Data-driven peer recommendation in higher education: A pilot study on academic reading

Abstract

Collaborative learning in tertiary education faces challenges such as limited teacher intervention and effective student pairing. This study addresses these issues by proposing a data-driven peer recommendation approach enhanced with learner profile visualisation. The system dynamically matches students based on evolving learning profiles, using an open learner model to improve transparency and decision-making. Implemented in a Japanese university, a pilot study in an academic reading course showed that peer feedback improved report scores, with visualisation aiding in selecting suitable peer reviewers. Comparisons across three recommendation rounds suggested that integrating recursive data accumulation strengthened personalised peer recommendations and encouraged greater participation. By demonstrating the workflow of peer learning implementation, this research also highlights the broader potential of data-driven systems to support collaborative learning in higher education.

Implications for practice or policy:

• Peer review activities with clear criteria can help students revise and improve writing, even within limited rounds.
• Data-driven analytics and recursive evidence accumulation can enable personalised peer recommendations while ensuring inclusivity by mitigating algorithmic biases.
• Visualised reviewer profiles via open learner models may enhance perceived feedback quality and student agency, though more controlled validation is needed.
• Policymakers could support flipped learning models that leverage data-driven personalised recommendations to enhance learner autonomy and peer collaboration.