From Classroom Computers to Artificial Intelligence
The recent history of educational technology in Spain can be told through a succession of technologies that arrived in schools amid great expectations. First came computers. Then the internet. Later came interactive whiteboards, learning platforms, tablets, and one-device-per-student programs. Today, artificial intelligence dominates headlines, educational conferences, and policy debates. Each of these tools seemed to herald an imminent transformation of teaching.
Yet while technology advanced rapidly, evidence about its educational impact accumulated much more slowly. Only in recent years has a sufficient body of research, experimental evaluations, and comparative studies emerged to identify relatively stable patterns across very different contexts.
The report Evidence-Based Integration and Effective Use of ICT and Generative AI in School Education in Spain, prepared by EsadeEcPol, builds on this accumulated evidence. Across 53 pages, it reviews international research on digital technologies and artificial intelligence, analyzes data from PISA, PIRLS, TIMSS, and TALIS, and examines the extent to which Spain’s education system possesses the conditions needed to use these tools effectively.
Spain has made remarkable progress in digitizing its schools. One hundred percent of educational institutions have internet access and 96.9% have operational Wi-Fi networks. Digital platforms are now part of everyday school life, and teachers’ adoption of AI tools is beginning to approach the average of more advanced countries.
However, EsadeEcPol’s review shows that technological expansion alone offers few clues about what actually happens inside classrooms. Two decades of evidence point in a different direction. Outcomes depend less on the presence of devices than on how they are used. Nor do they seem to depend primarily on the amount of technology available, but rather on the pedagogical and institutional conditions surrounding its use.
Based on this review, the report identifies several lessons that repeatedly emerge in international research. These findings help explain why some digital initiatives generate meaningful improvements in learning while others leave barely a trace. They also explain why the arrival of artificial intelligence is not necessarily altering the rules that educational research has been observing for years.
More Technology Does Not Always Mean More Learning
The report’s first conclusion is that the relationship between technology and learning is not linear. More devices, more platforms, or more hours of use do not automatically translate into better academic outcomes. In some cases, the opposite is true.
One of the strongest pieces of evidence comes from a 2019 J-PAL review of 126 studies on educational technology in primary and secondary education. Its conclusion was clear: programs focused solely on expanding access to computers or internet do not systematically improve grades or standardized test scores.
This does not mean they lack value. Many of these initiatives improve students’ digital skills, familiarize them with technological tools, and expand access to educational resources. However, their effects on academic learning are often limited when they are not accompanied by changes in teaching practices.
International research offers numerous examples. Programs such as One Laptop per Child increased access to devices without generating significant improvements in mathematics or reading. Technology was present, but educational practice remained essentially unchanged.
Spanish data point in a similar direction. Research based on PISA shows that the impact of technology depends heavily on the activities carried out with it. Some practices—such as searching for information or using digital tools integrated into learning—are associated with better outcomes. Others have neutral or even negative effects.
The type of device also matters. According to PISA 2022 data cited in the report, computer use in schools is associated with gains of up to 17 points in mathematics compared to students who never use them. Mobile phones paint a very different picture: frequent use for learning activities is associated with an approximately 11-point penalty. The difference likely lies not in the technology itself, but in the attention, supervision, and tasks typically associated with each device.
A similar pattern appears in intensity of use. Data from both PISA and PIRLS suggest an inverted U-shaped relationship: the best results are found among students who use digital resources moderately. Below that threshold, technology has little impact; above it, benefits begin to decline and may disappear altogether.
This suggests that greater exposure often comes with less focused use, weaker pedagogical supervision, or activities of limited educational value. As a result, research has gradually shifted attention away from devices and toward practices. What matters is not how much technology enters a school, but what role it plays once it gets there.
And it is precisely in this area that some of the most promising findings emerge: when technology is used to adapt instruction to the specific needs of each student.
The accumulated evidence from international research points to the same conclusion: educational technology is never better than the education system that receives it.
Where Technology Makes a Difference: Personalization
If evidence shows limited benefits when technology merely reproduces traditional practices, results improve substantially when it is used to adapt instruction to students’ actual learning levels. This is the most consistent pattern found throughout the research reviewed by EsadeEcPol.
