Assam’s AI test is not about machines but teachers

Siddharth Roy

(siddharth001.roy@gmail.com)

There is something deceptively simple about putting artificial intelligence into a classroom. By providing a teacher with a device, an internet connection, and an AI tool, it may seem like the future has arrived. However, merely introducing a new technology has never transformed education. The blackboard did not make a competent teacher, nor did the computer. Artificial intelligence will not either.

Assam's decision to train one lakh government-school teachers in artificial intelligence by May 2027 is therefore significant not because of the number, impressive as it is, but because of what it asks of the state's education system. The real test will be whether AI can help a teacher understand a child better, explain a difficult idea more clearly and make learning less dependent on geography and family income.

The scale of the experiment is considerable. Assam has more than 44,000 government and provincialised schools and over two lakh teachers. Training one lakh of them would touch a substantial part of the public school system. Millions of children could consequently encounter AI not first as consumers of entertainment or social media, but as an instrument of learning. That possibility deserves both optimism and caution.

For a teacher in a remote school, artificial intelligence could be remarkably useful. It could generate several explanations of the same mathematical concept for children at different levels of comprehension. It could help create classroom exercises in Assamese or other familiar languages. A science teacher could use it to design an experiment with inexpensive local materials. A teacher struggling to identify why a group of students is repeatedly making the same error could use AI to analyse their responses and suggest alternative approaches. In a state where distance remains an educational disadvantage, such assistance could matter.

But the danger lies in confusing assistance with substitution. A machine can produce a lesson plan in seconds. A machine cannot understand the reasons why a particular child has stopped answering questions. It cannot see the hesitation in a student's eyes, the anxiety behind silence or the social circumstances that make concentration difficult. The teacher remains the most important intelligence in the classroom.

This approach becomes particularly important in Assam because the state's educational problems are not primarily technological. UDISE+ 2024-25 data show that the secondary-school dropout rate in Assam remains considerably higher than that at the upper-primary level. Behind such statistics are adolescents who leave school for reasons that may include poverty, family responsibilities, weak learning foundations, migration and the perceived absence of economic returns from education. No chatbot can resolve these problems.

What AI can do is strengthen the teacher who confronts them. That distinction should shape Assam's entire training programme. Teaching AI cannot be reduced to teaching teachers how to write prompts. The more important lessons are about verification, judgement and responsibility. Teachers must learn that an AI system can produce an answer that sounds authoritative and is entirely wrong. They must learn how bias enters machine-generated content. They must understand plagiarism, copyright and the protection of children's personal information.

Most importantly, they must learn when not to use AI. That may be the hardest lesson of all.

Education is not simply the efficient transfer of information. A child struggling to solve a problem needs more than another answer. Occasionally the value of education lies precisely in allowing the child to struggle, fail, rethink and eventually discover. If AI is used to remove every moment of difficulty, it could produce students who have access to more answers but possess less intellectual resilience.

There is another concern. Assam must ensure that an AI divide does not develop alongside the existing digital divide. The state has made progress in basic school infrastructure. UDISE+ data indicate that electricity or solar power is available in the overwhelming majority of government and provincialised schools, while access to drinking water is also widespread. But having electricity is not the same as having dependable digital infrastructure. A school with intermittent connectivity, too few devices, and no technical support is unlikely to participate in the AI revolution on equal terms with a well-equipped urban school.

This is where Assam's geography becomes important. The state's river islands, tea-garden communities, forested areas, border districts and remote hill regions do not experience education in the same way as Guwahati. An AI policy designed around urban broadband connectivity could unintentionally widen the very inequalities it claims to address. Low-bandwidth applications, offline resources, shared devices and locally relevant content should therefore be part of the design from the beginning.

There is also a more subtle danger: the industrialisation of classrooms. AI can make it extraordinarily easy to produce worksheets, essays, quizzes, assessments and lesson plans. But an abundance of educational material is not the same as an abundance of education. If teachers begin accepting machine-generated content without interrogation, classrooms may become filled with polished but generic material. The appearance of efficiency could conceal an erosion of intellectual engagement.

The measure of success should consequently not be the number of teachers who complete an AI course. It should be whether teachers subsequently teach differently.

Does a student who struggled with mathematics begin to understand fractions? Does a child in a rural school gain access to better learning material? Does AI reduce the time teachers spend on routine administrative work and provide them more time with students? Do teachers become better at identifying misinformation? Do children learn to question machine-generated answers rather than simply accept them? These are harder questions than counting certificates, but they are the questions that matter.

Assam can also use this programme to address a problem that technology companies rarely discuss: language. India's AI future cannot be built exclusively in English. If artificial intelligence is to become a genuine public educational resource, it must understand the linguistic and cultural worlds in which children actually live. Assamese-language educational resources, local examples and culturally familiar contexts could turn AI from an imported technology into a genuinely Indian educational tool.

There is an irony here. The more powerful AI becomes, the more important the human teacher may become. Machines are increasingly effective at retrieving, organising and generating information. That increases the value of uniquely human functions of education, like curiosity, empathy, judgement, imagination, and moral reasoning. The teacher's role may gradually shift from being the principal source of information to being the person who teaches children how to interrogate information. That is a profound change.

Assam's one-lakh-teacher initiative should therefore be viewed not as a race to make its schools "AI-enabled", but as an opportunity to rethink what a teacher does in an age when information is abundant and answers are cheap.

The state should aspire to produce not teachers who know how to use every new AI tool, but teachers who know which tool is useful, which answer is suspect and which question must still be answered by a human being.

The success of Assam's AI experiment will not ultimately be visible on a computer screen. It will be visible in a classroom where a child who once remained silent raises a hand, asks a difficult question and is encouraged to think for herself. That is the intelligence education must protect.