Education Technology

Although this review primarily focuses on technology-aided interventions aimed directly at improving student achievement, there are important use cases and growing evidence on other important applications that target teachers and parents and may generate downstream effects on student outcomes.

Information to parents about schooling

The first set of interventions targets a ubiquitous challenge in all education systems: parents are, at best, only partially informed about their children’s participation and performance in school. ICT tools, such as text messages, can help address these gaps in automated ways without imposing huge additional burdens on parents.

Berlinski et al. (2025) provide evidence from Santiago, Chile, where 26% of parents at baseline could not correctly report their children’s grades in school and 48% could not approximate their child’s attendance over the past two weeks. Motivated by this, the authors randomise students (and schools) to receive weekly and monthly text-message updates for parents of primary school students, providing information on their children’s attendance, grades, and classroom behaviour. After 18 months, they find that math GPA improves by 0.09σ, school attendance improves by 1.1 percentage points, and increases were large for students more at-risk of low performance and low attendance. Similar positive gains in achievement from providing information to parents are also documented in Barrera-Osorio et al. (2020) in Colombia and Bettinger et al. (2021) in Brazil.

Evidence from LMIC settings outside Latin America, although from in-person interventions not using EdTech, indicates that similar information gaps are likely to be important there too (Dizon-Ross 2019, De Walque and Valente 2023). With the spread of mobile phones, and with electronic record-keeping now standard in many education systems, it is likely that scalable interventions using ICT tools would be both feasible and effective across a broad range of LMICs, including in sub-Saharan Africa, the MENA region, and south and southeast Asia.

Information asymmetries between parents and children are, of course, broader than just school attendance and performance. For example, Gallego et al. (2025) provided text messages to parents in Peru encouraging them to engage with a TV-based information intervention, called Decidiendo para un Futuro Mejor, which provided accessible information on the benefits of education, expected wage returns, and financial aid opportunities. The SMS-based encouragement led to reductions in dropout by 0.6–0.7 percentage points (from a 10.2% control group baseline).[1]1. An interesting example is presented by Gallego et al. (2020) who demonstrate that providing information to parents about their children's internet use reduced children's internet use by 6–10%.  The principal learning from this stream of work is that parents are responsive to SMS-based information on education that can lead to important downstream changes in behaviour and, potentially, school participation and achievement.[2]2. It is possible that the mechanism by which such text messages affect child outcomes is by increasing the salience of education to parents rather than the direct provision of information (Bettinger et al. 2021). That distinction is not materially important for assessing the potential for EdTech to improve outcomes.  ICT tools in this instance are clearly a mechanism that expands the possible scale, at potentially lower cost, for information provision.

Information about educational options

A separate class of EdTech-enabled information interventions focuses on making parents aware of different educational options that they may potentially be eligible for. This is especially important given that central school choice mechanisms are now increasingly common also in LMICs. These application portals frequently require students to apply for a set of educational options (e.g. high schools) with allocation subsequently determined through centralised allocation mechanisms (such as Deferred Acceptance or Immediate Acceptance). Thus, even as a default, these policies can often require students and parents to deal with internet-based application mechanisms.

Allende et al. (2019) evaluate a video-based intervention that tries to increase the awareness of neighbourhood school characteristics and the perceived returns to school quality in Chile, combined with report cards comparing schools available in the localities. They show that the treatment leads parents to choose schools that are further away and that the students perform ∼0.13σ better in math in fourth grade (five years after treatment). This suggests that even in a setting like Chile, which has had a long history of school choice, there is a potentially important role for tech-enabled interventions that reduce information frictions for households.

Agte et al. (2024), also in Chile, provide perhaps the best illustration of the potential of tech-enabled interventions to reduce informational frictions in centralised application processes. First, they demonstrate that parents are misinformed on multiple dimensions: they know fewer than 20% of the available options well (50% not at all), systematically underestimate the quality distribution of options they are not aware of, and overestimate the quality of their high-ranked options. They also embed randomised interventions into the design of the school explorer platform across two treatment arms: (i) providing personalised information about the joint distribution of nearby schools’ prices and quality scores and (ii) highlighting nearby low-cost high-quality schools. In a third RCT, they provide tailored feedback on the initial applications parents submitted, targeting parents’ misperceptions about known schools. These interventions all change parents’ information sets and application choices, with low-SES parents especially responsive to being informed about low-cost high-quality schools in their vicinity. Since these interventions are entirely run on internet-based platforms with administrative data, it is possible to scale these up at very low marginal costs.[3]3. In similar vein, see also Arteaga et al. (2022) who study the effect of 'smart matching platforms” in Chile (and New Haven), that warn applicants submitting applications with high risk of nonplacement. Warning applicants about their risky applications leads to lengthened applications, reduced risk of nonplacement, and enrolment in better schools.

