The End of the Assignment
Why exams, essays and grading no longer measure what we think they do
The illusion of authorship
The assignment still looks the same. It takes the form of an essay, a report, a carefully structured argument. It is submitted, read and graded as it always has been. On the surface, nothing appears to have changed. Except that everything has.
Students now work with systems that can generate structure, refine language and suggest arguments in seconds. This is no longer incidental or experimental; it is becoming part of the normal workflow. And yet, universities continue to treat the final product as evidence of individual understanding.
That assumption is no longer stable.
What are we actually grading?
For decades, assessment relied on a relatively simple premise: that output reflects effort and effort in turn reflects understanding. This chain has quietly been broken.
A well-written essay may still signal deep comprehension. But it may just as easily reflect partial understanding, combined with effective prompting and careful refinement. From the outside, these distinctions are increasingly difficult—if not impossible—to make.
The issue is not plagiarism. It is opacity. We no longer see how the work is produced. And without insight into the process, the meaning of the output becomes uncertain.
The vulnerability of take-home assessment
Take-home assignments were designed for a different academic reality, one in which access to sources required time, structuring an argument required effort and writing itself was inseparable from thinking. Artificial intelligence compresses all three.
What once demanded sustained engagement can now be partially externalised. Research becomes retrieval, structuring becomes suggestion and writing becomes generation. The assignment still exists, but the conditions that once gave it meaning have changed.
This is not a disciplinary problem. It is a structural one.
From assistance to substitution
There is a tendency to describe AI as a tool, but that description understates the shift that is taking place. A calculator assists calculation, yet leaves the reasoning intact. AI, by contrast, can begin to substitute elements of reasoning itself.
It does not merely help students articulate ideas; it can help generate them. This creates a grey zone that universities have not yet clearly defined: where does assistance end and where does substitution begin?
Students, however, are already navigating this space in practice.
The paradox of delegating intelligence
Something more subtle is unfolding beneath the surface. Students are increasingly learning to delegate cognitive tasks before they have fully mastered them themselves. They are becoming managers of processes before they have become practitioners of the craft.
To delegate reasoning before learning how to reason is to build on unstable ground. It creates a paradox in which students can orchestrate intelligent output without necessarily possessing the underlying capability.
We risk educating a generation that can direct intelligence, but not produce it independently—architects who have learned to design, but not to lay a single stone.
The return of the oral
As written output becomes less reliable as an indicator of understanding, older forms of assessment begin to reappear. Oral examinations, live defenses and in-room problem solving are returning—not out of nostalgia, but out of necessity.
These formats restore something that has gradually been lost: visibility. In an oral setting, reasoning unfolds in real time. Hesitation becomes meaningful. Uncertainty can be explored rather than concealed.
What is assessed is no longer only the answer, but the thinking that produces it. This is not a step backwards. It is a structural adjustment to new conditions.
The luxury of process
Learning has never been an efficient activity. It depends on friction—the slow construction of an argument, the rewriting of a paragraph, the uncertainty that precedes clarity.
Artificial intelligence reduces much of this friction. It offers structure before confusion has fully emerged and clarity before the underlying struggle has taken place. Yet it is precisely in that struggle that understanding is formed.
When students arrive too quickly at an answer, it becomes harder to know whether they have truly arrived at understanding. What is lost is not time, but depth.
The electric step
To understand the shift, consider a simple analogy.
Imagine a race designed to measure endurance. The goal is not speed, but the capacity to run. Yet participants are allowed to use an electric step, as long as it is framed as “support”.
At the finish line, we record times and celebrate the fastest competitors. But we no longer know who actually ran the distance.
This is the position of the modern assignment. It measures performance without visibility into effort and output without certainty about capability.
Synthetic competence
This dynamic produces what might be called synthetic competence: the ability to generate convincing performance without fully internalising the underlying knowledge.
Students can produce coherent essays, structured arguments and technically sound analyses, while still relying on external systems for key elements of reasoning. This is not necessarily dishonest, but it does create a gap between appearance and reality.
For universities, this is a critical risk. They may begin to certify competence that has not been fully developed.
Academic integrity after AI
Traditional frameworks of academic integrity were built around clear violations such as plagiarism, copying or unauthorised collaboration. Artificial intelligence does not fit neatly into these categories.
Its use is often partially permitted, sometimes encouraged and frequently invisible. As a result, integrity shifts from a matter of prohibition to a matter of responsibility.
Some institutions are beginning to respond by requiring transparency in AI use, including the documentation of prompts and reflection on process. Yet these measures only address part of the problem.
The deeper issue is that authorship itself has become more difficult to define.
The institutional dilemma
Universities now face a fundamental choice. They can attempt to restrict the use of AI, which is increasingly difficult to enforce and risks becoming disconnected from reality. They can ignore the shift, which is the easiest path in the short term but gradually undermines credibility.
Or they can redesign assessment from first principles.
Only the third option is sustainable, but it requires more than incremental change. It demands a reconsideration of what is actually being measured.
From product to process
If the final output can no longer be trusted as evidence of understanding, the focus of assessment must shift from product to process. The central question is no longer whether an answer is correct, but how that answer was produced.
This implies a different kind of evaluation, one that prioritises explanation, reflection and responsibility. Students must be able to articulate their reasoning, recognise the limits of AI-generated output and take ownership of the results they present.
These qualities are more difficult to measure, but they are closer to the core purpose of education.
Europe’s moment
In this transition, Europe occupies a distinctive position. It does not lead in the scale or speed of AI development, but it has a strong tradition of institutional accountability and public standards.
This creates an opportunity to redefine what credible assessment looks like in an AI-mediated environment—not by rejecting technology, but by embedding it within a framework of transparency and responsibility.
A system under examination
The assignment is not disappearing, but its function is changing. It can no longer serve as a reliable proxy for understanding. Instead, it becomes part of a broader process—an entry point into discussion, a basis for questioning, a moment within a larger trajectory of learning.
The real assessment shifts elsewhere: into explanation, into defense, into judgment.
The uncomfortable conclusion
The traditional essay has quietly become something else.
The traditional essay is quietly becoming a Turing test that higher education is increasingly failing. If institutions can no longer determine where the machine ends and the student begins, they are no longer certifying knowledge, but process management.
If we can no longer determine where the machine ends and the student begins, we are no longer measuring learning. We are measuring coordination.
If we can no longer trace the origin of thought, we can no longer claim to certify it.
This article is part of the series The University After AI, published in the Culture & Education section of Altair Media.
Photo by Vitaly Gariev on Unsplash
A student writing mathematical formulas on a blackboard—capturing a form of thinking that becomes harder to observe as AI increasingly shapes academic output.
