Heritage and Identity

Forty scripted lessons for Ontario Grade 2 Social Studies, Strand A — Changing Family and Community Traditions — published as a 191-page offline website. Alongside the lessons sit 40 projectable decks, 91 printables, 8 games, 10 assessments on a single knowledge / thinking / application / communication frame, and a curriculum map that quotes all 17 Strand A expectations word for word. The unit ends in a nine-lesson Identity and Heritage Museum whose exhibits are curated from artefacts the children already made, in Lessons 2, 3 and 4, 16, 17, 23 and 30.

Heritage units reach teachers as a set of activities with no stated sequence and no defined endpoint. Because the culminating task is not fixed at the start, it gets planned late and runs as a standalone craft week with no connection to the ten weeks before it. Work the children produced in October contributes nothing to it. The cost is a unit that fills classroom time without accumulating into anything a teacher can assess. The requirement was a unit where the final task is specified first and every lesson before it produces material that task consumes.

Grade 2 classroom teachers working to the Ontario curriculum — including teachers meeting Strand A for the first time, and anyone who has to pick a lesson up cold and run it that morning.

• Curriculum architecture — the unit structure, the culminating performance and the sequence that feeds it
• Lesson writing — all 40 lessons, scripted with timings, teacher moves and differentiation
• Assessment design — 10 assessments on one shared frame, with the criteria written for children
• Interaction and visual design — the site, the decks, the printables and the eight games
• Production design — one definition per lesson, so the lesson page, the deck and the printables cannot drift apart

What the design set out to achieve. These are objectives, not results.

• Design backward from a culminating performance so that every earlier lesson has a reason to exist
• Script each lesson closely enough that a teacher can run it without preparation the night before
• Put the verbatim expectation beside the lesson rather than in an appendix nobody opens
• Make the unit localisable, so the community studied is the buying teacher’s own
• Ship as offline HTML that opens on a locked-down school device with no network and no install

Each lesson is defined once rather than written three times. Its learning goal, success criteria, minute-by-minute timings, differentiation notes, verbatim expectation and assessment note live in one place, and the lesson page, the projectable deck and the printables all come from that one definition. Three artefacts, one source, so they cannot say different things after a revision. The unit is delivered as a folder of ordinary web pages that opens from a file, a memory stick or the hosted copy — no install, no accounts, nothing that breaks when a school network blocks a domain.

AI sits in the making of this unit, not inside it. Nothing here calls a model at runtime, no child’s work is sent anywhere, and the site would behave identically with the network unplugged. During the build, AI was used the way a fast, tireless drafting assistant is used: expanding a fixed lesson structure into narration, turning a scripted lesson into a first-pass slide sequence, and reading across all 191 pages for inconsistency at a scale that hand-checking does not reach. It was never asked what the unit should be about, where it should end, or which expectation a lesson meets.

The decisions a model cannot be handed are the ones that decide whether the unit works: which culminating performance is worth building 40 lessons toward, which artefacts genuinely feed it, where a seven-year-old’s attention actually runs out, and the difference between an expectation a lesson meets and one it merely touches. Every expectation-to-lesson mapping was checked against the Ministry wording by hand. So was the decision that the localised, community-agnostic version is the pedagogically correct one — a judgement about the curriculum, not about the market.

Backward design is easy to state and structurally awkward to build. It means committing to a final performance before you know whether the lessons in between can supply it, then going back and rewriting lessons whose artefacts turn out not to fit the exhibit they were supposed to fill. The second difficulty was drift: a lesson plan, a slide deck and a printable that say slightly different things are worse than any one of them alone, because the teacher discovers the mismatch mid-lesson. Defining each lesson once rather than writing it three times is the answer to that, and it constrains the writing as much as it protects it.

The record of the iteration is in the artefact rather than in a changelog: the mapping between each museum exhibit and the lesson that produces it is exactly what the revision passes produced. Committing to the museum first forced a sweep back through the sequence to confirm that every exhibit had a lesson that actually made the thing, and the lessons that could not supply their exhibit were the ones that changed.

The strongest thing about this unit is structural, which means it is invisible in any single screenshot — you cannot see backward design in a picture of a lesson. That is a genuine presentation problem and it is why this case study leans on the sequence rather than the pages. What I cannot tell you is what it did in a classroom. No learning measurement was designed into it and none has been collected, so there is nothing here about outcomes, only about intent. If I built it again I would add one thing: a short teacher-facing instrument that captures what the museum revealed, so the unit produces evidence about itself.

• Backward design
• Ontario curriculum mapping
• Lesson scripting
• Assessment design
• Content-model design
• Product localisation
• Offline-first delivery