HAX UbD Stage 2 — Thinking Like an Assessor

Build the scrapbook of acceptable evidence for a unit: frame at least one GRASPS performance task, map the six facets to rubric criteria, ensure at least one genuine problem (not just exercises), and run the validity self-test on every assessment. This is the most distinctive UbD skill — it reverses the natural instinct to design activities first.

The core move: think like an assessor, not an activity designer. An assessor asks "what sufficient, revealing evidence of understanding do I need?" An activity designer asks "what fun things will we do?" The first column produces aligned evidence; the second produces the apples unit (fun, but no understanding target).

When to Use

Trigger conditions:

• "What evidence do I need" / "design a performance task" / "build a GRASPS task"
• "What assessments prove understanding" / "write a rubric for understanding" / "is this assessment valid"
• You are hax-ubd-backward-design and have reached step 4 of its methodology

When NOT to use (with redirect):

• Fun-activity ideation with no evidence target → that is the activity-designer trap; refuse and redirect to Stage 1 first
• Page-level chunking → hax-content-chunking-audit
• Sequencing activities → hax-ubd-stage3

Scope

Mutating. Writes manifest.stage2 + assessment pages + editable-table rubrics. Does NOT sequence (Stage 3) or set goals (Stage 1).

Inputs

• manifest.stage1 — required. Stage 2 is derived from goals; do not invent evidence without targets.

Methodology

1. Map the scrapbook across the assessment continuum (UbD Fig 7.4/7.5): performance tasks (T) + academic prompts + quiz/test items + informal checks (all OE). Aim for variety + longitudinal spread — a scrapbook, not a single snapshot.

2. Frame ≥1 core GRASPS task (UbD Fig 7.7 prompts): Goal, Role, Audience, Situation, Performance/Product, Standards. The task must be authentic: realistically contextualized, requires judgment/innovation, asks the student to "do the subject," messy (constraints/noise/audience), criteria known in advance, rehearse-able with feedback. (Full prompt bank: hax-ubd-grasps companion.)

3. Problem vs. exercise check (UbD Fig 7.6): confirm ≥1 task is a genuine problem — no cues about how to frame/solve it, multiple approaches possible, realistically "noisy" setting, justified solution (not a single plugged-in answer). Exercises plug in on cue; problems require deciding which knowledge to use. Mark it isProblemNotExercise: true.

4. Six-facet rubric blueprint (UbD Fig 7.9): for each targeted understanding, pick the facet(s) that yield the strongest evidence and build editable-table analytic-trait criteria. Verbs per facet:

   • Explanation → explain/justify/verify/prove
   • Interpretation → interpret/make-meaning/tell significance
   • Application → apply/transfer/use in new context
   • Perspective → critique/compare/question assumptions
   • Empathy → role-play/relate
   • Self-Knowledge → self-assess/reflect Include at least one impact criterion (effect on audience / real-world consequence), not just correctness + craft. (Full facet lens: hax-ubd-six-facets companion.)

5. Validity self-test (../hax-ubd-backward-design/references/validity-self-test.md): run the 7 questions on every proposed assessment. Target: all "very-unlikely." Flag any "likely" as a defect and apply the matching redesign pattern before accepting.

6. Reliability check: ensure multiple measures over time (progress-donut / simple-progress / promise-progress + grade-book), not one high-stakes event. One valid task can still be anomalous; a pattern of evidence justifies the inference.

HAX Component Map (Stage 2)

Evidence type	Component	Notes
GRASPS performance task	authored page + video-player/media-playlist/audio-player/image-compare-slider (authentic context) + simple-fields (product) + editable-table (rubric/standards)	the "do the subject" page
Academic prompts	stop-note / self-check + simple-fields (constructed response)	no essay element → simple-fields
Quiz & test items	multiple-choice, fill-in-the-blanks, matching-question, sorting-question, tagging-question	selected/closed response only
Informal checks	self-check, flash-card, stop-note	ongoing, ungraded
Six-facet rubric	editable-table (analytic-trait)	no rubric element → editable-table
Longitudinal evidence	progress-donut, simple-progress, promise-progress, grade-book	"scrapbook not snapshot"
Perspective evidence	image-compare-slider, a11y-compare-image	compare viewpoints
Portfolio exhibit	clean-portfolio-theme, glossy-portfolio-theme	for Exhibit stage

Full map (all stages): ../hax-ubd-backward-design/references/ubd-element-map.md.

Output Format

# Stage 2 — Acceptable Evidence: {unit}
## Evidence scrapbook
- T (performance): {GRASPS summary} → page {slug}; facets: {...}; problem-not-exercise: yes; targets: {U}
- OE (academic prompts): {...} → stop-note + simple-fields
- OE (quiz/test): {...} → {components}
- OE (informal checks): {...} → {components}
## GRASPS task {id}
G: {...} | R: {...} | A: {...} | S: {...} | P: {...} | S (standards): {...}
## Rubric (facet-mapped)
| Criterion | Facet | Level 4 | Level 1 |
## Validity self-test
{7 questions → answers; defects + redesign}
## Reliability
{longitudinal evidence plan: components + cadence}
## HAX handoff
{page + component list per evidence item}

Acceptance criteria

• ≥1 GRASPS task with all 6 letters filled and isProblemNotExercise: true.
• Every performanceTasks[*].targetsUnderstanding exists verbatim in manifest.stage1.understandings (Stage 1 ↔ Stage 2 alignment invariant).
• Rubric criteria each map to a named facet; ≥1 criterion is an impact criterion, not just correctness/craft.
• Validity self-test: all 7 answers "very-unlikely" (or defects logged with a redesign plan).
• Selected-response items are NOT the only evidence of understanding for any targeted U — every U has ≥1 performance-based evidence source.
• longitudinalEvidence is non-empty (reliability: scrapbook, not snapshot).

Gotchas

• "Interesting project" ≠ valid evidence (UbD Fig 7.6). A fun task that doesn't reveal the targeted understanding is a trap — redesign or drop it. Validity is about what you can infer, not whether the task is engaging.
• Don't let multiple-choice carry understanding alone. Selected response is "insufficient and sometimes misleading" for understanding. It rounds out the scrapbook for K/S; it cannot be the sole evidence of an enduring understanding.
• Authenticity is about context + judgment + "doing the subject," not just "hands-on." A cooking activity can be hands-on and inauthentic if it doesn't reveal the targeted science understanding.
• A valid indirect test can still be inauthentic; an authentic task can still be invalid. Run the self-test regardless of format.
• Don't duplicate the rubric's craft criteria as understanding criteria. A lovely product with limited understanding (self-test Q4) is the classic validity failure — keep understanding criteria separate from and weighted above presentation.

Dependencies

• Reads: manifest.stage1 (required); grad-blooms (facet ↔ cognitive level)
• Hands off to: hax-ubd-stage3 (which reads manifest.stage2)
• Companions: hax-ubd-grasps (GRASPS prompt bank), hax-ubd-six-facets (facet lens + rubric criteria)
• Uses: hax-design-system (DDD tokens on rendered components)

References

• ../hax-ubd-backward-design/references/manifest-schema.md — stage2 fields
• ../hax-ubd-backward-design/references/ubd-element-map.md — Stage 2 component map
• ../hax-ubd-backward-design/references/validity-self-test.md — the 7-question self-test + defect→redesign patterns
• ../hax-ubd-backward-design/references/design-standards.md — standards 5–9
• PRAW RULES.md: ~/Documents/git/haxtheweb/praw/RULES.md
• Source: Wiggins & McTighe (2005), UbD Ch. 7–8.