A machine learning analysis of school-level drivers of organisational innovativeness

Abstract This study applied explainable machine learning (ML) to identify school-level conditions that predict organisational innovativeness using OECD TALIS 2018 principal data from 47 education systems. Six machine learning algorithms were compared, and XGBoost achieved the best predictive performance. SHAP (SHapley Additive exPlanations) was then applied to interpret the model and rank the most influential factors. The results show that organisational innovativeness is shaped primarily by relational and cultural conditions rather than by isolated structural inputs. School climate emerged as the strongest predictor, particularly the encouragement of staff-initiated projects, mutual support, shared responsibility, participatory decision-making, and cooperation with the local community. Shortage of time with students, by contrast, reduced predicted innovativeness. Principals’ job satisfaction was the second most influential factor. Leadership practices that promote teacher cooperation for pedagogical development showed a modest positive association. Diversity beliefs among teachers and access to mentoring programmes contributed small but generally positive signals. Overall, the findings suggest that organisational innovativeness in schools depends mainly on a supportive and participatory school climate, principals’ job satisfaction, instructional leadership practice, diversity belief and structured mentoring.