The Critical Thinking Gap in Education

In fall 2025, 90 percent of faculty surveyed by the American Association of Colleges and Universities said generative AI will diminish students’ critical thinking skills. The number should alarm us. It also reveals something these instructors rarely ask out loud: if a new tool can erode critical thinking this quickly, it raises the question of whether instructors were ever really teaching it in the first place.

As artificial intelligence has continued to infiltrate classrooms and dorm rooms, much of the public narrative has focused on preserving critical thinking in the age of AI. Conversations focus on cognitive offloading, how the technology is reducing friction in the learning process and the risks of dependency on these systems. And yet, while we tout the importance of critical thinking skills for college graduates, we do very little to explicitly teach those skills. This situation has produced a long-standing but often overlooked crisis in that ability, one that predated AI. In fact, one major study found that at least 45 percent of students show no significant improvement in complex reasoning and writing over their first two years of college.

Failure of the By-Product Model

Most of us inherited a pedagogical model in which critical thinking was the residue of teaching and learning rigorous content. We assigned hard readings, demanded evidence-based arguments and graded for analytical quality with a hyperfocus on the content we were trying to teach: early modern philosophy, discrete mathematics, Mandarin, you name it. This thinking was supposed to develop in the manipulation and application of the concepts and methods in our home disciplines. Maybe we got lucky and taught this ability without really meaning to — but the evidence above seems to indicate that often wasn’t the case.

The idea that a college education at least nudges learners toward critical thinking is, in many cases, correct. Many professors promote and model this ability through their disciplinary content; that is not in dispute. What we are arguing is that embedding and attempting to teach it while also trying to pass on disciplinary content, even when done carefully, is not the same thing as teaching it explicitly, and that distinction matters even more now than it used to.

Examining the effectiveness of that by-product approach is an empirical question, not just a philosophical one, and the emerging data gives us some uncomfortable answers. The evidence suggests strategies like structured dialogue and exposure to authentic problems are effective for teaching analytical reasoning, but only when they are deliberately designed into instruction, rather than assumed to emerge from disciplinary content alone. Even when students pick up these skills from disciplinary instruction, evidence shows that they are unable to generalize them to other contexts when transfer isn’t explicitly practiced.

Related: Students Question If They Matter on Campus

We have managed some success thus far because students develop a semblance of these skills through the friction of research, drafting and revision. However, as AI erodes our ability to force friction into the learning process outside the classroom, we even more urgently need to turn to developing such skills directly rather than through proxies.

Increasing evidence shows that student use of AI tools for education is inevitable. The Digital Education Council’s global student survey found that 86 percent of learners already use AI in their studies, with 24 percent using it daily. At this point, we have to ask whether we have built a strong enough foundation of independent thinking for that use to be productive rather than substitutive.

A 2025 study reported a significant negative correlation between frequent AI tool use and critical thinking scores, with younger users showing the strongest effects. Notably, however, another 2025 study from Microsoft Research and Carnegie Mellon researchers found that self-confidence in one’s own abilities predicts higher levels of critical thinking by knowledge workers even when they use AI. Educators should note this finding, as self-confidence in thinking is exactly what explicit instruction can teach.

When such a large proportion of us say a tool can erode the thinking we claim to develop, we are making an admission about how fragile that development was in the first place. If we want students who can actually think critically in a world saturated with AI-generated content, we have to stop treating critical thinking as an emergent property of higher education and start treating it as a discipline that requires instructor training, explicit instruction for students, deliberate practice and its own place in our curriculum.

From a By-Product Model to a Focus Model

What this means concretely is that making critical thinking a focus rather than a by-product requires a shift in what we believe our job is and a willingness to teach something most of us were never trained to teach. Those of us who were trained to teach reasoning as an explicit skill set nevertheless often presented those skills as rigidly formal. Formal logic, often listed as satisfying critical thinking core-curriculum standards at the college level, is perhaps the most notable attempt at teaching strong thinking skills, yet it rarely transfers to other domains.

New disciplines are emerging to make reasoning instruction practicable, measurable and transferable. One such discipline, often called argument mapping, visualization or diagramming, attempts to visualize and systematize informal reasoning. Rather than represent reasoning as a formal linear proof, or rely on free-form classroom discussion with no shared structure, argument mapping is a method in which a professor or her students make claims and inferences explicit on a visual canvas. Research shows that explicit instruction in argument visualization produces statistically significant gains in students’ analytical reasoning and argument comprehension on standardized assessments. It is also a method that can be used when reasoning gets tricky in nearly any discipline, making it well suited to building transferable skills.

Related: Get to know Belts and standards in Taekwondo and BJJ

Across the research, three principles recur. First, students need vocabulary. They have to be able to identify and explain argumentative structures. As a starting point, they need to, for example, be able to read or listen to an argument and know what the author’s ultimate conclusion is (which turns out to be far from a universal skill for entering first-year students). Second, they need deliberate practice with feedback. This requires moments when the thinking itself is the subject of instruction and assessment. When they assert that the conclusion is the opposite of what the author actually said, they need help diagnosing their mental moves, and they need help constructing a new approach before trying again. Third, they need metacognitive scaffolding. They need to develop the ability to monitor and critique their own reasoning through introspection on how they are thinking, especially when new topics and contexts arise.

In our current by-product model, students do critical thinking without knowing that they are doing it, and faculty assess what the critical thinking is supposed to produce without assessing the actual thinking process. In the focus model, the thinking is surfaced, named, practiced and evaluated.

It can then be integrated and applied to the disciplinary topics and issues that animate the classroom experience, especially for faculty.

The focus model does not require adding a new course to every major. For example, first-year seminars, writing courses and general education requirements are natural homes for explicit critical thinking instruction that scales without redesigning the majors.

Explicit, cross-disciplinary critical thinking instruction helps students recognize the transferable structure of the skills, not just their domain-specific application. This means the skills they develop through this instruction can apply to a variety of disciplines and contexts and become lifelong skills.

Related: ED Backs Out of NASFAA Conference Sessions

For the first time in decades, there is widespread faculty consensus that critical thinking is at risk, institutional attention on the problem and a concrete reason to move discussions about how to address it beyond campus lecture series and committee discussions.

That alignment will not last forever.

If we are going to continue to tell students, parents and employers that a college degree means the graduate can think critically, we need to be able to demonstrate how we taught that skill and how we assessed whether the student developed it. Moving forward, institutions that build reputations for genuinely developing critical thinking will have a meaningful advantage in an environment where employers increasingly doubt the value of a degree. Indeed, employers rank critical thinking as the most important skill colleges should teach, yet only 49 percent say recent graduates are “very well prepared” in it.

Most of us entered this profession because we believe in the capacity of our students to learn to evaluate, question, synthesize and create and in the power of rigorous, deliberate thought. We owe our students more than the hope that such thinking will develop. We owe them the deliberate, explicit, transparent, sustained instruction that gives them the tools to think for themselves.

AI’s real threat is to the illusion that we were teaching critical thinking all along.