Can AI weaken critical thinking? What educators and researchers say

Earlier this year, education researcher Lixiang Yang realized that students could outsource all the work for their university courses to artificial intelligence.

Yang, who teaches at Tsinghua University in Beijing, China, was using an AI agent called Codex to set up online courses on a learning-management platform. But it quickly became clear that a skilled AI could independently complete every task, answer questions and participate in discussions “without the learner engaging intellectually”, Yang says.

The experience convinced Yang that teaching students to reason and think for themselves is “more important than ever”. He is not alone in that concern. In a 2025 survey of more than 1,000 US faculty and staff conducted by the Association of American Universities, 90% of respondents said they believe generative AI will reduce students’ critical-thinking skills. According to this year’s global survey of more than 8,000 respondents, around 20% of higher-education students who use AI say they find it harder to deal with problems without the technology.

“Learning occurs as a result of cognitive processing,” says Daniel Willingham, a cognitive psychologist at the University of Virginia in Charlottesville. “Cognitive processing typically takes effort, and AI effectively removes that.”

Science, in which people must critically evaluate theories and data, is one of the fields potentially affected by the AI revolution in large language models (LLMs). “There is a risk that students will outsource the very activities that develop scientific thinking, such as forming hypotheses, interpreting evidence, considering alternatives and grappling with uncertainty,” says Thomas Nygren, an education and critical-thinking researcher at Uppsala University in Sweden.

But all is not lost. Education and cognitive experts have studied critical thinking and how to develop it for years. They have practical advice for preserving and strengthening cognitive skills in the age of AI. “Critical thinking and building higher-order thinking skills has become even more important,” says Stéphane Vincent-Lancrin, who led a 10-year programme on critical thinking at the Organisation for Economic Co-operation and Development (OECD) in Paris.

What is critical thinking?

If you ask five people to define critical thinking, you might get five different answers. “Critical thinking is like the pinnacle of human cognition,” Willingham says. “It’s kind of a basket term that captures the cases that people think about a lot.”

Will AI ruin our skills? Early results are in, and they’re not good

One reason for the diversity of definitions is that, although the term is widely used in education and everyday discussions, it is less common among researchers who study cognition. Instead, they tend to focus on distinct, high-level processes involving different but overlapping mechanisms in the brain, including memory, reasoning, decision-making and problem-solving.

Researchers identify several common strategies for critical thinking. One is the ability to consider possibilities and perspectives different from one’s own, as scientists do when weighing competing hypotheses.

Critical thinking also requires knowledge and skills in a particular field. Being able to think critically in one discipline does not necessarily mean being able to do so in another. For example, one study found that neurologists were not very good at diagnosing cases of heart disease.¹

Is AI weakening critical-thinking skills?

Concerns about declining student intelligence are not new. In 1983, a report by the US National Commission on Educational Excellence warned that many 17-year-olds lacked “the ‘higher’ intellectual skills that should be expected of them”.

Over the past few decades, critical thinking has been a goal of many education systems, says Vincent-Lancrin. It is now considered an education priority by the United Nations cultural agency UNESCO and is part of the school curriculum in almost all OECD countries.

Vincent-Lancrin gives the example of a course in which students investigate the causes of type 2 diabetes and revise their thinking as they learn about different perspectives, including the contribution of genes and the influence of the environment.

However, the inclusion of critical thinking in curricula does not mean that the goal has been achieved. “There’s not much reason to think we did a great job” of teaching critical thinking in the past, Willingham says. “And now AI has come along and undone everything we’ve achieved.”

Teachers in schools and universities have many stories of students who are less able to solve problems or evaluate information independently because they outsource their work to AI. Research also supports the possibility that AI can erode skills.

A 2024 study found that middle-school students given access to generative AI tools such as ChatGPT became better at solving mathematics problems. But after access was removed, the students performed worse than those who had not used the tool.²

Overall, research has not yet clearly demonstrated how generative AI is affecting learning and critical thinking. AI tools are relatively new and diverse, and studies so far have produced contradictory results. There have also been too few rigorous or long-term studies.³ “Tools can be helpful, but they can also be detrimental,” says Yang, who reviewed the literature. “It goes both ways, actually.”

Some research suggests that AI can help people learn when it provides feedback and tasks that challenge students to think and solve problems, rather than simply supplying ready-made answers. “It’s all about how you structure your tasks,” Yang says.

However, Willingham and Yang acknowledge that people often want quick answers rather than using AI in more demanding ways. When Yang asked students to use an AI system designed to prompt their own thinking, some shut it down and switched to ChatGPT instead. “It’s kind of shocking, but it’s the truth, right?” he says.

Future research could provide clearer answers. In June, the London-based non-profit research group Education Endowment Foundation (EEF) announced that it would spend up to £2.5 million (US$3.4 million) on rigorous research into the effects of generative AI on learning, understanding, memory and problem-solving.

“I’m glad someone did this research,” Willingham says, although he expects the outlook to be bleak. “It’s going to take a while to figure out this slippage, but it feels pretty inevitable,” he says.

How can teachers develop critical thinking?

Critical thinking varies by discipline, which is one reason it is difficult to teach. “It’s not general, it’s not free-floating. You have to teach it within specific subjects,” says Paul Kirschner, an educational psychologist at the Dutch Open University in Heerlen.

This means giving students the specific knowledge and reasoning skills required in a particular field. Thinking critically in history might require support from records and artefacts, whereas critical thinking in science might require an understanding of the relationships between theories, hypotheses, methods and evidence.

A second challenge is that critical thinking often involves recognizing reasoning strategies in the abstract. Students might learn about the difference between correlation and causation in class, for example, but fail to notice when one is mistaken for the other in the real world.

Willingham recommends that teachers identify the most important reasoning skills, incorporate them into the curriculum and provide sufficient practice through a variety of problem types. A consistent curriculum that builds knowledge, skills and critical thinking across subjects and over many years is also important, Kirschner says.

The OECD project led by Vincent-Lancrin was one of the largest efforts to examine how critical thinking can be taught in schools. It involved around 800 teachers and included the development and testing of teaching materials, such as example lesson plans and syllabuses.

Is the internet and AI affecting our memory? What the science says

Vincent-Lancrin acknowledges that there is not much rigorous evidence about which approaches are most effective for teaching critical thinking. One reason is that the skill is more difficult to measure in a standardized way than mathematics or reading. Reliable measurement is necessary to determine whether a teaching strategy is working.

The OECD is working with Macat, a London-based education company, to develop standardized assessments of critical thinking for secondary-school and undergraduate students.