Evaluating Program Alignment Effects on Teacher Emotional Exhaustion: Machine Learning Versus Conventional Parametric Mediation Models

• In Event: Emotion, Belief, and Relational Elements of Teaching and Teacher Preparation

Abstract

Emotional exhaustion represents a critical early-career teacher risk factor. We applied and compared two mediation approaches—targeted learning framework using ensemble machine learning algorithms (ML) and Structural Equation Modeling (SEM)—to estimate the effect of program alignment on emotional exhaustion among 508 U.S. pre-service teachers. Both methods decompose effects through mediators: subject-specific knowledge and preparation for diversity. Results showed comparable total effects of −0.26 (95% CI: [−0.50, −0.02]) in ML and −0.35 (CI: [−0.66, −0.01]) in SEM, though ML attributed more to indirect effects (−0.15; CI: [−0.23, −0.07]) than SEM (−0.11; CI: [−0.23, −0.03]). Program alignment plays a key role in supporting teachers’ emotional well-being, while result disparities suggest more complex, nonlinear interactions best captured by data-adaptive models.

Authors

• Jean Baptiste Habarurema, University of Cincinnati

• Benjamin Kelcey, University of Cincinnati

• Amota Ataneka, University of Cincinnati

• Leigh McLean, University of Delaware

• Peter A. Youngs, University of Virginia