Explanatory Machine Learning Methods for Teacher Effectiveness

• In Event: Machine Learning Methods in Educational Research

Abstract

We develop and apply explanatory machine learning methods for delineating teacher effectiveness and compare them to each other and the general linear model. Methodologically, we integrate machine learning methods with the general linear model through surrogate regressions to produce methods with high predictive efficacy and high interpretability. Substantively, we examine the extent to which we can we develop machine learning architectures to identify and explain teacher profiles, pathways and practices that produce student learning. A core hypothesis was that explanatory machine learning methods will be able to generate unique insights on what drives student learning in different contexts and provide actionable avenues for improving theories, assessments, surveys, teacher development plans, and the quality of evidence generated through studies.

Authors

• Benjamin Kelcey, University of Cincinnati

• Amota Ataneka, University of Cincinnati

• Yanli Xie, Florida State University

• Fangxing Bai, Montana State University