In contrast to the impressive body of work in high income settings, such as the US, Chile, or Europe, work on technology-enabled interventions in school choice mechanisms in LMICs remains extremely limited. Yet, the information that does exist suggests that such interventions could potentially lead to large improvements in these settings as well. For instance, Ajayi (2024) shows that high-scoring students from disadvantaged educational backgrounds are significantly more likely to rank low-performing schools above high-performing schools. Likewise, Romero and Singh (2024) who study a large voucher-like affirmative action policy in India show that many students only rank one option even when allowed to provide a rank-ordered choice list to maximise chances of placement. These patterns suggest a potential role for information provision in these settings; given that parents and students already have to engage with the internet portals in order to apply, incorporating this information in the application portal would be one potential way of reducing these information asymmetries.[4]4. That being said, it is far from guaranteed that such interventions would be as successful in high-disadvantage populations in Ghana or India than in high-income settings. Ajayi et al. (2025) show that providing such information did affect school enrolment decisions but not matriculation rates: households faced further frictions such as household shocks that prevented enrolment in better schools from translating into graduation. The importance of multiple frictions, which disproportionately affect low-SES households even within their setting, is also stressed in Romero and Singh (2024). For large shares of eligible populations, an important challenge in LMIC settings like India may be to just get them on the portal (rather than the choices they make there).

EdTech to support teachers

EdTech products reviewed earlier, such as learning software, also frequently include modules to support teachers (e.g. by giving them diagnostic feedback on the learning levels of individual students). More importantly, implementation of any interventions in the classroom typically relies substantially on teacher support and motivation. In addition to such products, however, there are several potential uses of technology that are primarily teacher-facing, without requiring any direct engagement by students.

One potentially important application is in providing the structure and support for implementing structured pedagogy, which is identified in recent reviews as a particularly promising approach to address the ‘learning crisis’ in LMICs. These structured pedagogy interventions frequently rely on very detailed lesson plans and teacher support materials; in principle, digital tools could make scripted lesson plans and relevant teacher-support materials much more widely available (and easier to deploy in class teaching).[5]5. For a useful example from the US, see Jackson and Makarin (2018) who show that being randomly assigned to receiving a license for online lesson plans and social media based support in using it increased test scores by 0.09σ in Virginia.

The potential for this use is highlighted by the results of Gray-Lobe et al. (2022) who study the composite effects of enrolling in private schools which use a highly-scripted lesson plan using tablets provided to every teacher. At the end of two years, they find very large causal effects of roughly 0.9σ of being enrolled in these schools. The authors highlight that the tablets were important for the transmission of lesson plans to the teachers, in training them to adopt the high degree of standardisation envisaged by the Bridge schools, and in monitoring teacher attendance and compliance with standardisation. However, we do not know whether the same gains could have been achieved without the tablet computers, or indeed other features of the schools.[6]6. Note that the empirical strategy in Gray-Lobe et al. (2022), based on the randomised allocation of vouchers, identifies the total effect of enrolment in Bridge schools. Since these schools differ from counterfactual options also in staffing, teacher incentives and school infrastructure, it is difficult to attribute the causal effect specifically to the use of tablet-based lesson plans.  Trials specifically focused on EdTech-enabled structured pedagogy in LMICs are currently underway in Ghana and Uganda, which may be expected to provide greater insight on the potential effects without changing other aspects of the school environment.[7]7. See the description of intervention, Inspiring Teacher's Tools for Foundational Learning Improvement (TFLI), and trial here on the JPAL website.  However, evidence from Kenya’s PRIMR programme for foundational literacy finds large gains from structured pedagogy, but not differentiated by whether only teacher supervisors were provided tablets, teachers were provided tablets, or students were provided e-readers (while costs were much higher in the more device-intensive treatment arms) (Piper et al. 2016).

ICT tools could potentially also be useful in improving teacher training, or enabling their delivery at cheaper costs. For instance, video conferencing tools are now common even in LMIC education systems for teacher training at large scale. Potentially, these could also be used for the delivery of teacher mentoring. While the logistical convenience of these tools is easy to recognise, and apparent in their uptake across very different geographies, a major risk is that of dilution of the intensity with which the intervention is implemented. The clearest example of this is provided by Cilliers et al. (2022) who experimentally compare in-person teacher training with virtual teacher training targeted at English language teachers in South Africa. While initial results indicated similar effects across the two arms after one year, the in-person training was both more effective and more cost-effective after three years.

Overall, while there is considerable enthusiasm for using EdTech to target teachers, and potential mechanisms by which it may improve teacher productivity, we are not aware of compelling examples yet where it has been shown to do so at moderate-to-large scales in LMICs.

For full reference list see the end of the conclusion chapter.