Before students use AI, they should prove they don’t need it

Key points:

• What, exactly, does a degree certify in the age of AI?
• The real work of AI and instructional technology is creative
• Reflections on Magnifica Humanitas for college and university instructors
• For more news on students and AI, visit eCN’s AI in Education hub

Universities are attempting to adapt to artificial intelligence while considering mostly the wrong questions.

Should AI be banned? Should students disclose use? How much AI use is acceptable? Are students abusing it? Should institutions purchase better detection software? Should they stop using detection software?

While these questions dominate the discussion, the more important objective is making sure students can meaningfully function without AI.

For most of history, performance was treated as evidence of competence. A strong essay suggested strong writing skills. A persuasive presentation suggested communication ability. A successful research project suggested analytical capability. The work itself served as proof. Outside assistance existed, but it was limited, unevenly distributed, and rarely capable of performing the task on the student’s behalf. Tutors, editors, and even ghostwriters could influence major projects, but few students had access to instant assistance for every reading assignment, homework exercise, draft, presentation, classroom activity, or moment of confusion.

Artificial intelligence has broken a premise that universities have relied upon for centuries: the assumption that strong performance usually reflects genuine competence. The implications extend beyond assessment. If universities can no longer confidently distinguish performance from competence, they risk exacerbating challenges that already exist. Employers have long complained about graduates who possess credentials but lack practical skills. Students increasingly question the value of expensive degrees and uncertain career prospects. Grade inflation has raised doubts about what academic distinctions actually signify. In such an environment, a diploma must represent more than completed coursework. It must continue to serve as a credible signal that its holder possesses knowledge, judgment, skills, and the ability to perform independently. If artificial intelligence weakens that signal, universities may find themselves confronting not merely an assessment problem, but a new legitimacy problem: what exactly does a degree certify in the age of AI?

Today, students can produce polished essays, presentations, reports, and summaries without necessarily possessing the underlying skills those assignments were designed to develop. The output may be excellent. The student’s capabilities may be much less of a given.

This is also not primarily a cheating problem–it is a development problem.

Many discussions about AI focus on academic integrity. While important, this emphasis risks overlooking a more fundamental challenge. Even when students use AI entirely within institutional rules, they still most likely bypass the struggle through which learning often occurs.

A student who uses AI to improve an argument is learning. A student who uses AI because they cannot construct the argument in the first place may not be.  A student who uses AI to explain an unfamiliar concept is learning. A student who uses AI to complete an assignment without understanding the concept, or the assignment itself, is merely submitting. A student who uses AI to identify weaknesses in a draft is engaged in the process. A student who uses AI to generate the draft from scratch may never develop the underlying skill at all. The distinction matters.

Universities remain concerned with detecting AI, even as the distinction becomes increasingly meaningless in many practical contexts. Outside the classroom, few employers care whether AI helped draft an email, summarize a report, or organize information. What matters is whether the individual understands the content, can evaluate it critically, and can function without the tool when necessary. Universities should perhaps focus less on detecting artificial intelligence and more on teaching, developing, and verifying competence.

Universities have never been primarily in the business of information. Even before artificial intelligence, most graduates eventually forgot much of the factual content they memorized for examinations. What remained were the habits and capabilities developed along the way: analytical thinking, communication, research, judgment, time management, attention to detail, professional discipline, resilience, and the ability to learn unfamiliar material independently. The cliché is largely true. The most enduring product of higher education is not knowledge itself, but the intellectual and professional habits acquired while pursuing it.

Years ago, while grading a student competition, I came across an unusual observation: Stress is to students what salary is to employees. At first glance, it sounded absurd. Yet there is a deeper truth behind it. Universities do not merely reward students with grades. They expose them to deadlines, expectations, criticism, uncertainty, and responsibility within a relatively controlled environment. Students learn not only accounting, chemistry, history, or law. They learn how to function when under pressure. Resilience is not an accidental by-product of education. It is one of its intended outcomes.

The same can be said of ideas. Most graduates do not remember every lecture, chapter, or examination question. Yet many remember a particular argument, an uncomfortable realization, a moment when a long-held assumption fell apart, or a conversation that changed how they viewed the world. These intellectual turning points rarely arrive fully formed. They emerge through reading, discussion, confusion, disagreement, reflection, and struggle. Ideas become meaningful not when they are delivered, but when they are wrestled with. That process becomes much harder when every moment of uncertainty can be instantly outsourced to a machine, allowing the answer to arrive before the understanding that normally produces it.

Rather than focusing exclusively on what students produce, institutions may need to pay greater attention to what students can independently do.

Universities often speak as though artificial intelligence has created an unprecedented assessment crisis. It has not. The underlying problem is ancient: how do you ensure that the work being evaluated belongs to the person receiving the credit?

