On September 8, the world of education will turn its attention to a single dataset. The release of the PISA 2025 results is poised to reignite one of the most contentious debates of our digital era: does technology in the classroom help or hinder learning? Having analyzed the trends buried in over a decade of this global assessment data, I’m confident I know what the headlines will say. I’m even more certain that a fundamental, and potentially dangerous, misinterpretation is about to unfold.
For years, the story told by the Programme for International Student Assessment (PISA) has been strikingly clear. Conducted by the OECD, this triennial test evaluates 15-year-olds in mathematics, reading, and science across dozens of countries. Alongside the exam, students answer detailed questionnaires about their school lives, including how much time they spend using digital devices for learning. The consistent, global pattern has been a linear decline: as classroom tech use increases, academic performance largely goes down.
But I predict this 2025 report will look different. This time, the graphs will likely show students who use a moderate amount of technology—say, between one minute and three hours a day—outperforming their peers who use none. Commentators will quickly point to this curve as evidence of an educational technology “sweet spot.” The narrative will shift overnight. Years of cautionary data will be swept aside in favor of a new, comforting conclusion: the problem was never the screens themselves, but our inability to use them in moderation.
While this new interpretation might feel like a resolution, it misses a far more critical story hidden within the numbers. This apparent sweet spot isn’t a sudden discovery of 2025. Its emergence has been meticulously tracked and documented across previous PISA cycles, growing stronger as digital tools have become ubiquitous. And the stronger it gets, the less it actually tells us about technology’s benefits and the more it reveals about the statistical mirage created by its own saturation.
To understand why, we must rewind. The first flicker of this pattern appeared in the 2012 data. While most countries showed the familiar downward slope, three nations stood out: Australia, Denmark, and Sweden. In these systems, students reporting moderate tech use seemed to have an edge over those reporting none. This created a subtle inverted-U shape on the graph. These weren’t random outliers. As noted by researchers delving into international educational trends, they were among the most technologically saturated school systems in the entire assessment. This coincidence suggested a provocative idea. Perhaps these countries weren’t revealing an optimal dose of tech, but what happens when the group of students not using technology becomes so small and unusual that it ceases to be a reliable baseline for comparison.
| Year | Group Size (Non-Users) | Performance Gap |
|---|---|---|
| 2012 | Not specific | Not evident |
| 2015 | Smaller | Larger |
| 2018 | 0.6% Denmark | Higher |
| 2022 | 15% | Wider |
The hypothesis was simple yet powerful. If technological saturation was altering the composition of the “zero-use” group, then the apparent advantage of light use should grow as that group shrinks. By the 2015 assessment, as devices flooded classrooms globally, this correlation solidified dramatically. Across 47 education systems, the smaller the group of non-users, the larger the performance gap appeared in favor of light users. Crucially, this analysis, which I’ve replicated using school fixed-effects models to compare students within the same schools, held even after accounting for socioeconomic factors. The pattern was not about wealth or resources; it was about prevalence.
- Counter-argument: Countries getting better at tech use
- Data showed contradiction: 9 of 10 systems declined
- Non-user group fell significantly
- 2018 data confirmed inference
- COVID-19 effect on data
- Unique outlier with tech becoming essential
The 2018 data turned the flicker into a beacon. In nations like Denmark, students who used no technology at school represented a mere 0.6% of test-takers. In Japan, that figure was still 46%. The term “non-user” no longer described the same reality everywhere. In a highly saturated system, the few students not using tech likely differed from their peers in profound ways—perhaps due to learning disabilities, strict parental choices, or lack of access at home. The statistical fingerprint was unmistakable: the rarer the non-user, the worse they performed, and the better light use looked by comparison. The switch to computer-based testing in 2015 explained some early shifts, but it couldn’t account for why the divergence kept widening through 2018.
Then came the earthquake of COVID-19. The 2022 PISA cycle, conducted amid global school closures and a forced revolution in remote learning, is a unique outlier. Educational technology use became not just common but essential for basic schooling. Consequently, the zero-use group shrank to a historic low of just 15% of students globally. Yet, even in this radically altered landscape, the same statistical rule held. The smaller the non-user group in a country, the larger the apparent benefit of light technology use. The inverted-U pattern appeared in a staggering 68 out of 75 countries, creating a seemingly universal law of EdTech efficacy.
But this is where we must pause. The near-ubiquity of the pattern is its own warning sign. When almost every student uses technology, finding that the highest-performing group comes from within the tech-using majority tells us shockingly little about causation. It’s like trying to determine if smartphones make adults more productive by only studying the tiny, self-selected group of people who voluntarily abstain from owning one in 2024. That group is not a representative sample of the general population; it’s an anomaly. Their lower productivity wouldn’t prove that smartphones cause success, just that the abstainers are different.
This is the core dilemma PISA 2025 will face but likely fail to highlight. We are losing our control group. As digital technology becomes the default setting for education, the “no tech” cohort is vanishing. When it becomes a tiny, atypical sliver of the student population, outperforming them is a low bar. It does not mean technology caused the higher performance. It may simply mean that the students still in that group face compounded challenges that no amount of technology could easily solve.
So, as we await the September 8 release, my predictions are twofold. First, the inverted-U “sweet spot” will be more pronounced than ever. Second, it will be widely heralded as proof that moderate technology use is the key to learning. We will hear that the relationship is “nonlinear” and that past concerns were overblown.
We should greet this narrative with intense skepticism. The existing longitudinal data provides a compelling alternative explanation: the apparent benefit is a statistical artifact of technological saturation, not evidence of its pedagogical power. This doesn’t prove technology is worthless, but it starkly illustrates that the PISA data, in its current form, can no longer answer the question we most want to ask. When the beautiful graphs appear, remember that “performing better than the remaining non-users” is not at all the same thing as “performing better than the average student would perform without technology.” Our ability to discern the difference is fading with every new device rolled into a classroom. The real story of PISA 2025 won’t be about a sweet spot; it will be about the silent disappearance of the comparison we need to find one.
