Generative AI has made schoolwork faster and smoother than ever before. Across campus, students report using it to summarize reading and help draft essays, and faculty are rapidly experimenting with it in the classroom and through research as well. This new technology removes friction from academic research. However, friction is often an issue in higher education. Are we accelerating learning or simply outsourcing to AI? Where is the line between supporting student ideas and completely replacing their cognitive efforts?
At Oberlin College and the Conservatory of Music, where we recently marked our first steps into AI with the Year of AI Exploration, we saw these kinds of questions about AI play out in real time. At a faculty workshop, educators questioned the line between using AI to streamline feedback and the risk of students submitting sophisticated work that they cannot explain. A common theme emerged during the student sessions. Few felt confident that they understood where support ended and overdependence began.
Behind all the hype, potential, and efficiency, there are quiet risks that higher education has yet to adequately address. What would happen if we stopped thinking for ourselves?
What is cognitive debt?
Humans have always found creative and innovative ways to use tools designed to ease our burdens. In the digital age, note-taking devices, calculators, or setting reminders on your phone are all tools designed to support what psychologists call cognitive offload. Offloading is often very beneficial, freeing up mental space for deeper work such as reflection and analysis.
But unlike a calculator or a notebook, AI does more than just store information. It interprets it, synthesizes it, and generates an answer. Similar to technical debt in software development, cognitive debt accumulates over time. This is built by cognitive offloading providing short-term convenience and replacing sustained engagement, a core principle of education at all levels. Although students may feel more efficient in the moment, they risk losing their ability to think critically, solve problems independently, and retain knowledge over time.
Why this is important for learning
Learning doesn’t always have to feel easy. In fact, deep learning requires “productive struggles,” such as when students grapple with confusion, test formative ideas, and navigate uncertainty. AI can shorten that process.
At one campus workshop, a faculty member shared a perfect example of this problem. A student in a foreign language class submitted a beautifully written answer that appeared to be generated by an AI, but was having a hard time discussing even the basic concepts behind it with the class. This work seemed powerful on the surface, but there was no underlying understanding.
When a tool quickly explains a concept or rewrites an argument, it removes friction that helps the brain build strong neural connections. According to a 2025 article in frontiers of psychologyhabitual offloading to AI reduces processing depth. If students consistently ignore their efforts, the effects will be felt later in the form of decreased reasoning and problem-solving skills.
In other words, cognitive debt not only impacts today’s challenges but also threatens students’ long-term academic trajectories.
The real problem is not technology
Generative AI has proven to be a polarizing but undeniably useful tool in higher education. Proactively break down accessibility barriers, help faculty close learning gaps, and expand staff efficiency and research capacity. But it also poses a problem for academic integrity, fosters cognitive offload, and invades every aspect of our daily work, from Zoom meetings to proactive email drafts. Amid such dramatic changes, it is very tempting to view AI purely as a technology issue and simply ban it from campuses.
However, this challenge stems from three fundamental institutional gaps regarding AI.
1. Aligned leadership and organizational governance
2. Intersection with pedagogy/andragogy
3. Formal guidance, policies and processes
At the end of the day, this is not a technical issue, but a leadership and governance issue. Guidelines are located throughout the map. We saw this firsthand in our year of AI exploration. In one department, professors warned students not to use AI at all, while in another they encouraged free experimentation. Meanwhile, new enterprise tools like Gemini were being rolled out. Students felt confused by these mixed messages, making it difficult to make ethical decisions about how and when to use these tools. Oberlin is not alone in this. Many higher education institutions I have connected with over the past nine months share similar examples of gray areas and blind spots. This type of mismatch creates real risks.
- confusion. Students do not know about authorship, academic integrity, or even the correct use of AI tools.
- Equity gap. Students who become proficient in AI will gain a disproportionate advantage over other students.
- Overdependence. Without clear and communicative guardrails, convenience will always prevail.
