What does a student's history add to forecasts of their university trajectory? A new arXiv preprint presents an "information-timing audit" that isolates the predictive contribution of historical summaries in a rule-based university trajectory simulator. The work is framed around calendar controls and the timing of information, asking whether past plans and past states matter beyond what current plans already reveal.
The study runs seventeen ridge specifications plus a persistence baseline, separating inputs into current plans, past-state summaries, past-plan summaries, and calendar-only features. This setup is designed to decompose the predictive value of each information type, rather than simply comparing a full model against an empty one.
The abstract does not include findings, so the outcome of the audit is not yet visible from the snippet. The contribution appears to be methodological: a framework for testing when and how historical data should be incorporated into trajectory forecasting models.