Bangladesh's education system, from primary school through the HSC examination, has been built around marks earned through memorisation. GPA-5 has become less a measure of understanding and more a test of how faithfully a student can reproduce a model answer, formula, or paragraph learnt by heart. That approach was already inadequate for preparing young people for work; AI has made it virtually obsolete.
Human thinking operates in two broad modes. The lower-order cognitive mode - memorising facts, understanding them, and applying them to familiar problems - is what curricula, coaching centres, and board exams have spent decades optimising. The higher-order mode - analysing relationships, weighing motives, making ethical judgements, and building new solutions for problems that have no memorised answer - is mentally demanding and largely unrewarded by the system.
AI has advanced disruptively in the first mode. Storing information, retrieving it, and matching patterns are probability exercises, and machines do them faster than any student. A tool that can write a competent English composition or solve a board exam maths problem in seconds doesn't just help a struggling student. It makes the entire skill the system tests for less valuable.
"We are the ideal environment for AI to erode independent thinking," the argument goes. For decades, notebooks and commercial guides have sold Bangladeshi students pre-written answers to memorise. AI automates and personalises the same shortcut at zero cost and infinite scale. The incentive for a student to actually think is disappearing faster here than in most places.
Left unaddressed, this risks producing a generation fluent in prompting AI but not trained to evaluate, question, or build on what it gives them. Psychologists call this "cognitive offloading." Locally, it's a high-tech evolution of the GPA-5 problem - a powerful crutch that perfectly mimics the student's writing style.
Reform assessment before curriculum
The response must not be banning AI in classrooms. The real task is redesigning what students are asked to do, so AI becomes a sparring partner for higher-order thinking rather than a substitute for it. Estonia's "AI Leap" programme, which is rolling out AI tutors nationally at the secondary level, is a useful reference point - but it's built for a system that tests reasoning. Bangladesh's hardly does, and that is precisely where reform must begin.
Assessment must be reformed before curriculum. As long as SSC and HSC exams reward reproduction of memorised answers, no curriculum change will matter. Students and coaching centres will find new ways to game the old test. Introducing components that require analysis, argument, and original reasoning, even gradually, would do more to redirect classroom behaviour than any technology mandate.
Higher-order training should start earlier. Analytical and evaluative thinking is currently treated as something reserved for university research. It should begin in lower secondary school - roughly classes 7 and 8 - through assignments that ask students to weigh motives, debate ethics, or propose original solutions, tasks a chatbot can assist with but not complete.
Teachers are the bottleneck
Bangladesh's AI-in-education conversation has focused heavily on student-facing tools and far too little on teachers. A teacher who cannot design an assignment resistant to copy-paste AI use, or understand what these tools can do, cannot be expected to guide students through it. Any serious policy has to start there. For educators looking to build those skills, AI for Teachers Courses cover assignment design and classroom integration strategies.
The ICT Division and university education departments are better positioned to run flexible, evidence-generating pilots in select districts or grade levels than a ministry bureaucracy is to design and enforce a national policy on a new technology. Those pilots must also be honest about the urban-rural and internet divides. What is learned from them, tracked systematically by universities and the UGC, should shape eventual national rules. For teachers in lower secondary settings, AI for Secondary School Teachers offers targeted guidance on using AI tools in classrooms where students are just beginning to develop analytical skills.
AI tutors that question rather than answer should also be piloted. Rather than leaving students with unsupervised access to general-purpose chatbots, the NCTB and ICT Division could pilot Socratic AI tools - tuned to probe and guide rather than resolve - in a small number of schools before wider rollout.
Why this matters for educators
The Industrial Revolution redefined physical labour and forced schools to teach differently. AI is now doing the same to mental labour. For teachers and education professionals in Bangladesh, the practical takeaway is immediate: exam papers that reward memorised answers will be gamed by AI within a year, and the only durable response is to change what gets assessed. That means building assignments that require students to evaluate sources, construct arguments, and propose original solutions - tasks a chatbot can assist with but not complete. Teachers who start redesigning their assessments now, rather than waiting for a national policy, will be the ones whose students actually learn to think.
Your membership also unlocks:
