AI Tutor output artifacts and progress insight policy

DEC-0085 - AI Tutor output artifacts and progress insight policy

Context

The expected AI Tutor experience includes more than text chat: mini exercises, structured plans, and visual-style outputs (mindmap/chart/progress explanation). Current docs do not lock these output categories.

Decision

AI Tutor output artifact baseline:

• concise_answer: short explanation and clarification.
• guided_checklist: step-by-step learning plan.
• mini_exercise: immediate small practice task in chat.
• diagram_or_mindmap_ready: structured output suitable for mindmap/diagram rendering.
• progress_insight: learner-progress explanation with timeframe and metric context.

Progress insight rule:

• Progress/analytics explanation is on-demand (user-requested), not forced.
• Output must include metric window context (for example today/7-day/30-day where available).

Action continuity rule:

• Each non-trivial output should include at least one actionable follow-up:
  • continue in chat, or
  • deep link/CTA to relevant platform exercise/resource when available.

Decision Value

• Converts AI Tutor from generic chat to actionable learning workspace.
• Improves continuity between explanation and practice action.
• Makes output behavior testable for UX and QA.

Rationale

A locked output taxonomy prevents inconsistent response styles and enables systematic UX mapping for each artifact type.

Implications

• Product/UX impact:
  • AI Tutor UI should support artifact-specific rendering states.
• Data/logic impact:
  • response pipeline needs artifact type tagging and CTA routing metadata.
• Operational impact:
  • analytics can track artifact-level usefulness and completion follow-through.

Alternatives considered

• Option A: one generic text response format.
• Option B: explicit multi-artifact output taxonomy with action continuity (selected).

Open follow-ups

• Resolved by DEC-0090: layered rendering fallback contract (visual_native -> structured_graph_text -> guided_checklist_fallback).

Decision quality check

Decision quality check: DEC-0085

• Score: 10/12
• Weak dimensions: Decision testability (1), Operational feasibility (1)
• Action: Promoted with rendering-fallback follow-up.