Artificial intelligence is no longer just a matter of academic fraud. It is forcing universities to redefine what they are truly certifying: knowledge, a method, or the ability to exercise judgment.
When the student has already walked through the door
The shift did not take place on corporate boards. It began in dorm rooms, libraries, and late-night conversations with ChatGPT or Claude—places where people ask a machine to explain a theory, summarize a text, rephrase an argument, or write a line of code.
More than 80% of students now reportedly use generative AI tools on a regular basis. The figure itself is less important than what it reveals: the use of AI is no longer marginal. It often precedes institutional policy. Universities thus find themselves in an uncomfortable but fruitful situation: they can no longer decide whether AI will enter higher education. They must decide what role to assign it.
At Rice University, the shift is telling. Three years ago, the use of AI was prohibited. Two years later, it became tolerated, provided it was disclosed. Today, the university encourages it, even providing small grants to fund the redesign of courses to incorporate these methods. This shift is not a capitulation. It is the acknowledgment of a fact: employers already expect graduates to know how to work with AI.
A degree can no longer serve merely as proof of the answer
The most sensitive issue isn’t cheating. It runs much deeper. If a machine can produce a correct essay, a convincing summary, or functional code, what does traditional assessment actually measure anymore?
For decades, universities have often conflated the evidence of work with learning itself: the submitted paper, the grade earned, the passed exam. But AI makes it impossible to maintain this confusion. An assignment can be flawless without the student having actually thought about it. The statement is blunt, but true: to write is to think. To delegate writing entirely is sometimes to delegate the very act of reasoning itself.
This is where the real divide lies—not between institutions that allow AI and those that reject it, but between those that incorporate it as a shortcut to production and those that use it as a tool for discernment. In a first-year writing course at Rice, students compare several versions of the same essay: one generated by ChatGPT, one by Claude, one written by themselves, and one based on a simple outline. The exercise isn’t about hiding the AI, but about examining it. Where does it simplify? Where does it flatten the text? Where does it invent a fluidity that resembles thought, without always having the depth of thought?
Perhaps this is where the university rediscovers its primary purpose: no longer to protect students from the machine, but to teach them not to be intimidated by it.
Details
Rice University receives approximately 40,000 applications for a class of about 1,400 students. Despite the temptation to automate the admissions process, the university maintains a three-person review of each application and conducts interviews with nearly half of the applicants. In a world obsessed with optimization, this choice speaks volumes: some decisions do not necessarily become more accurate simply because they are made faster.
The new rare ability: Judgment
The core competency of the “augmented” student will not be merely mastery of a tool. It will be the ability to evaluate the tool’s output. What AI produces is often plausible, sometimes brilliant, and frequently inadequate. But one must still understand why.
Tomorrow’s graduates will therefore need to possess a dual perspective: enough knowledge to question the machine, and enough method to avoid following it blindly. This is a higher standard, not a lower one. AI does not eliminate the need for knowledge; it makes the gap more visible between those who know how to ask questions and those who are content merely to receive information.
This distinction risks becoming as much a social one as an academic one. Students who receive the best support will learn to orchestrate, compare, critique, and simulate. The others will use AI as a crutch: fast and well-formulated, but intellectually stunting. Higher education must prevent AI from becoming a catalyst for cognitive inequality.
College doesn’t just train workers
Another tension runs through the debate. AI is pushing universities toward employability: optimized résumés, mock interviews, career counseling, short-term certifications, and closer ties with businesses. In the United States, three- to twelve-month micro-certifications are already experiencing rapid growth, particularly in cloud and technical skills. AI will further accelerate this trend.
But it would be a mistake to reduce the university to this single function. It is not just about getting a job. It also teaches students how to live with others, how to advocate for an idea, how to change their minds, and how to become part of an intellectual community. That is why the residential experience retains special value: at Rice, 75% of students live on campus, in a system where community life is just as important as class time.
The machine can tailor its explanations. It can assist in the creation of educational content. It can help a student understand a concept in their own words. But it cannot replace the friction of a seminar, the discomfort of a challenge, or the slow process of developing one’s own thoughts.
Protecting What Cannot Be Delegated
By 2035, the university will likely have adopted AI in its courses, services, counseling, and operations. The question is not whether it will do so, but to what extent.
We’ll need to experiment—a lot. Train teachers. Revise assessments. Integrate AI into the curriculum rather than leaving it in the blind spots. But we’ll also need to draw clear boundaries: academic freedom, admissions, bias, sensitive decisions, and the development of judgment.
Universities change slowly, until the moment comes when they no longer have a choice. That moment has arrived. The risk is not that students will use a machine to learn. The risk is that they will cease to distinguish between learning and producing.
The value of a degree may soon depend on this promise alone: not to certify that one knows how to answer, but that one still knows how to think before answering.
Cette publication est également disponible en :
