An extended community of inquiry framework for monitoring and predicting online peer learning participation

Purpose Online learning communities on social media platforms can support peer learning, but educators often lack theoretically grounded and measurable approaches for monitoring how participation and discourse evolve across a semester. This study proposes an extended Community of Inquiry (CoI) evaluation framework that integrates Social, Teaching, and Cognitive Presence with a fourth behavioural dimension, Student Presence. Design/methodology/approach A sequential exploratory mixed-method design was adopted. Qualitative analysis of prior literature and semester-long observations of two large first-year engineering course Facebook groups (each enrolling 800–1000 students) informed an indicator-based coding scheme, applied quantitatively over Weeks 1–13. Predictive modelling used a persistence baseline, a multi-output Random Forest, and a multilayer perceptron under time-aware evaluation protocols. Findings Social Presence was enquiry-driven and peaked in Weeks 3–4; Teaching Presence was frontloaded and primarily reactive; Cognitive Presence was shallow, dominated by remembering and analysing. Student participation was consumption-oriented, with observers consistently outnumbering posters. Random Forest achieved consistent poster prediction (R2 ˜ 0.48–0.49), while observers and non-members remained difficult to forecast due to structural interdependence. Permutation importance identified remembering and evaluating as the most influential cognitive predictors. Research limitations/implications The dataset comprises 13 weekly observations from a single platform and institution, limiting generalisability. Future work should collect multi-cohort data, introduce lagged predictors, and explore individual-level modelling. Practical implications The framework provides instructors with an early-warning system for low poster activity, enabling timely, evidence-based interventions to support online peer learning communities. Originality/value This study makes three contributions: a multi-dimensional coding scheme grounded in the extended CoI framework; a data-driven analytics pipeline enabling descriptive monitoring and predictive modelling of participation roles; and an integrated evaluation framework that combines theory-grounded indicator coding with transparent machine learning to produce actionable insights from social media learning data.