. Ahmad and his colleagues used a structured questionnaire administered to 285 students. The students came from multiple universities in Pakistan and China, two countries where AI adoption in education has been aggressive and where the cultural attitudes toward technology differ from Western norms.

The researchers used SmartPLS, a statistical software package designed for structural equation modeling. That is a fancy way of saying they built a model that could separate the effects of AI from other factors that might cause laziness or poor decision making. They measured their constructs using validated scales. They checked for reliability and validity. The methodology is standard for this type of social science research.

What makes the study notable is not the method but the timing. Published in 2023, it caught the wave of AI adoption just as it crested. The paper has already been cited 317 times, which in academic terms is a signal that other researchers found it useful enough to build on.

There is a complication. The paper was later retracted. The retraction notice does not allege fraud or data fabrication. It cites concerns about the peer review process and the editorial handling of the manuscript. That matters. But it does not mean the findings are wrong. It means the academic community is still debating how to handle research on a topic that is moving faster than the journals can keep up.

What This Research Does Not Prove

Let me be clear about what Ahmad et al. (2023) did not find. They did not find that AI causes laziness in the way that a virus causes a disease. They found a statistical association. Students who use AI more report being lazier. But it is possible that lazier students are simply more drawn to AI tools. The causality could run in either direction.

The study also did not measure academic outcomes. It did not track grades, completion rates, or long term career success. It measured self reported attitudes and behaviors. Students told the researchers they felt lazier and less capable of making decisions. That is real data. But it is not the same as objective measures of performance.

And the sample, while carefully selected, is not globally representative. Students in Pakistan and China face different educational pressures than students in North America or Europe. The cultural context matters. In societies where rote memorization and exam performance are heavily emphasized, the temptation to use AI as a shortcut may be stronger.

None of these limitations invalidate the findings. They just define their boundaries. The study is a warning sign, not a final verdict.

The Privacy Problem Nobody Is Solving

The second largest effect in the study was privacy and security. 68.6 percent of the variance in privacy concerns was tied to AI use. That number is almost identical to the laziness figure, and it tells a different story.

Students know their data is being collected. They know their writing is being analyzed. They know that every prompt they type into an AI tool becomes part of a training set that the company owns. And they use the tools anyway.

This is not naivety. It is a rational response to a system that has already made the choice for them. If every professor expects polished, AI assisted writing, then refusing to use AI is not a principled stand. It is a competitive disadvantage. The students in Ahmad et al. (2023) understood this. They reported high levels of concern about privacy and simultaneously high levels of AI use. They were trapped between two bad options: sacrifice their privacy or sacrifice their grades.

The researchers argue that this is a design problem. The tools are built to extract data, not to protect users. When you design a system that forces students to choose between privacy and academic success, you have already failed. The question is whether educators are willing to redesign the system.

The 27.7 Percent That Changes Everything

The loss of decision making ability is the smallest effect in the study. But it might be the most important.

Think about what decision making means in an educational context. It means choosing which sources to trust. It means evaluating conflicting evidence. It means deciding when you have enough information to form a conclusion. These are the skills that education is supposed to teach. If AI is eroding them, even by 27.7 percent, then the tool is undermining the very purpose of the institution using it.

Ahmad et al. (2023) measured this as a loss of human decision making. That phrasing matters. They are not talking about students making bad decisions. They are talking about students making fewer decisions. The AI makes the choice, and the student ratifies it. Over time, the student loses the ability to recognize that a choice was even made.

This is the quiet damage. It does not show up in exam scores. It does not get flagged by plagiarism detectors. It manifests as a student who can produce a perfect essay but cannot explain why one argument is stronger than another. It looks like competence. It is actually dependency.

Why the Cheating Frame Misses the Point

Most of the public conversation about AI in education has been about cheating. Can students use AI to write their papers? How do we detect it? Should we ban it?

Ahmad et al. (2023) suggest that this framing is dangerously narrow. Cheating is a behavioral problem with a behavioral solution. You can design better exams. You can use detection software. You can enforce honor codes. But laziness, loss of decision making, and privacy erosion are structural problems. They are built into the technology itself.

A student who uses AI to cheat knows they are cheating. They are making an active choice to violate rules. A student who uses AI to help them think is not cheating. They are following the rules. But the effect on their cognitive development may be the same. The tool does the work. The student does not practice the skill.

This is what the researchers mean when they warn about summoning devils. The danger is not that students will break the rules. The danger is that they will follow the rules perfectly and still lose the ability to think for themselves.

What This Actually Means

The study by Ahmad et al. (2023) is not a call to ban AI. It is a call to be honest about what we are trading.

• ▸If you are an educator, stop pretending that AI is just another tool like a calculator. A calculator performs a mechanical operation. AI performs a cognitive one. When you assign work that AI can complete, you are no longer teaching students. You are teaching the AI. Redesign assignments so that the process matters more than the product.

• ▸If you are a student, recognize that every time you ask AI to generate a first draft, you are skipping the hardest and most important part of learning. The struggle to organize your thoughts is not a bug. It is the feature. The study measured a 68.9 percent increase in laziness. That is not a judgment. It is a description of what happens when you outsource thinking.

• ▸If you are a policymaker, pay attention to the privacy number. 68.6 percent of privacy concerns tied to AI means that students are aware of the surveillance but feel powerless to resist. Regulation should focus on giving students real choices, not just terms of service agreements that nobody reads.

• ▸If you are a researcher, the retraction of this paper should not make you dismiss its findings. It should make you replicate them. The questions Ahmad et al. raised are too important to ignore because of a procedural dispute. The world needs more data on this, not less.

• ▸If you are anyone who cares about education, the 27.7 percent loss of decision making ability is the number to watch. It is small now. It will not stay small. Every generation of AI tools gets better at hiding the work they do. Every generation of students gets more practice letting the machine decide. The question is whether we will notice before the number hits 50 percent.

The devils are already here. The question is whether we are brave enough to look at them directly.