Social predictors of inquiry learning climate in physics education: Insights from explainable machine learning

Understanding the social factors that shape the learning climate in physics education contexts that rely on inquiry-based approaches is increasingly vital. This study aimed to model and predict the most important social aspects, namely, Instructor Support, Peer Support, and Interpretation, that contribute to students' perceptions of the Inquiry Learning Climate (ILC). Three hundred twenty-two Indonesian high school students (aged 15–18) participated in the study. Data were analyzed using a combination of Spearman correlation, the Boruta feature selection algorithm, Fast-and-Frugal Trees (FFT), and SHapley Additive Explanations (SHAP). Results revealed that all social factors significantly correlated with students' ILC perceptions, with Interpretation emerging as the strongest predictor. The FFT model also identified Interpretation as the most informative initial cue in classifying students' perceptions. At the same time, SHAP analysis confirmed that mutual understanding, especially students’ perceptions of being understood by teachers and peers, played a central role in shaping the learning climate. These findings underscore that the success of inquiry-based instruction in physics is deeply contingent on the quality of two-way communication in the classroom. The study contributes theoretically and practically by offering an interpretable machine learning approach to identify actionable social predictors that support the development of socially responsive and inquiry-oriented learning climates.