ICT use and student achievement: Evidence from PIRLS 2021 using machine learning techniques

In this paper, we analyze the relationship between students’ ICT use and reading achievement using PIRLS 2021, adopting a cross-country approach with data on 138,100 fourth-grade pupils from 29 European education systems. Specifically, we focus on examining learning-related ICT practices at home and in the classroom and their relationship with reading achievement. To do this, we employ a machine-learning framework based on counterfactual prediction methods and a high-dimensional set of covariates to estimate differences in predicted outcomes across students with different ICT-related profiles, while also exploring heterogeneity by gender and socioeconomic status. We quantify uncertainty using resampling-based confidence intervals and assess robustness across alternative specifications and sample restrictions. Results suggest that learning-related ICT use at home is positively associated with reading achievement, particularly for activities related to online information search and the identification of reliable sources. These relationships appear to be stronger among socioeconomically disadvantaged students. By contrast, frequent use of digital tools in the classroom does not appear to be associated with better reading outcomes and, in some cases, are negatively associated with student performance. These findings suggest that education policies should prioritize supporting purposeful ICT use at home, particularly among socioeconomically disadvantaged students, while ensuring that classroom digital initiatives Are pedagogically meaningful rather than merely technology-driven.