The PISA Results Armenia Feared

Listen to the AI generated audio article. 

I was about to close my working day on Tuesday, September 8, 2026, when an OECD newsletter landed in my inbox: the new PISA results were out. I had an immediate flash back to spring 2019, when, in my capacity as Armenia’s Deputy Minister of Education, I initiated negotiations for our participation in the Programme for International Student Assessment, better known as PISA. At one point, after months of negotiations, the process stalled. Among the concerns was an uncomfortable but widely understood expectation: Armenia’s results would be bad. That fear was not entirely misplaced.

The 2026 results are now available, and they are sobering. Armenian 15-year-olds scored 369 points in science, 348 in reading and 376 in mathematics, compared with OECD averages of 482, 461 and 463 respectively. Armenia scored 354 in computational problem solving, against an OECD average of 500. Most alarming, 58.1% of Armenian students performed below Level 2, which is the baseline level of proficiency, in all three core subjects. Across OECD countries, that figure was 19.7%. Only 0.8% of Armenian students were top performers in at least one of the three subjects, compared with an OECD average of 11.9%.

Okay, and then what? I have read and heard the explanations that around 90% of respondents had studied under the new curriculum for only about six months. I understand better than most that improvements in learning outcomes follow lengthy education reforms. I also personally know dedicated professionals who have worked genuinely and tirelessly to improve and advance our students’ learning. But these results should neither be softened nor explained away: they reveal a serious learning problem.

Let’s properly read the results, beyond the headlines and wishful thinking. Armenia participated in PISA for the first time in 2025, which means, there is no earlier Armenian PISA result against which to measure improvement or decline; the assessment provides a baseline, and therefore cannot establish whether the curriculum reform improved learning, failed to improve it or prevented outcomes from deteriorating further. We cannot claim what the data does not show!

This distinction is important, and it matters not to protect a reform, but to protect the integrity of the conclusions we draw from evidence. Though, it is fair to observe, the might of curricula reform is exaggerated: a new curriculum is not a learning outcome. For example, in a 2023 publication, Wang and her colleagues reviewed 156 PISA studies and found that mathematics achievement is associated with a much wider learning ecosystem: teacher qualifications, adequate instructional time, curriculum-content coverage, attendance, school staffing, classroom behavior, home educational resources and socioeconomic conditions. A factor that is consistently positively associated with math achievement is the family socio-economic status, and at the same time student absenteeism and lack of punctuality, class repeating and dropout rate, school prevalence of student misbehavior, shortage of teachers and general staff, and student-centered instruction is negatively associated with math achievement. This last factor is extremely interesting, as it may sound counterintuitive, but the trick is in the labeling: a curriculum may claim it is “student-centered” but it is insufficient. Outcomes depend on how the approach is implemented, how much teacher guidance students receive, and whether teachers and learners are adequately prepared for it. 

PISA Is Not a Curriculum Checklist

PISA subject tests, unlike the questionnaires of the Main Survey, are not publicly available. What is publicly available are the OECD Assessment and Analytical Frameworks, together with selected released and sample questions. These provide a picture of the kinds of competencies PISA assesses. These tests primarily ask whether 15-year-olds can use their knowledge in unfamiliar situations: interpret a text, evaluate evidence, reason mathematically, solve a problem without relying on a rehearsed procedure: bottom line – can they transfer knowledge they have developed to circumstances yet unknown to them.

A country may formally cover every topic in the PISA frameworks and still perform poorly if students encounter those topics mainly through memorization. Conversely, a curriculum may prepare students well without reproducing the PISA framework if teaching develops comprehension, inquiry, reasoning and transfer.

Matching PISA content against Armenian curriculum documents, while useful, will not explain the current results. The caveat is in the intended curriculum versus what teachers actually teach, what students experience and what assessments reward. Curricula research has long established that a reform can upgrade national standards without producing an equivalent change in classroom practice. Implementation is shaped by teacher interpretation of the reform, their knowledge and beliefs, professional support, school leadership, assessment requirements, and the materials available to teachers and learners. This one… I am not entirely sure the teacher community in Armenia has access to materials that support the reform implementation. 

My doctoral research comparing curriculum frameworks in Armenia and Singapore revealed that competency-based reform in Armenia introduced important elements associated with deeper learning, including inquiry, repeated practice and application of knowledge. Yet the curriculum documents offered limited pedagogical guidance in areas such as formative feedback, scaffolding, spacing, interleaving and purposeful revisiting of knowledge. Research by Hwang and colleagues found that the relationship between instructional practices, achievement and socio-economic equity was context-dependent. Even more, pedagogical labels such as “inquiry-based,” “student-centered” or “competency-based” do not guarantee the same classroom practice or the same effect everywhere. A curriculum can point a system in a new direction without automatically giving every teacher the practical architecture required to reach the destination.

