SRS Review Signals Beyond a Completion Checkbox

Watching a lesson and recalling it days later are different events. A useful review ledger stores the evidence needed to tell them apart instead of promoting both to one “completed” state.

Start with a transparent interval ladder

The baseline schedule uses D0, D1, D3, D7, D14, D30, D60, D90, D180, and D365. Again resets the stage, Hard keeps a conservative interval, Good advances one step, and Easy advances two. The function returns both the new stage and the next local review date.

Store signals before estimating mastery

The review row can record rating, response time, hints, modality, and transfer performance. Practice attempts and incorrect answers live in separate ledgers. This keeps raw evidence available if the mastery formula changes.

An early system should not display a confident percentage from one answer. It can instead show a range and lower confidence when sample size is small, delayed recall is missing, or hints were used.

Verify transitions and idempotency

Automated tests cover the interval ladder and confirm that retrying the same review request does not create a second record. The temporary D1 run stored an initial review and an identical retry with one resulting row and the expected next date. The model remains a scheduling policy, not a clinical or psychological assessment; it should be recalibrated only with consented learning outcomes and documented uncertainty.