Skip to content

SOP-050 — Learning design

Guide

<!-- last-reviewed: 2026-06-01 · owner: Design Research (Instructional Designer) -->

SOP-050 — Learning design

SOP · owner: Design Research (Instructional Designer). Layer: SOP/Runbook (docs/canon.md). The fixed lesson structure (HOOK → PIT NOTES → MAP → EXPLANATION → EXAMPLES → CHECKS) lives in docs/concept.md — not restated here. Activation note (2026-06-01): this role activates at platform Build start, which has not begun (product data layer deferred, ADR-0006). The procedure is defined and stub-first; it becomes live at Build, and its efficacy-criteria step is sequencing-locked behind the DA Director’s scoring rubric (which must exist first). docs/platform/ does not exist yet and is not created empty.

Interface Contract

  • Inputs: lesson structure + scoring rubric ← from <Data & Analytics Director> · trigger: document changed
  • Outputs: framework set (objectives / assessment / scaffolding / efficacy) → to Design Research Director · trigger: review requested
  • Outputs: difficulty/objective metadata → to AI Systems Director · trigger: requirements issued
  • Outputs: efficacy criteria + calibration recommendation → to Data & Analytics Director · trigger: requirements issued
  • Outputs: finalized framework → to Documentarian · trigger: document changed
  • Outputs: framework set → to Archivist · trigger: artifact approved
  • Handoff trigger (in): platform Build begins (or a new lesson type / material rubric change), and the lesson structure + scoring rubric are available to design against
  • Handoff trigger (out): the framework set is signed off by the Director and the metadata/criteria are issued to AI Systems and DA
  • Definition of Done before handoff:
    • [ ] Each learning objective written at a named Bloom’s level (correct action verb; “understand” is not a Bloom’s verb)
    • [ ] Objectives mapped to lesson-structure sections (HOOK…CHECKS)
    • [ ] Recall-vs-understanding filter applied to every item type, result documented
    • [ ] Difficulty tiers + classification rubric operationalizable (not “feels hard”)
    • [ ] Efficacy criteria evaluable (named behavior, interval, method, threshold) — only after the rubric exists
    • [ ] Every pedagogical claim cites a learning-science basis
    • [ ] Director sign-off obtained (member ships nothing unsigned)
    • [ ] Learnings captured to docs/knowledge/ if any
    • [ ] Framework handed to the Archivist (0 drift)

Triggers reused (document changed, review requested, requirements issued, artifact approved) all exist in docs/handoff-triggers.md — no new trigger is proposed (reuse-beats-add; requirements issued fits the spec-to-implementer handoff to AI Systems / DA). The requirements issued outputs to the AI Systems Director and the DA Director are absorbed by those Directors’ requirements issued fan-in sinks (SOP-064, SOP-060; §3.1 R1 — matched); the inbound rubric/structure from the DA Director is consumed via a placeholder-producer sink. The Design-Research-Director edge reconciles as pending (not an Interface-Contract identity here, §3.1.1 R3).

Purpose

Produce the pedagogical backbone — learning objectives, assessment specs, difficulty scaffolding, and efficacy criteria — so the platform’s rubric measures genuine learning, not completion, and the AI assembly system has legible objective and difficulty metadata to honour.

When to use

At platform Build start; when a new lesson type is added; when the rubric weights change materially; or when accumulating learner-outcome data contradicts the efficacy criteria. Not used before Build (the role is activation-gated).

Flow

flowchart LR
  A["Read lesson structure<br/>+ scoring rubric"] --> B["Define objectives<br/>(Bloom's-tagged)"]
  B --> C["Design assessment<br/>(recall-vs-understanding filter)"]
  C --> D["Specify difficulty<br/>scaffolding"]
  D --> E["Write efficacy criteria<br/>(rubric must exist)"]
  E --> F["DRD sign-off →<br/>issue to AI Systems / DA"]
  classDef s fill:#1d4ed8,color:#fff,stroke:#1e3a8a
  class A,B,C,D,E,F s

Text fallback: read the fixed lesson structure and the DA Director’s scoring rubric → define learning objectives at named Bloom’s levels and map them to lesson-structure sections → design assessment specs and apply the recall-vs-understanding filter to every item type → specify difficulty tiers, an operationalizable classification rubric, and sequencing rules → write evaluable efficacy criteria (only once the rubric exists) → obtain Director sign-off, then issue metadata to AI Systems and criteria/calibration to DA.

Steps

  1. Orient. Read docs/concept.md (the fixed lesson structure) and the DA Director’s scoring rubric. Do not produce a framework without first reading the existing structure, rubric, and assembly logic — reinventing is waste, contradicting silently is a violation.
  2. Define objectives. Per lesson type: name it, write the objective at the appropriate Bloom’s level (action-verb format, level named explicitly), map it to the structural sections that scaffold/deliver/assess it, and specify the success evidence. Write to docs/platform/learning-objectives.md.
  3. Design assessment. Per CHECKS component: select item format(s) justified against the objective; write item specifications (stem template, distractor rules, passing criterion); apply the recall-vs-understanding filter to every item type and document the result. Write to docs/platform/assessment-spec.md.
  4. Specify scaffolding. Define 3–5 difficulty tiers with precise entry/exit competencies; produce an operationalizable chunk-classification rubric; specify sequencing rules and re-entry paths. Write to docs/platform/difficulty-scaffolding.md.
  5. Write efficacy criteria. For each rubric metric, write an evaluable criterion (named behavior, interval, method, threshold); flag proxy-vs-outcome mismatches. Gate: this step requires the rubric to exist — if it does not, stop and flag the sequencing lock to the Director. Write to docs/platform/learning-efficacy-criteria.md.
  6. Sign off and issue. Obtain Director review (member ships nothing unsigned). Issue difficulty/objective metadata to AI Systems in an encodable form; issue efficacy criteria + any reports/rubric-calibration-<YYYY-MM-DD>.md to DA.
  7. Hand off finalized frameworks to the Documentarian (canon) and the Archivist; capture learnings.

Roles / approvers

Run by the Instructional Designer. The Design Research Director signs off before anything is finalized or issued. The DA Director owns rubric implementation (ID provides criteria only); AI Systems owns the assembly pipeline (ID provides encodable metadata).

Guardrails

  • Define criteria, not weights — never touch rubric weights/formulas; that is DA’s.
  • Bloom’s precision — name the level; use the correct verb; precision determines what assessment is appropriate.
  • Recall-vs-understanding filter — an item answerable by surface pattern-matching fails and is redesigned.
  • Evidence over assertion — every pedagogical claim cites a learning-science basis.
  • No content, no code — specify the framework; do not author chunks or write code.
  • No personal learner data — anonymise/aggregate; personal data is PA-only.
  • Do not restate docs/concept.md’s lesson structure — link to it.

See also: docs/concept.md · docs/units/design-research/charter.md · SOP-049 (gate evidence) · docs/handoff-triggers.md.