While many institutions report strong interest in AI tools among faculty, research highlights that many still struggle with low self-efficacy regarding ethical risks. We saw this very theme emerge in our own faculty, staff, and student surveys.
What students actually need
Rather than needing “less AI,” students need better guidance on how to use AI as a scaffold rather than a shortcut.
In student surveys and feedback sessions, students said they wanted clear guidelines, training, and organizational adjustments. They simply didn’t know what a “good” use of AI looked like. Once we made our expectations clear, they were much more willing to engage. This played out in a variety of ways. Students completed AI literacy training, the Student Senate took the lead in proposing an Academic Integrity update to the Student Code of Conduct, and students led roundtables and public forums on the environmental impact of AI and its use at Oberlin, a carbon-neutral institution starting in 2025.
At a practical level, we have an obligation to teach students basic AI literacy skills as AI becomes more ingrained in our culture through software, applications, and products and becomes a required skill for employment. But we must also teach students to notice whether AI is supporting or replacing their thinking.
Here are three simple requirements that institutions can adopt:
- Ask students to describe the output of the AI. If students use AI to generate an answer, they must be able to explain it without tools.
- Teach students to ask questions of the tool. Turn AI from an answering machine to a starting point for questions. Students should always ask, “Why did the AI suggest this?” What assumptions are you making? Whose perspective is missing?
- A fundamental flaw in AI, especially large-scale language models, is the perpetuation of biases in the training data. Students must be able to spot these biases and make informed decisions.
- It is also important to understand that AI does not evaluate truth, fairness, or relevance.
- Incorporate “pause points” for critical thinking: Encourage students to use AI to organize ideas early on, but require group or independent work for interpretation and final synthesis. I tell faculty to consider allowing students to use AI during the acquisition phase of learning, but use their own creative and critical thinking skills during the application and evaluation phase.
What should institutions do next?
Universities don’t need a 50-page policy on AI. We need a clear, easy-to-use framework that bridges the gap between innovation and responsibility. These frameworks also need to be fluid enough to be used by a variety of stakeholders, including curriculum departments, staff groups, student groups, administration, and researchers.
We have found that the most effective strategy is to embed small-scale practical guidance directly into these groups through means such as staff and department handbooks, sample course syllabus language, and institutional review board and grant language for AI in research, to name just a few. And yes, we finally created an organizational AI policy, along with countless resources, help guides, workshops, listening sessions, and more. It is important to note that these elements were realized organically through a cross-sectoral and community-based approach.
We launched the Year of AI Exploration and invited our faculty, staff, and student community to join the conversation. Rather than inheriting IT policies or organizational mandates, we chose to listen to our employees and consider culture. And it worked. Now, instead of pointing to text-centric policies or university-wide mandates, faculty simply add a few lines to assignments and syllabi explaining exactly when AI is appropriate.
To make this work at scale, we need shared accountability.
- Teachers must use their pedagogical expertise to define where AI supports learning and where it inhibits learning.
- Staff should rely on AI for operational efficiency, but must apply human judgment to maintain ownership of the final product.
- Students need to consider AI ethically and combine it with their own critical thinking skills and integrity.
moment of opportunity
There are options in higher education. Ignoring AI is doing students a disservice in preparing them for the world of tomorrow. On the other hand, blindly adopting AI risks increasing cognitive offload and cognitive debt to society. We risk producing faster but less capable students who passively consume information from tools they do not fully understand.
But if we address this issue directly and highlight the curiosity, vigilance, and concern of each community, we can teach students how to think with AI without giving up their ability to think for themselves. AI offers unprecedented speed and scale, but lacks true human judgment.
Successful institutions will not be the first to adopt AI. They just pay a premium for early access. Successful institutions will be those that focus on how AI intersects with their values, culture, and users. They amplify the enduring power of human intelligence, ethics, creativity, and empathy and teach students to use it wisely.
Chris Drennen is director of academic technology and instructional support at Oberlin College and Conservatory of Music.
