Perspectives

“This programme will reach 30,000 teachers.”
It is the kind of number that wins funding. It is also, very often, a projection wearing the clothes of a result.
Most teacher training in low-income education systems runs on the cascade model: train a small group centrally, and rely on them to train the next group, who train the next. It is cheap, fast, and easy to report. It reaches enormous numbers on paper.
And it usually does not change what happens in the classroom.
The research is remarkably consistent on why. At each hand-off down the chain, the content thins out. The people asked to train their peers were prepared to teach children, not to train adults. And nowhere in the model is there any follow-up in the actual classroom, which is the one thing that changes how a teacher teaches.
One review named the mechanism perfectly: “spray and pray.” Spray information at as many people as possible, and pray something sticks.
Here is the arithmetic laid bare. In one documented cascade, twelve trained facilitators were recorded as eventually reaching 8,200 people. The headline is 8,200. The number actually prepared with any depth is twelve. Everything in between rests on transfer surviving relay after relay, unfunded and unobserved.
The distinction that matters is simple:
A cascade spreads information and counts who was reached. What changes a classroom is ongoing guided practice: a teacher working at a specific skill, in their own room, with someone watching and helping, until it holds.
The two look alike on a report. They cost very differently to produce. Only one of them reaches a child.
At TDSO we build the second kind. We prepare teacher trainers over years, not workshops, and we build that capability into the institutions that are already responsible for training Cambodia’s teachers, so it stays after we leave. We plan ourselves out of the work rather than create a dependency on us.

Why does TDSO still say M&E, when the sector has moved on to MEAL, MERL and MEARL?
Not because we reject accountability, learning or research. Because those letters name different kinds of things.
Monitoring and evaluation form an instrument: specialist work that gains from being concentrated in dedicated, independent hands. Accountability and organisational adaptation are obligations, conditions of holding a management role at all. An instrument can be given to a unit. An obligation cannot; a unit absorbs the appearance of it while the substance quietly thins elsewhere.
That is the structural risk the composite acronyms invite: responsibilities that belong to everyone who manages acquire an address on the organogram, and everyone else stops looking.

Evidence can be delegated. Responsibility cannot.

Cambodia’s education sector is placing growing institutional confidence in artificial intelligence as a response to longstanding gaps in teaching quality, infrastructure, and equity. This essay argues that confidence is misplaced. Across MoEYS programmes such as Live Teaching 2026, the Primary Learning Platform, and Sala Digital, and among development partners including UNESCO/STEPCam, the same structural logic repeats: AI and digital tools are deployed as compensatory mechanisms ahead of the conditions that would make them effective. The result is not improvement but the reproduction of existing inequalities at a higher level of technological complexity. Drawing on cognitive science, bilingual education research, and Cambodian educational scholarship, the essay identifies three structural harms: AI widens the access gap it claims to close, introduces cultural and economic bias through Western-centric training data, and erodes the foundational knowledge and critical capacity that genuine learning requires. The argument concludes that the priority must be teacher development, a Cambodian core knowledge curriculum, genuine bilingual education infrastructure, and honest investment in physical school conditions, not AI platforms that assume into existence the very conditions they cannot provide.

Cambodia’s education sector is placing growing institutional confidence in artificial intelligence as a response to longstanding gaps in teaching quality, infrastructure, and equity. This essay argues that confidence is misplaced. Across MoEYS programmes such as Live Teaching 2026, the Primary Learning Platform, and Sala Digital, and among development partners including UNESCO/STEPCam, the same structural logic repeats: AI and digital tools are deployed as compensatory mechanisms ahead of the conditions that would make them effective. The result is not improvement but the reproduction of existing inequalities at a higher level of technological complexity. Drawing on cognitive science, bilingual education research, and Cambodian educational scholarship, the essay identifies three structural harms: AI widens the access gap it claims to close, introduces cultural and economic bias through Western-centric training data, and erodes the foundational knowledge and critical capacity that genuine learning requires. The argument concludes that the priority must be teacher development, a Cambodian core knowledge curriculum, genuine bilingual education infrastructure, and honest investment in physical school conditions, not AI platforms that assume into existence the very conditions they cannot provide.

The dominant model of grant funding channels resources through projects designed to advance donors’ priorities rather than the core programmes and institutional needs of the organisations and public bodies receiving support. Across the education sector, this approach fragments local capacity, sidelines ministries and teacher training institutions, and produces outcomes that are difficult to sustain. Donors have a legitimate role in setting standards for how funded work is conducted, requiring inclusion, environmental responsibility, financial transparency, and a focus on real outcomes rather than administrative outputs. What they should not do is determine what the work is. TDSO calls on donors and funding bodies to draw a clear distinction between the floor they set and the ceiling they must not impose, and to shift towards funding arrangements that actively support coordination, strengthen national systems, and place substantive programme decisions in the hands of those closest to the learners they are meant to serve.