In any classroom, students differ in pace, prior knowledge, and needs. Adapting instruction to this diversity has historically been one of the most difficult tasks for education systems. Digital tools can automate part of this process through activities and sequences that continuously adapt to student performance.
The report highlights Mindspark, one of the most widely studied adaptive learning programs internationally. After four and a half months of use, students improved by 0.37 standard deviations in mathematics and 0.23 in language. In educational research, these are substantial effects.
Other experiences show similar trends. A computer-assisted learning program in India produced gains of 0.35 standard deviations in the first year and 0.47 in the second. In the United States, SimCalc improved mathematics performance by between 0.50 and 0.63 standard deviations.
Despite differences in context, age, curricula, and methodology, these programs share one characteristic: technology does not replace teaching. It complements it, providing differentiated support where uniform instruction struggles.
This conclusion is especially relevant for understanding current interest in artificial intelligence. Much public debate has focused on chatbots’ ability to generate text or solve exercises. The report points elsewhere: AI’s greatest educational value may lie in its ability to personalize instruction at a scale previously difficult to achieve.
Evidence on artificial intelligence remains limited compared to that available for ICT more broadly. Still, early findings reinforce similar conclusions. Benefits appear when technology supports learning and helps students progress step by step. Risks increase when it replaces the cognitive effort students should perform themselves.
A recent experiment cited in the report illustrates this difference. Students who used a conventional chatbot initially improved their mathematics performance. However, once the tool was removed, they performed worse than students who had never used it. By contrast, students who worked with an AI tutor designed to provide hints and guidance without directly giving answers retained the benefits.
This finding encapsulates one of the central lessons of recent research: technology produces its greatest benefits when it expands the capabilities of students and teachers. When it replaces them, those benefits disappear.
System Capacity
The report’s conclusions lead to a more practical question: if we know which uses of technology work best, what conditions does an education system need to implement them?
EsadeEcPol’s answer shifts attention away from infrastructure. The main obstacles no longer seem to lie in connectivity or devices, but in the system’s capacity to integrate technology consistently and sustainably.
Teachers provide a revealing example. TALIS 2024 data show that Spanish teachers work under significant organizational pressure. Administrative burden stands 12.6 points above the OECD average. The need to adapt to curricular changes exceeds the international average by 23.8 points, while administrative requirements add another 17.3 points. Teachers also spend an average of 18 hours per week on non-teaching tasks—around one and a half hours more than the European average.
Many effective technological interventions require precisely what education systems often lack most: time. Time to plan activities, interpret student data, adapt materials, provide feedback, and experiment with new methodologies.
Artificial intelligence may help in this area. Surveys cited in the report suggest that teachers using AI tools weekly estimate time savings of nearly six hours per week, with more than 80% identifying benefits in administrative and organizational tasks. Spain’s National Office of Foresight and Strategy estimates that AI could free up as much as one working day per week.
Yet providing tools is not enough. Around 35% of Spanish secondary teachers report using AI in their work. Among non-users, three out of four cite lack of training as the main barrier.
The report identifies one of the system’s major weaknesses: Spain has educational regulations, digital competence frameworks, and AI regulation, but still lacks operational guidance to help schools and teachers translate these tools into concrete teaching practices.
International experience points in a common direction. The most successful systems are not necessarily those that invest the most, but those that build institutional capacity, train teachers, and develop long-term strategies. Technological integration resembles a cumulative process of piloting, evaluating, correcting, and trying again more than a one-time innovation.
Technology Is Only as Good as the System That Receives It
Educational technology produces its best results when it is embedded in systems capable of giving it a clear purpose. Devices, digital platforms, and artificial intelligence can expand learning opportunities, facilitate personalization, and improve teaching processes. But their effects depend largely on the conditions under which they are used.
Two decades of evidence point consistently in the same direction: technology alone rarely transforms education. Systems that combine technological tools with pedagogical objectives, teacher training, and monitoring mechanisms achieve far more consistent results.
Spain enters this new phase with widespread digital infrastructure and a stronger evidence base than ever before. Research shows that technology can improve learning—but only under specific conditions. The system’s ability to create those conditions will be decisive.
Ultimately, the accumulated evidence from international research points to the same conclusion: educational technology is never better than the education system that receives it.
