All guidesTeacher Insights
Teacher Insights
AI student: Where the class struggles, and what was taught but never tested.
Teaching adapts only when teaching can see. Insights gives teachers class-level views built from the same evidence the learners’ twins hold — without ever exposing an individual learner.
What it is
The coverage auditor crosses the concept graph against the assessment graph: concepts that were taught but never assessed, and assessment edges pointing at concepts nothing teaches. Each concept carries a class-average mastery chip.
The confusion radar is an eight-week timeline of where friction happened — misconceptions surfacing, discussion volume, quiz misses — so a teacher can see which weeks the class wobbled and what to re-teach now.
Why — the pedagogical rationale
The most actionable teaching signal is a gap between intention and evidence: a concept you taught twice but never assessed is invisible in a gradebook, and a week where misconceptions spiked is invisible a month later. Both need the concept-level view, not the score level.
Privacy is the hard constraint: a teacher who could read individual twins could read individual struggles. Every insight is an aggregate over the class — the pattern is visible, the person is not.
How we achieve it
Both views are pure SQL aggregates over the twin’s evidence and misconception stores — no LLM, no per-learner rows crossing the boundary, owner-or-admin gated. The radar buckets by the same Monday-anchored weeks the learners’ plans use, so the two sides of the conversation share a calendar.
Where to find it
Teacher Insights, per course: the twin section, the coverage auditor, and the confusion radar sit after the engagement overview.