A student submits an elegant essay but cannot explain its central argument. A teacher produces a lesson plan in seconds but has not checked its examples. A school adopts an assessment tool without knowing where children’s data will be stored. Each case shows the same problem: technological convenience can conceal educational failure. Pakistan is no longer deciding whether artificial intelligence should enter its classrooms. Punjab announced mandatory AI use in schools in January and has since promised training for 160,000 teachers. The Higher Education Commission has made an AI course compulsory across undergraduate and postgraduate degree programmes from this autumn and developed a draft framework on generative AI in universities. The harder question remains: when does its use strengthen learning and respect the learner?
AI offers possibilities worth pursuing. It can suggest explanations, generate practice exercises and help teachers differentiate material. But a system that completes an assignment may save time while removing the intellectual work the assignment was supposed to teach. The word that should govern the answer is judgment. A language model generates text by predicting likely sequences of words; fluency alone does not establish truth or understanding. Judgment is the capacity to decide what counts as a good reason, to notice when an answer is fluent but wrong, and to take responsibility for a conclusion. A machine can assist with evaluating reasons and identifying errors, but responsibility remains with people and institutions. Education exists, in part, to form that capacity. Any use of AI that leaves it unformed, however impressive the output, has failed at one of education’s central purposes. The governing principle, then, is the protection of independent thinking. Asking a chatbot to explain why an argument is weak can support learning; submitting its replacement argument as one’s own conceals authorship and understanding. The measure of success is what learners can subsequently explain, apply and produce for themselves. This requires clear rules, and here the HEC draft makes a sensible beginning. It asks universities to distinguish permitted, restricted and prohibited uses of AI, communicate expectations through course outlines, and require disclosure of AI assistance. It also addresses fabricated references and invented data as matters of academic misconduct. Students deserve to know these distinctions before assessment begins, and teachers should be equally transparent when AI contributes to marking.
The draft also recognizes a safeguard every university committee should observe: an AI-detection report alone should not determine a student’s guilt. Consider why. In 2023, Stanford researchers tested seven detectors on 91 TOEFL essays written by non-native English speakers. Across the detectors, the average false-positive rate was approximately 61 per cent, while essays by American eighth-graders were classified with much greater accuracy. The findings concern tools tested at that time, rather than every detector available today. Nevertheless, they demonstrate the danger of treating a software score as proof. Many Pakistani students write academic English as an additional language. Reliance on detection software risks penalizing honest students whose writing follows predictable language patterns. An accusation must rest on evidence a student can answer: drafts, a conversation about the work, an explanation of the sources.
Assessment must, therefore, change. Where marks reward polished reproduction, outsourcing becomes attractive. Drafts, classroom writing and oral explanations reveal the development of understanding. They cost time; a lecturer with three hundred students cannot hold three hundred vivas without institutional support, and that cost must be budgeted rather than wished away. Privacy deserves equal attention. Where schools introduce AI platforms, they must establish what student data those platforms collect. Teachers should not enter identifiable student records or confidential assessments into unapproved services. Institutions should establish what a provider collects, how long it keeps it and whether it trains on submissions. Families need understandable explanations.
Fairness extends beyond access to a device. A policy announcement does not ensure classroom provision; the distance between a circular issued in Lahore and a classroom in Rajanpur is measured in electricity, connectivity and trained staff. Schools should ask whether a tool works in Urdu and the regional languages, and whether paid features buy an advantage that assessment mistakes for ability. Wherever AI use is mandated, an accessible alternative must exist. Accuracy remains a professional obligation. AI produces convincing statements and references that require verification, especially on history, religion or contested public questions. One classroom exercise makes the point: ask learners to examine an AI response, locate two supporting sources, identify an unsupported claim, then rewrite the answer and explain their changes. Teacher development should extend beyond prompt-writing to include verification, assessment design and the protection of independent thinking.
The HEC draft envisages institutional oversight and mechanisms for addressing misuse. Their value will depend on whether universities investigate concerns fairly, resolve complaints promptly and explain their decisions. A committee should be judged by its response to a student’s complaint as well as by its reports. Decisions about grades or discipline should remain subject to meaningful human review. The ethical use of AI will ultimately be tested in ordinary classroom choices. Does the teacher verify an explanation before presenting it? Can the student defend the submitted argument? Can a family question how its child’s data is handled? Pakistan has welcomed the innovation; it must now make the conditions for responsible use clear and workable. The aim should be students who use intelligent systems with informed judgment and retain the confidence to think, question and learn independently. Developing that capacity must remain central to education.
—The writer is Commoner from 44th Common Educationist — Founder of WHI Institute.based in Sargodha.
