Open data, private learners: a de-identified student activity and performance dataset for learning analytics

Key findings from this study

• The study presents a de-identified clickstream dataset from three academic years of first-year bachelor instruction suitable for public research use.
• The authors report that comprehensive documentation of de-identification procedures accompanied the dataset release to ensure methodological transparency.
• The researchers demonstrate that privacy and utility validation confirmed both adequate learner protection and retained analytical value.

Overview

Learning Analytics research increasingly depends on digital trace data from learner interactions with educational platforms. However, sensitive personal information embedded in such datasets raises privacy concerns that inhibit open science practices. The authors present a de-identified clickstream dataset from two first-year bachelor courses at KU Leuven spanning three academic years. The dataset enables LA solution development while maintaining learner privacy through transparent de-identification documentation and empirical validation.

Methods and approach

The researchers collected clickstream data across three consecutive academic years from students enrolled in two first-year bachelor courses. De-identification procedures were systematically applied to remove or transform personally identifiable information. The authors then conducted privacy and utility validation assessments on the resulting dataset. Validation examined both the effectiveness of anonymization techniques and the dataset's continued value for analytics research.

Results

The study demonstrates that comprehensive de-identification documentation accompanies the public dataset, enabling reproducibility and transparency in the anonymization process. Privacy validation confirmed that de-identification procedures adequately protect learner identity. Utility validation established that the anonymized dataset retains sufficient information for developing and evaluating Learning Analytics models and frameworks. The authors report that this balance between privacy protection and analytical utility enables collaborative research across institutional boundaries.

Implications

Open datasets with rigorous de-identification support the democratization of Learning Analytics research. Institutions can contribute data to collective research efforts without compromising learner privacy through transparent methodological documentation. The validation approach presented offers a replicable model for other educational contexts considering public data release. Establishing such datasets may accelerate framework development and cross-domain knowledge transfer in the Learning Analytics field.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.