Unveiling Educational Inequity Pathways: A Causal Machine Learning Analysis of Australian Mathematics Achievement

• In Event: How Technology Shapes Educational Opportunity: Promise and Pitfalls

Abstract

This study employs Invariant Risk Minimization (IRM) to identify stable causal mediators of educational inequity across social, ethnic, gender, and treatment dimensions using PISA 2022 Australian data (N=13,021). Integrating Bourdieu's theory with machine learning methods, we reveal that home ICT access and school attendance consistently mediate achievement gaps across equity dimensions. Domain-specific mechanisms emerged: social equity was mediated through behavioral patterns (tardiness), ethnic equity through digital access and linguistic factors, and gender equity through psychological mediators including career aspiration clarity. Simulation-based interventions demonstrate that improving ICT access from none to one device yields 48-point gains, while regular attendance produces 34-point improvements. These findings challenge deficit-based explanations by revealing how structural disadvantage operates through relational mechanisms, offering evidence-based targets for equity interventions.

Authors

• Hui Tao, University of Adelaide

• Ruoxin Wang, University of South Australia

• Haiyao Cao, University of Adelaide

• Jinan Zou, University of Adelaide