I was never very good at exams. I do not mean to say I never studied, or that I disliked learning. In school and college, my notes had a habit of disappearing on the eve of exams. Once they were simply stolen. The people who borrowed or stole them often scored better than I did.What I struggled with was the exam situation itself.I still remember one school exam where I had entered what athletes might call “the zone”. I was writing quickly, almost automatically, moving from one answer to another without overthinking. Then a teacher stopped near my desk and began watching me write. I froze. Completely. After a while, she announced to the class that students could not simply “freeze” during exams like this.That did not help.Illustration: Pariplab Chakraborty.Over time I became increasingly bad at high-pressure, timed examinations, especially in environments where I felt watched, judged or overwhelmed. During my MA, I often struggled with the mathematics and statistics portions of the programme. I collapsed in several exams. Ironically, after graduation I went back and taught myself much of that material independently, eventually developing a highly quantitative PhD proposal that I later abandoned because I realised I was more interested in political economy, history, ethnographic observations and older traditions of economic thought.For a long time, I treated this as a personal inadequacy. But I increasingly wonder whether universities often confuse success within a particular evaluative format with intelligence itself.Is AI ‘destroying’ education?That question returned to me repeatedly over the past two years as universities have become consumed by anxiety over artificial intelligence.Much of the public conversation assumes there was once a clear distinction between “authentic” intellectual work and technologically mediated work, and that AI has suddenly blurred this boundary. I am not sure the distinction was ever so clear.When I wrote my MA dissertation about 15 years ago, I remember being deeply uncertain about which ideas were “mine” and which emerged from months of immersion in books, lectures, conversations and notes. I read obsessively at the time. Sentences, arguments and turns of phrase from many sources accumulated in my head faster than I could disentangle them. Had today’s AI-detection systems existed then, I suspect parts of that dissertation may well have been flagged despite being written entirely by me.That uncertainty feels important today because universities increasingly respond to AI not by rethinking evaluation itself, but by intensifying surveillance.Proctoring AI using AI proctoringThis semester, faculty in my own institution spent hours collectively trying to interpret the logic of online AI-assisted proctoring systems during examinations.Students were asked to perform 360-degree room scans while also being permitted to use phones to upload handwritten answers. Faculty debated whether looking down too frequently constituted suspicious behaviour even when students were solving mathematical problems on paper. One student was flagged for “disappearing” because of the lack of adequate lighting in the room. Another for standing up repeatedly. Another because someone briefly entered the room with a charger. The exam environment itself was filled with anxiety, suspicion, bordering on the absurd.But I found it interesting that this environment didn’t feel too alien because while the technologies were new, the underlying reasons for these anxieties were not. I had seen versions of this long before AI entered classrooms. I knew students who understood the material deeply but performed badly in exams. I knew others who were excellent at figuring out what teachers wanted from them. I knew people who became articulate only outside classrooms, after the pressure of evaluation disappeared. I also knew students who flourished within formal academic structures and went on to do excellent work later in life. What counted as a “good student” was never as obvious as universities often pretend it is.What exactly are we evaluating when we evaluate students?In retrospect, I am not even sure that my confusion about authorship during my MA dissertation was unusual.Most serious reading changes the way one thinks and writes. Certain phrases stay in your head. So do rhythms of speech, argumentative habits and ways of framing ideas. Over time, these influences become difficult to disentangle from what we call our “own” thinking.I remember once using the word “decent” in an email to describe something I was actually quite satisfied with. I suspect I had absorbed that usage from years of watching English football interviews. But someone senior interpreted it as a sign of dissatisfaction or passive aggression. The same word had traversed through entirely different worlds of meaning before arriving in that exchange.The panic around ‘authenticity’Part of the current panic around AI seems to emerge from a much older anxiety within education systems — the fear that institutions may not actually know how to reliably recognise learning, originality or intellectual effort.When I teach, I sometimes notice students using phrases that clearly originated in classroom discussions. Occasionally, they repeat an analogy or formulation almost exactly as I once used it. Part of me feels proud that they were paying attention closely enough to remember it. Part of me also feels slightly embarrassed because I had not realised that an offhand remark I made while teaching on four hours of sleep and several cups of coffee would become, for someone else, an important way of understanding a concept. But that is also how intellectual life often works. We carry fragments of other people’s language, arguments or humour with us long after we encounter them. Sometimes we consciously cite them. Sometimes we no longer remember where they came from.Traditional plagiarism systems at least attempted to identify direct overlaps with identifiable existing sources. AI-detection systems operate differently. They often make probabilistic judgments about whether writing resembles machine-generated text. That is a far odder claim. The software is now attempting to infer how a piece of writing came into existence. And that uncertainty has started producing strange consequences inside universities.Recently, literary circles found themselves debating whether submissions to the Commonwealth Short Story Prize may have been AI-generated. Universities across the world are facing similar disputes in classrooms with respect to essays, dissertations and examinations. Students now encounter situations where they are asked not just to submit work, but somehow prove the authenticity of their thinking process itself.In my own institution, this surfaced during disagreements around evaluations of undergraduate dissertation proposals. Faculty members disagreed sharply over how much weight should be given to software-generated assessments of originality. Some students had clearly used AI tools in some capacity. Others insisted they had written independently. But what interested me was not simply whether AI had been used. It was the assumption that these disputes could be cleanly solved through surveillance, detection systems or blanket prohibitions.This semester, I experimented with requiring students to disclose how they had used AI in assignments. The intention was less to “catch” students than to force some honesty about practices that were already becoming normalised within academic life. I also wanted students to reflect on how dependent they were becoming on tools that increasingly seemed impossible to avoid.What surprised me most this semester was not that some students were using AI extensively. It was that one student who refused to use it at all. They told me quite directly that they did not want to become dependent on these systems while learning. But they also admitted something else. They increasingly felt that they were at a disadvantage. Other students were using LLMs to summarise readings, generate notes, brainstorm assignments and prepare faster for exams. Meanwhile, they were still trying to read everything themselves while also keeping up with coursework, internships, applications and also the general pressure to remain competitive inside a rapidly intensifying university environment.Listening to this, I found myself thinking about my own years as a student. Is this situation entirely new? Students have always navigated unequal educational worlds. Some had access to coaching centres and private tutoring. Some inherited old notes, solved papers and exam strategies from seniors. Some were simply better at guessing what teachers wanted to hear. Some were calmer under pressure. Some came from backgrounds where university life was already familiar. Others entered these spaces constantly unsure whether they truly belonged there. Some genuinely loved reading and thinking but froze during exams. Others barely engaged with the material but became highly skilled at navigating the system around it.AI did not suddenly create these uneven relationships to learning. What it seems to have done is expose how naive many of our older assumptions about evaluating merit and original thinking already were.Rohith Jyothish is assistant professor at the Jindal School of International Affairs, O.P. Jindal Global University, Sonipat and academic dean of the Master’s in diplomacy, law & business programme.This piece was first published on The India Cable – a premium newsletter from The Wire – and has been updated and republished here. To subscribe to The India Cable, click here.