Before AI in education, make sure children can learn and reason

Bangladesh’s education debate has recently taken an unexpected but welcome turn. As governments everywhere announce artificial intelligence initiatives, Education Minister A.N.M. Ehsanul Hoque Milon has argued for something less fashionable but more fundamental: children must first acquire strong reading, reasoning and verification skills if they are to use AI safely and effectively.

The minister made this case at Unesco’s Digital Learning Week 2026 in Paris, where more than 25 ministers and designated education representatives adopted a joint statement on sustaining education as a human right and a common good in the age of AI. Milon spoke in the segment devoted specifically to age-appropriate AI.

Tap here to add The Daily Star as a trusted source

His intervention stood out because it resisted the prevailing pressure to equate technological adoption with educational progress. AI may now be unavoidable, he acknowledged, but its success should not be measured by the numbers of users, pilot projects or participating institutions. Instead, Bangladesh should judge it by learning outcomes, effects on teachers’ workloads, equity, safety and cost-effectiveness. Teachers should participate in decisions about pilots, procurement, evaluation and correction, while automated systems should not independently determine recruitment, promotion or disciplinary decisions.

This is the right kind of de-escalation. It does not reject AI or ask Bangladesh to retreat from technological advancement. It steps back from the competition to adopt first and asks the question that should precede every education technology decision: what problem are we trying to solve, and what must already be in place for the technology to help?

Unesco’s roundtable also rejected technological solutionism. Its theme was “Sustaining the public purposes of education in the age of AI,” and the ministerial deliberation focused on four questions: safety, age appropriateness, procurement and the real cost of ownership, and AI sovereignty. These are questions of public authority, not merely computing. Who decides when a child is ready to use AI? Who controls student data? What conditions must vendors satisfy before entering schools? Can a ministry audit a system, understand its full cost and withdraw without losing its data or becoming dependent on one company?

The resulting statement, supported by 25 education ministers, calls for auditable systems, teacher participation, developmentally appropriate use, protection of learner data, inclusive and multilingual design, and calculation of total ownership costs. It asks providers to demonstrate educational benefit before deployment and accept accountability for harm. Most importantly, it states that learning is not served when AI replaces human relationships and professional judgment.

These principles are timely because Bangladesh is preparing to give AI, robotics, machine learning and cybersecurity greater prominence in textbooks for classes 6 to 10. Introducing students to emerging technologies is sensible. But introducing new terminology is easier than ensuring students possess the literacy, numeracy and reasoning required to understand it. Without those foundations, an AI curriculum could encourage dependence on tools whose outputs students cannot question.

Bangladesh’s Five-Year Strategic Framework for Reform and Development records learning poverty—the inability to read and understand a simple text by age ten—at 50 percent and seeks to reduce it to 25 percent by FY2030 and 22 percent by FY2031. That target should be treated as the first milestone of the national AI strategy.

A ten-year-old who cannot comprehend a short Bangla passage will not become an independent evaluator of AI-generated information. Generative tools can produce fluent but inaccurate answers. Detecting those errors requires reading comprehension, subject knowledge and the confidence to question an authoritative-sounding machine. Foundational learning is therefore no longer only preparation for conventional schooling. It is the first layer of AI safety.

This connection also tells us what government should do next. The most urgent AI-readiness intervention may be ensuring that every struggling learner receives structured support in reading and numeracy. A recent roundtable organised by The Daily Star argued that remedial education should be built into regular classroom teaching, teacher training, assessment, mentoring and the next Primary Education Development Programme (PEDP)—not left to temporary projects. Participants also stressed that teachers must diagnose which child is behind, in which subject and by how much before selecting the remedy.

That model should be implemented systematically. Schools need regular diagnostic assessments, protected time for catch-up instruction, teaching at the child’s actual level, suitable Bangla reading materials, teacher mentoring and additional support where staffing is weakest. Success should be measured by learning gained, especially among disadvantaged children—not by the money spent, materials distributed or training sessions completed.

AI tools could then be piloted as complements to teachers—to help identify learning gaps, adapt practice materials or reduce routine administrative work. Pilots should disclose costs and data arrangements, compare outcomes against credible alternatives, and include rural, low-income and Bangla-medium schools. Teachers must be co-designers and evaluators, not passive recipients.

Cost and sovereignty deserve equal attention. Bangladesh should not confuse access to a free AI interface with ownership of an education solution. Costs can emerge later through subscriptions, connectivity, devices, data storage, teacher training and dependence on proprietary platforms. Unesco’s AI for Good initiative has highlighted the potential of smaller, energy-efficient models to make AI more affordable and accessible in resource-constrained settings. For Bangladesh, that strengthens the case for frugal, interoperable and locally adaptable systems, including tools that work well in Bangla, rather than automatic dependence on the largest imported models.

Bangladesh now has an opportunity to turn the minister’s international message into a domestic reform principle. Every proposed AI initiative in education should have to answer five questions: Which learning problem will it solve? Are the intended students developmentally ready? What evidence will establish educational benefit? How will teachers shape and oversee its use? And what are the full financial, data and dependency costs?

The government’s decision to modernise the curriculum can still become transformative. Transformation will come from repairing the foundations, restoring teachers’ agency and subjecting technology to educational purposes rather than reorganising education around technology.

Bangladesh should certainly prepare its children for a world shaped by AI. But the country’s AI future must begin with a child who can read, reason, verify and learn.

Dr M Niaz Asadullah is professor of economics at Dhaka University and a visiting professor at the University of Reading in the UK.

Views expressed in this article are the author's own. 

Follow The Daily Star Opinion on Facebook for the latest opinions, commentaries, and analyses by experts and professionals. To contribute your article or letter to The Daily Star Opinion, see our guidelines for submission.