Imperial China addressed the problem by locking candidates alone in examination cells and making them produce the work themselves. Modern universities possess countless equivalents: oral defenses, debates, presentations, handwritten examinations, open ended exam questions, case discussions, and live questioning. We already know how to test competence. What we lack is not methodology. We lack the willingness to accept the cost, inconvenience, and complexity that genuine, daily verification requires.

The solution is introducing a “Human Intelligence Certificate” before the start of every degree program.

The certificate would not demonstrate proficiency in artificial intelligence. Most likely almost everyone will soon possess that. Instead, it would demonstrate AI independence: the ability to perform essential academic and professional tasks without AI assistance. Students who earn it would demonstrate that they can write, analyze, research, evaluate sources, solve problems, and communicate effectively without relying on artificial intelligence.

Students might be required to write under supervised conditions, participate in oral examinations, defend arguments in discussions, conduct independent research, evaluate sources, and solve problems without technological support. Only after demonstrating competence would they be permitted, and maybe even encouraged, to integrate AI into their work more freely.

Most certifications exist to confirm competence. Lawyers must pass examinations before practicing law. Doctors must demonstrate proficiency before treating patients. Drivers must show they can operate a vehicle safely before receiving a license. A Human Intelligence Certificate would be different.

Rather than certifying a student’s ability to use a tool, it would certify that they can function without it. Before relying on AI to write, analyze, summarize, or solve problems, students would first demonstrate that they can perform those tasks independently.

The certificate would not prove that students know how to use artificial intelligence. It would prove that they do not depend on it. Before relying on the intellectual equivalent of autopilot, students would first demonstrate that they can fly the aircraft themselves.

Why should students be allowed to outsource a skill before demonstrating possession of that skill?

The best use of AI is as a shortcut–but only after the underlying skill already exists. It can learn my writing style and draft an email, identify weaknesses in an argument, suggest improvements, highlight unusual or overlooked ideas, summarize long texts, organize information, and help me work more efficiently. In these situations, AI accelerates tasks I am already capable of doing myself. It saves time without replacing understanding. The danger arises when the shortcut becomes a substitute for the skill rather than a tool that enhances it. A shortcut is valuable because it shortens a journey we know how to complete; it becomes a problem when it replaces the journey altogether.

The strongest AI users are rarely those who depend on AI. They are those who can challenge it. They can recognize weak arguments, identify errors, verify sources, and detect hallucinations. They understand enough to know when the machine is wrong. In other words, they are precisely the people who do not actually need AI in order to function.

Employers have spent decades complaining that graduates possess credentials but lack competencies. Artificial intelligence risks widening that gap. The challenge facing universities is not merely producing impressive work. It is producing capable graduates.

The COVID-19 pandemic offered a useful lesson. For years, educational technology advocates imagined a future in which physical barriers to learning would largely disappear. Students could learn from anywhere. Lectures could be streamed. Information could be accessed instantly. Then the world was forced to test that vision at scale. The results were mixed.

While technology expanded access and flexibility, it did not automatically create discipline, motivation, engagement, or intellectual curiosity. Many students struggled despite having unprecedented access to information. The experience served as a reminder that access is not the same as education, and technology is rarely the great equalizer it is claimed to be. The students who benefited most were often those who already possessed the habits, discipline, and skills necessary to take advantage of it.

The experience reminded us of something important about education.

Education is not merely the transfer of information, if it were, universities would have disappeared long ago. Learning depends upon structure, accountability, challenge, discussion, mentorship, and human interaction. Artificial intelligence faces the same limitation. It can provide information, explanations, and feedback. It cannot eliminate the need for intellectual effort.

The greatest educational change brought by artificial intelligence may not be that students have answers. It may be that students are no longer forced to sit alone with questions.

For most of human history, thinking was often a solitary activity. Students wrestled with uncertainty, confusion, and incomplete understanding. Today, an answer is always seconds away. Yet confusion is not necessarily evidence that learning has failed. It is often evidence that learning has begun.

Indeed, universities may need to move in precisely the opposite direction.

As information becomes abundant, the value of higher education may increasingly lie in cultivating judgment.

We already know that information is no longer the scarce resource – attention is.

We naturally prefer shortcuts. Artificial intelligence may be the most powerful educational shortcut ever invented. The challenge is that shortcuts are valuable only when they shorten a journey we are capable of completing ourselves. A shortcut cannot replace the journey.

The good news is that universities do not need to invent an entirely new educational model. A tried, tested, and remarkably effective one already exists: legal reasoning, particularly the tradition of common law reasoning.

Contrary to popular stereotypes, legal education is not primarily about memorizing rules. It is training in disciplined thinking. Students learn to work from facts to principles, compare situations, distinguish meaningful differences, reconcile competing interpretations, and defend conclusions under scrutiny.

Most professions operate under uncertainty. Lawyers rarely know every fact. Managers rarely possess perfect information. Policymakers rarely understand all consequences. Yet universities often assess students through assignments that imply certainty exists. Legal reasoning is valuable precisely because it teaches students how to think when certainty is unavailable.

The Socratic method pushes this process further. Students are required not merely to provide answers but to explain, defend, and refine their reasoning. Weak assumptions are exposed. Contradictions emerge. Certainty becomes more difficult to maintain.

Artificial intelligence is remarkably good at producing answers. The future may belong not to those who can generate information, but to those who can evaluate, challenge, and refine it.

Many of the challenges facing universities today are not information problems. Students have access to more information than any generation in history. Instead, institutions increasingly face challenges involving attention, engagement, critical thinking, resilience, and independent judgment. In this environment, one of the most valuable educational experiences may be learning how to think through uncertainty rather than merely retrieve information.

The idea is not entirely without precedent. Many universities already offer pre-university intensive programs designed to prepare students for academic life. My own institution operates a Pre-University Intensive English Program (PIEP), which gives incoming Chinese students a short but concentrated introduction to studying, communicating, and functioning in an English-language academic environment. Universities routinely recognize that students benefit from a brief period of preparation before beginning their degree studies. The question is whether the age of artificial intelligence may require a similar introduction, not to English, but to reasoning, analysis, argumentation, judgment, and independent thought. Chinese universities routinely require military training before students begin their studies. In an age increasingly shaped by artificial intelligence, one might reasonably ask whether a short period of intellectual training deserves similar consideration.

A month of law school for everyone

Probably not a course about legal rules. It would be a course about thinking. Students would read dense materials with a pencil in hand, underline key passages, annotate margins, identify assumptions, distinguish facts from opinions, and learn to extract meaning rather than merely consume information. They would brief cases, apply FIRAC and IRAC frameworks, identify legal issues, connect rules to facts, distinguish relevant facts from irrelevant ones, and explain why a small factual difference can produce an entirely different outcome. They would write objective analyses and persuasive arguments, draft memoranda and position papers, debate controversial issues, argue both sides of the same question, defend conclusions under challenge, and evaluate competing interpretations of the same evidence. They would learn to spot logical fallacies, challenge unsupported assumptions, identify weaknesses in reasoning, and separate confidence from credibility. They would engage in Socratic discussions in which every answer invites another question. The goal would not be to teach students law. The goal would be to teach them how to think when information is abundant, answers are cheap, and certainty is unavailable.

Legal education is not without its own shortcomings. The skills rewarded in law school are not always the same as those rewarded at different stages of a legal career. Students may be encouraged to focus on policy, theory, and big ideas, only to encounter detail-oriented document review, drafting, and research early in practice, followed later by management, client development, and business generation. Yet despite these imperfections, what many law schools describe as “lawyering skills” may be unusually well suited to the challenges posed by artificial intelligence.

Students learn to operate within closed universes of information, conduct structured research, distinguish facts from assumptions, move from objective analysis to persuasive advocacy, evaluate competing arguments, identify weaknesses in reasoning, and defend conclusions under scrutiny. Not everyone should become a lawyer. But in an age of abundant information and machine-generated answers, everyone may benefit from learning a little more like one. The model is inspired by legal education, but draws equally from philosophy, rhetoric, history, and the broader liberal arts tradition.

What might such a program look like? Not surprisingly, low-tech. For a month, students would work largely without screens, AI tools, or digital assistance. They would read, write, discuss, debate, annotate, analyze, and defend ideas in person. The goal would not be to reject technology, but to establish a baseline of independent competence. Concentrating this process into a short, intensive period may actually be easier than attempting to police AI use across every class, assignment, presentation, and examination over the course of an entire degree. Students who successfully demonstrate the required skills could then be trusted to use artificial intelligence however they see fit. Students who fall short would receive additional training and another opportunity to demonstrate competence. The objective would not be exclusion. It would be development. The principle is simple: before students are permitted to rely on AI, they should first be certified that they do not need it.

The irony of the AI era is that the students who benefit most from artificial intelligence are often the students who need it least. They already possess the skills necessary to question, verify, refine, and improve its output. For them, AI helps shift effort upward on the pyramid of intellectual work, away from formatting, searching, collecting, organizing, and drafting, and toward analysis, synthesis, creativity, judgment, and decision-making. The students who feel they “need” AI most may be the ones most likely to outsource the very abilities universities exist to develop.

The future of higher education is not teaching students how to use artificial intelligence. That will soon be assumed. The more important challenge may be ensuring that students can still demonstrate judgment, reasoning, resilience, and independent thought when artificial intelligence is readily available.

The original Turing Test asked whether machines could imitate human intelligence. Universities may increasingly face the reverse challenge: proving that humans still can.