wildpixel/iStock/Getty Images Plus
This year, for the first time in history, college seniors will graduate never having experienced a full semester without generative AI. The technology certainly has its many benefits, but when it replaces the mental struggles of learning, students lose something irreplaceable: the slow, difficult work that builds expertise and strengthens critical thinking.
As education experts from two California State University (CSU) campuses, we view overreliance on AI in much the same way as our public health colleagues view students’ academically motivated misuse of stimulants. Indeed, we view America’s educational institutions as key to preventing a potential epidemic of intellectual atrophy. Instead of making faculty obsolete, AI has intensified the need for their expertise and guidance. Like vaccines or exercise, authentic intellectual struggle, which well-prepared teachers foster, is preventative medicine for the mind.
At San Diego State University (SDSU), where two of us teach, our approach has been informed by data collected from 32,833 SDSU students over the past three years through a widely replicated AI survey. Anonymous findings revealed that by fall 2025, 98 percent of students surveyed had used AI.
In addition to the survey, we conducted interviews and informal focus groups to gain a deeper understanding of how students use AI. The interviews revealed a significant and striking divide, with students falling into two distinct clusters.
On one side are “appropriators,” who let AI do the work. Appropriators outsource cognitive effort, prioritize efficiency over intellectual growth and appropriate AI’s output as their own. The other group consists of “collaborators,” who engage AI tools to sharpen their own thinking, using AI as peer reviewers, tutors or thought partners whose outputs these students vet carefully.
For instance, Ian (name changed for anonymity) said in the add-on interview that if he is struggling with a paper, “I explicitly tell it, do not write this for me. I need to do this on my own, but I want help with understanding what I should write about.” After uploading a draft of his essay, Ian asked if it was on track. What happened next? “I took these responses, changed my essay, and then I asked about how it was after I changed it.” In this collaboration, Ian retained authorial responsibility.
Generally, intentional cheating aside, appropriators didn’t realize they were using shortcuts, undermining their own learning. They often understood themselves as using AI ethically and responsibly, even when they were essentially just copying answers the AI provided. Gabriel (name changed for anonymity), for instance, believed he was using AI just to “help” even though he depended on it to provide the explanation for statistics assignments.
Notably, the data revealed a threshold that divided students into two groups: AI appropriators and AI collaborators. Leveling up depended on how professors treated AI in the classroom. That is, rather than making professors redundant, AI has only amplified their importance: Interviews affirmed that most students need and want instructor guidance to cross the threshold into self-aware, responsible, collaborative AI use.
Because student overreliance on AI can have a devastating impact, for instance on subject matter mastery and cognitive development (not to mention emotional or mental health), drawing on public health’s harm reduction approach, which accounts for context, is both apt and worthwhile. Pharmacists don’t dispense prescriptions without explaining how to use them. Community health workers don’t recommend preventive actions without offering relevant support and justifications. Likewise, teacher-based scaffolding for AI use is important as a preventive measure.
Faculty can play a critical role in cultivating what we term ”good AI hygiene” among students if universities empower and equip them to do so. To start, we must enable faculty to offer meaningful assignments or, as students say, “relatable” work. Part of this involves explicitly conveying to students, in personalized terms, the value of intellectual effort for strengthening critical thinking skills—of treating AI as a training partner, not a servant.
Faculty must also be given the means (e.g., through professional development) to model ethical and responsible use. This includes demonstrating the importance of subject-matter expertise in assessing AI output and developing curricula and syllabi with clear AI guidelines to reduce uncertainty around academic integrity.
At SDSU, for example, we’ve developed a publicly available Academic Applications of AI Micro-Credential program, which offers support and a range of scenarios and simulations to help faculty achieve these goals. Additionally, SDSU has established Generative AI guidelines, and has supported some of the extra labor that fostering AI literacy requires. The university also continuously convenes stakeholder groups across the campus to ensure instruction evolves at the same velocity as the tools themselves.
Higher education’s job is to cultivate capable thinkers. Rather than making faculty irrelevant, AI has shown us how important they are to optimizing the developmental potential of our young people in the service of a long, healthy life of the mind. Investing in faculty for the fight ahead is vital.
