Privacy-Preserving Distributed Link Predictions Among Peers in Online Classrooms Using Federated Learning

Federated Learning enables privacy-preserving analysis of student interactions: The paper demonstrates that machine learning models can predict student interactions across multiple classrooms without requiring raw student data to be centralized on a single server, addressing a key privacy concern in educational settings.

Collaborative training improves prediction accuracy across classrooms: By using federated learning to train models across multiple classrooms simultaneously, the approach identifies common interaction patterns that wouldn't be visible when analyzing each classroom in isolation, leading to better predictions of future student connections.

Model personalization balances shared insights with classroom differences: The framework uses personalization techniques to adapt the collaborative model to each individual classroom's unique dynamics, recognizing that student interaction patterns may vary between different learning environments while still benefiting from cross-classroom learning.

Privacy-preserving ML can enhance educational outcomes: Rather than requiring schools to share sensitive student interaction data, this approach allows educational institutions to collaborate on improving predictive models for social learning networks while maintaining data privacy and institutional autonomy.