Learning analytics · Python · Machine learning · Assessment integrity
Organization: Independent project
Role: Designer and developer
Situation
Many LMS platforms process SCORM assessment files in a way that lets a learner retrieve correct answers and retake questions without score penalty. The result is grade inflation that goes undetected until someone reviews results manually.
Analysis
Exploitation leaves a behavioral signature in the SCORM interaction data — patterns of wrong-answer repetition and retry timing that don’t occur in honest attempts.
Design
A Python-based system that ingests SCORM reports, scores each attempt against per-question baselines, and flags likely exploitation. The current version uses a vectorized pipeline with robust latency parsing and a supervised-model validation harness. Testing identified repeated-wrong-answer counts as the strongest discriminative features.
Result
An auditable tool that corrects grades to reflect actual learner performance. Built and tested entirely on synthetic data — question banks, employee records, and realistic SCORM reports generated by an included script. None of the data is drawn from any employer’s assessments.
Deliverables
- GitHub repository: scorm_check (includes the synthetic data generator)
- Flagged-attempt output example — coming soon
