AI forces universities to rethink term papers: the cheating problem predates ChatGPT

The appearance of ChatGPT and other generative models did not create the problem of dishonest university work from scratch. According to Dr. Nafisa Baba-Ahmed, AI merely made mainstream and convenient a method of "outsourcing thinking" that had existed in universities long before the current boom.

The Problem Is Not New

Teachers' frustration with AI's impact on critical thinking is easy to understand. Many universities are already having an anxious conversation about how to preserve students' independent work when a bot can quickly assemble an essay, argument structure, and even polished academic style. But Baba-Ahmed suggests looking not only at new technology but at the old assessment model, in which the final text could long be obtained not entirely through one's own intellect.

In her work with students on academic writing, she essentially describes a familiar picture from before the era of generative AI. When the assignment format allows it, thinking begins to be outsourced: some commission essays, others use archives from previous years, still others pass successful model answers between cohorts, fourth use tutors, friends, or senior colleagues too heavily. That is, the problem is not that AI corrupted an ideal system, but that it arrived in a system with already open loopholes.

How AI Accelerated It

The main change, in her view, is not in the nature of the violation itself, but in its scale. Previously, such a shortcut required money, connections, time to search, or a ready template from a previous cohort. Now much can be obtained in minutes: generate a draft, ask the model to restructure the argument, polish the style, or adapt old text for a new assignment. AI did not invent the workaround, but sharply lowered its cost and entry threshold.

"AI didn't invent this behavior.

It merely streamlined an already existing way of cutting corners."

Because of this, the debate about whether "AI kills thinking" becomes too narrow. If a university assesses primarily the smoothness of the final text rather than the independence of reasoning, the technology simply makes the old vulnerability more visible. The problem manifests not because students suddenly stopped thinking, but because the system too often accepted a well-formatted result as evidence of genuine understanding. This is exactly why the current panic looks not like a new disease, but a late diagnosis.

What Universities Should Check

From this emerges a more uncomfortable but also more useful question for universities: what exactly should a coursework, essay, or other written work demonstrate. If the task is to see whether a student understood the material, can connect ideas, build an argument, and draw conclusions, a neat text alone may not be enough. AI simply forces this to be stated aloud and stops pretending that the previous format reliably measured everything on its own.

Baba-Ahmed suggests not romanticizing the pre-AI past. Instead of trying to "return to how it was," universities will have to redefine what exactly they want to see in student work and where the boundary lies between tool assistance and substitution of original thought. In practical terms, this means reviewing not only the rules but the very logic of the assignment.

It is no longer about cosmetic bans, but about a more honest answer to the question of what exactly counts here as evidence of learning.

• Understanding of the topic, not just polished presentation
• Independent argument, not assembly from ready-made phrases
• Ability to analyze, not rewrite a template
• Personal coherence of reasoning, not flawless tone
• Actual demonstration of thinking, not just the final file

What This Means

For universities, AI turned out to be not so much the root cause of a crisis as a stress test for old assessment methods. It showed that the question is no longer about banning the next tool, but about which assignments truly require independent thinking and allow distinguishing genuine understanding from neatly assembled imitation.