The public debate over artificial intelligence in higher education revolves around a familiar concern: cheating. Will students use chatbots to write essays? Can instructors tell? Should universities ban this technology? Adopt it?
These concerns are understandable. But focusing so much on cheating misses the larger transformation already underway, which extends well beyond student misbehavior and even the classroom.
Universities are adopting AI in many areas of institutional life. Some uses are largely invisible, such as resource allocation systems, report “at risk” studentsoptimize lesson planning or automate routine administrative decisions. Other uses are more visible. Students use AI tools to summarize and study, instructors use them to create assignments and programs, and researchers use them to write code, scan literature, and compress hours of tedious work into minutes.
People can use AI to cheat or skip tasks. But the many uses of AI in higher education, and the changes they portend, raise a much deeper question: As machines become more capable of doing the work of research and learning, what happens to higher education? What is university for?
Over the past eight years, we have explored the moral implications of a pervasive commitment to AI as part of a joint research project between the Center for Applied Ethics at UMass Boston and the Institute of Ethics and Emerging Technologies. In a recent white paperwe argue that as AI systems become more autonomous, the ethical challenges of its use in higher education increase, as do its potential consequences.
As these technologies become more effective at producing knowledge work – designing courses, writing papers, suggesting experiments, and summarizing difficult texts – they are not only making universities more productive. They risk destroying the learning and mentoring ecosystem on which these institutions are built and depend.
Non-autonomous AI
Consider three types of AI systems and their respective impacts on academic life:
AI-based software is already being used in higher education in admissions exampurchase, academic advice and institutional risk assessment. These are considered “non-autonomous” systems because they automate tasks, but a person is “in the loop” and uses these systems as tools.
These technologies can pose a risk to student privacy and data security. They can also be biased. And they often lack transparency in determining the sources of these problems. Who has access to student data? How are “risk scores” generated? How can we prevent systems from reproducing inequalities or treating certain students as problems to be managed?
These questions are serious, but they are not conceptually new, at least in the field of computer science. Universities typically have compliance offices, institutional review boards, and governance mechanisms designed to help address or mitigate these risks, although they sometimes fail to achieve these goals.
Hybrid AI
Hybrid systems encompass a range of tools, including AI-assisted tutoring chatbots, personalized feedback tools, and automated writing assistance. They often rely on generative AI technologiesespecially large language models. While human users set the overall goals, the intermediate steps the system takes to achieve them are often unspecified.
Hybrid systems are increasingly shaping everyday academic work. Students use them as writing companions, tutors, brainstorming partners and explainers on demand. Teachers use them to generate rubrics, write lessons, and design programs. Researchers use them to summarize papers, comment on drafts, design experiments, and generate code.
This is where the “cheating” conversation belongs. As students and faculty increasingly rely on technology for help, it’s reasonable to wonder what types of learning might be lost along the way. But hybrid systems also raise more complex ethical questions.
Eric Lee for the Washington Post via Getty Images
This is a question of transparency. AI chatbots offer natural language interfaces that make it difficult to detect when you are interacting with a human and when you are interacting with an automated agent. This can be alienating and distracting to those who interact with them. A student reviewing material for a test should be able to tell whether they are speaking with their teaching assistant or a robot. A student reading comments on an essay needs to know whether it was written by their instructor. In such cases, any less transparency would be alienating to all involved and would shift the focus of academic interactions from learning to the means or technology of learning. Researchers at the University of Pittsburgh have shown that these dynamics give rise to feelings of uncertainty, anxiety and distrust for students. These are problematic results.
A second ethical question concerns responsibility and intellectual credit. If an instructor uses AI to write an assignment and a student uses AI to write an answer, who is doing the assessment and what exactly is being assessed? If comments are partly machine-generated, who is responsible when they mislead, discourage, or incorporate hidden assumptions? And when AI contributes substantially to the synthesis or writing of research, universities need for clearer standards regarding authorship and responsibility – not only for students, but also for teachers.
Finally, the crucial question of cognitive unloading arises. AI can reduce the drudgery, and that’s not bad in itself. But it can also distance people from the skill-building aspects of learning, like generating ideas, tackling confusion, revising a clumsy draft, and learning to spot your own mistakes.
Autonomous agents
The biggest changes could come from systems that look less like assistants and more like agents. Even if truly autonomous technologies remain a dream, the dream of a researcher “in a box” – an agentic AI system capable of carrying out studies autonomously – is becoming more and more realistic.
NurPhoto/Getty Images
Agent tools should “free up time” for work which focuses on more human abilities like empathy and problem solving. In education, this may mean that professors can still teach broadly, but more of the day-to-day work of teaching can be outsourced to systems optimized for efficiency and scale. Likewise, in research, the trajectory is moving towards systems capable of increasingly automating the research cycle. In some areas it already looks like robotic laboratories that operate continuouslyautomate large parts of the experimentation and even select new tests based on previous results.
At first glance, this may seem like a welcome productivity boost. But universities are not information factories; they are systems of practice. They draw on a pool of graduate students and early career scholars who learn to teach and do research by participating in this same work. If autonomous agents take on more of the “routine” responsibilities that historically served as an on-ramp to academic life, the university could continue to produce courses and publications while quietly reducing the opportunity structures that support expertise over time.
The same dynamic applies to undergraduates, although in a different register. When AI systems can provide explanations, outlines, solutions, and study plans on demand, the temptation is to offload the hardest parts of learning. To the industry pushing AI into universities, it may seem like this type of work is “inefficient” and that students would be better off letting a machine do it. But it is the very nature of this struggle that builds lasting understanding. Cognitive psychology has shown that students grow intellectually by doing the work of drafting, revising, failing, retrying, tackling confusion, and revising weak arguments. It’s the work of learning how to learn.
Taken together, these developments suggest that the greatest risk posed by automation in higher education is not simply the replacement of particular tasks by machines, but also the erosion of the broader ecosystem of practices that have long supported teaching, research, and learning.
An uncomfortable inflection point
So what are universities for in a world where knowledge work is increasingly automated?
One possible answer is to consider the university above all as an engine for the production of diplomas and knowledge. There, the central question is: do students obtain diplomas? Are articles and findings generated? If autonomous systems can deliver these results more effectively, then the institution has every reason to adopt them.
But another response views the university as something more than a production machine, recognizing that the value of higher education lies partly in the ecosystem itself. This model attributes intrinsic value to the set of opportunities through which novices become experts, to the mentoring structures through which judgment and responsibility are cultivated, and to the educational design that encourages productive struggle rather than optimizing it. Here, what matters is not only whether knowledge and credentials are produced, but also how they are produced and what kinds of people, capacities and communities are formed in the process. In this version, the university is supposed to serve nothing less than an ecosystem that reliably trains human expertise and judgment.
In a world where knowledge work itself is increasingly automated, we believe universities must ask themselves what higher education owes to its students, its early career researchers, and the society it serves. The answers will determine not only how AI is adopted, but also what the modern university will become.