ACT Government Education Directorate

Education Outcomes and Ethically Aware Machine Learning

Claire Benn, Sarita Rosenstock, Colin Klein, Chelle Adamson, Finn Latimore, and Dan Steinberg

The ACT Education Directorate, HMI and the Gradient Institute have leveraged their expertise in aiding the Directorate improve educational outcomes for Primary and Secondary schools in the ACT. The Education Directorate has recently come into possession of data that capture factors that may influence the educational outcomes of these students from both inside (classroom engagement, student teacher relations, etc.) and outside (family support, home safety, etc.) the school gate. We used this anonymised data in conjunction with causal analysis to better understand the factors under the control of the Directorate that influence the educational outcomes of students. It was a two-stage project.

Gradient staff reviewed existing scientific literature exploring the factors that underlie student success, guided by the Education Directorate experts. After data processing, performed some preliminary causal modelling of the relationships between both school and non-school factors and educational outcomes in the ACT with a focus on identifying factors that lead to children failing to thrive in education.

In parallel, HMI staff identified value-laden questions and assumptions as they arose throughout the project. HMI, in conversation with the education department, helped in the identification and unpacking of the theoretical questions and frameworks needed to inform the discussion of interventions. For example: What does ‘success’ in this area mean? How should interventions be prioritised? Which subgroups should be the target of interventions: the ones who will improve most easily? The ones who are currently doing the worst? The ones with the biggest gap between potential and achievements?