Data

• Clickstream, assessment and LMS data: what each can and cannot evidence
• Consent, minors, and institutional review requirements
• Anonymisation limits when timestamps and trajectories re-identify students

Methods

• Knowledge tracing and mastery estimation
• Clustering learner trajectories and the instability of unsupervised groupings
• Natural language processing on open responses and reflective writing
• Text mining literature at scale for systematic reviews

Rigour

• Algorithmic bias against under-represented learner groups
• Overfitting to a single cohort, institution or platform
• Pre-registration, open materials and reproducible analysis pipelines

Publication

• Reporting effect sizes and uncertainty rather than model accuracy alone
• Writing methods sections that another researcher could actually rerun
• Translating findings into practice without overclaiming causality