# Learner Variability and Agency

Version: 2026-07-22.3  
Status: design plan requiring practitioner and learner validation

## Adjustable learning features

- Shift+H changes future mentor support: automatic, more guidance, or more challenge.
- P changes future lesson pace: slow, standard, or fast.
- V skips the active AI recommendation without an evidence penalty.
- H shows or hides responsive tutorial hints.

The canvas is keyboard focusable, mentor messages have a live text equivalent,
and reduced-motion styling is provided. Terrain editing uses left-click/right-
click on desktop and dedicated hold buttons on touch devices; Q and E are
intentionally unbound. A remappable keyboard terrain-editing method and human
accessibility testing are still required.

## Automatic challenge bands

There is no level selector. The mentor uses three task profiles: middle-school, high-school, and college. A conservative evidence policy changes the profile only after completed mentor assessments. It requires repeated success across concepts before moving upward and repeated need for scaffolding before moving downward.

The system does not infer age, actual school placement, intelligence, diagnosis, or mastery. It excludes movement speed, edit speed, pointer use, keyboard use, and protected attributes from band changes. The selected band may still be wrong; the Shift+H support control and V skip control provide immediate user agency.

## Whole-learner limits

The current model represents interaction preferences and observed task performance only. It does not claim to model motivation, culture, language background, disability, social context, emotional state, prior schooling, or interests outside gameplay. Expanding the model requires consent, necessity, privacy review, co-design, and evidence that the change benefits learners.

## Feedback and co-design evidence still needed

Digital Promise review requires evidence that diverse learners and educators influenced the design and that findings changed the product. Keep dated recruitment methods, session protocols, participant-role descriptions, accessibility accommodations, findings, product decisions, release notes, and follow-up validation. Do not collect more personal data than the study needs.