Modeling Explainable Human-AI Collaboration for Enhancing Instructional Decision Support System Effectiveness

Abstract

The integration of artificial intelligence (AI) in education has transformed learning environments into data-driven instructional decision support systems (DSS). However, system effectiveness is contingent not only on technical capability but on how users interpret, evaluate, and implement AI-generated outputs. Grounded in the DeLone and McLean IS Success Model and Task-Technology Fit (TTF) theory, this study models explainability-driven human–AI collaboration as a predictor of instructional DSS effectiveness. Three core dimensions are examined: Human Oversight Capability (HOC), AI Interaction Competence (AIC), and Perceived Algorithmic Explainability (PAE), as predictors of Human-AI Collaborative Teaching Effectiveness (HCTE). A quantitative cross-sectional study was conducted with 316 teachers across Indonesia. Multiple regression and Classification Regression Tree (CRT) analyses were employed. Results indicate that PAE is the strongest predictor (β = 0.567), followed by HOC (β = 0.182) and AIC (β = 0.125), with the model explaining 65.0% of HCTE variance (R2 = 0.650, p < 0.001). CRT confirms that human–AI collaborative application capability is the highest-leverage indicator. These findings advance an IS-grounded PHI framework (PAE → HOC → AIC) positioning explainability as the foundational cognitive layer of effective AI-enabled instructional decision support.