But back to Armenia, the PISA report allows us to look beneath headline scores. What we find there is not one simple failure but a series of revealing tensions.

Asking for Help, But Not Questioning the Answer

Beyond the headline numbers, the cross-tabulated data reveal some interesting patterns. One concerning finding is that more than 30% of students in Armenia’s highest quartile for “self-regulated help-seeking,” the ability to recognize when support is needed and deliberately seek guidance, did not agree that information obtained from one source should be checked against other reliable sources. This points to a disconnect between knowing how to seek support and knowing how to evaluate information, and this distinction matters enormously in an increasingly digital and AI-mediated information environment. 

So, imagine, a student encounters difficulties, asks their teacher, parent, peer, or ChatGPT, and they treat the answer received as authoritative and final. Have we not heard of critical thinking? And about critical thinking predominantly being a pedagogical phenomenon? But there is another data point here. The analysis shows that Armenian students are adopting AI faster than schools are developing the capacity and guidance to use it educationally (Armenia -0.63, OECD average -0.03). Although 75.9% of respondents in Armenia reported that they had learned to evaluate AI-generated information, this is a self-reported measure rather than evidence of demonstrated competence. But here is where I get curious: how a student unable to validate the source of information could accurately evaluate the AI-generated information? Could a reform of the educational standards change this? 

The reality is that a new media-literacy course, or a similar endeavor is not a solution. Source evaluation must become routine across subjects. Students should compare accounts, trace origin, identify evidence, distinguish claims from opinions and explain why one source is more credible than another. Linked to this is classroom and national assessments, which must reward such practices, or else “think critically” will remain largely aspirational.

When Students Answer Quickly and Incorrectly

Another finding deserves considerably more attention than the conventional ranking provides: 23.1% of Armenian student reading responses were classified as both fast and incorrect, compared with an OECD average of 8.9%. This may reflect weak reading comprehension, unfamiliarity with computer-based testing, or difficulty sustaining attention, but I would wonder if this were a reflection of low motivation in an assessment carrying no personal consequences. Research on low-stakes assessments shows that when test results carry no personal consequences, some students invest less effort, and rapid responses can signal disengagement, which is itself a significant finding. It is possible that the results are based on the combination of various factors, but I would tackle motivation vis-à-vis assessment: in the motivation research assessment can be an extrinsic reward which, unlike intrinsic motivation, does not really cultivate curiosity, persistence and intellectual agency. 

The Reading Problem Is Not Confined to Reading

Reading, the weakest domain of the test results, transcends to the other domains: how much reading comprehension is needed for math and science? Well, as much as one can get. Students need it to interpret scientific evidence, extract a mathematical problem from a written scenario and distinguish relevant from irrelevant information.

While the derived information is important for both genders, the results revealed an alarming discrepancy: girls in Armenia outperformed boys by 18 points in science and 40 points in reading. Note, OECD average of the reading gap was 30 points. The problem here to tackle has little to do with the curricula reform: the underlying factors could be attendance, motivation and classroom practices that affect boys and girls differently.

Shall We Skip a Class?

Is it a culture of truancy, or are the classes simply not interesting? In the OECD countries on average 66% of students have reported that they have not missed a single class or a school day in the reported period, while in Armenia about 72% “confessed” in the test that they have skipped either a class or a day. Many of us who grew up in the post-Soviet education system may smile at this, but missing class means missing instruction, particularly when the curriculum is already dense. If we refer to the 2023 Wang’s article again, absenteeism and lack of punctuality were among the relatively small number of factors consistently associated with poor results. No curriculum can produce its intended effects if students are not present to experience it.

What Armenia Should Do With the Diagnosis

The decision to participate in PISA was ultimately a decision to stop avoiding an uncomfortable diagnosis. At some point, the system cannot continue to “protect” itself from evidence. But  evidence must be used with equal courage and care.

The results point to serious gaps in subject knowledge and proficiency. But beyond that, they reveal difficulties in persisting, reading deeply, verifying information, exercising critical thinking, and in so many other things. Then there is the data on truancy: with so many ecosystem factors potentially behind it, where do we even start…?

Social media is full of surface level posts and “analyses” these days: some praising those in charge, others framing the results to suit their criticism. I have news: none of that is going to make a difference. What Armenia needs is an implementation-focused research and policy agenda. We need classroom evidence on instructional practices, actual learning time, attendance, learning materials, and hopefully, finally, including textbooks. Teacher professional learning and not a check-the-box attestation. 

Most importantly, Armenia should resist responding to disappointing results with another wholesale rewriting of curriculum documents. The results are bad, and knowing that is better. And yet, are we, the collective we: the teacher, the researcher, the policymaker, the parent, the student, the expert, ready to roll up our sleeves and move beyond masking a change as progress?

Also see