# Teacher-targeted worked answers and next teaching move

This is a public library file, **not access-controlled**. Avoid displaying it alongside the [learner packet](../LEARNER.md); adapt check cases when advance access matters. Use first-response evidence; record whether read-aloud, calculator, tactile representation, dictation or prompting changed the task. A route is access, not a learner label. All values below are invented.

## Day-by-day look-fors

| Day | Expected evidence | If missing, next move |
| --- | --- | --- |
| 11 | A and B are same-team sources; four-layer 20 recorded vs 10 model; poster purpose and quiet qualification both noticed. | Read headline and small print with equal emphasis; contrast maker, number and method. |
| 12 | `P(0..4)=160,80,40,20,10`; `P(5)=5` remains unobserved; `2^-2=1/4`. | Build eight equal marks, then return to reciprocal notation. |
| 13 | Signed residuals `0,0,+4,+6,+10`; no cause inferred. | Physically align each prediction tile and record tile; change cause statement to a test question. |
| 14 | Protocol controls paper, lamp, geometry, ambient light, sensor/measurement, runs, recording and risk. Actual new readings only when approved and performed. | Ask a second group to attempt one instruction; turn a vague “same setup” into positions/settings. |
| 15 | See separate worked Check A below. | Re-teach the one weak component using new numbers, then reassess later. |
| 16 | At n=4 D gives 19–21, middle 20; clustered in this fictional setup, not universal or independent. | Contrast repeat on one desk with another documented setup. |
| 17 | `r=.55` has smaller mean absolute gap than `.50` for B, but neither is a proved paper property. | Sort `fits five numbers` from `predicts another setup`; plan a new test. |
| 18 | “The reduction” is traced to team/action/values; “dependable” unsupported; revisions remain attributed. | Keep one useful abstract noun with its referent, then remove unwarranted certainty. |
| 19 | Two panels show decision, `10` model vs `20` recorded, source/method limit and next check. Actual Arts evidence requires a designed media artefact. | User-test first glance; revise the biggest unsupported word. |
| 20 | See separate worked Check B below. | Target exponent, repeat/independent distinction or operational controls separately. |

## Check A · first response only

1. `P(3)=96×2^-3=96×(1/8)=12`. Negative exponent expresses reciprocal, not a negative quantity. An equivalent repeated division `96→48→24→12` is full reasoning.
2. Three-card record is **18**; `record − prediction = 18−12 = +6` arbitrary units. This conflicts with the notice's exact-half assertion for all additional cards; it neither identifies a cause nor a real material property.
3. “The improvement” nominalises and presumes a desirable change. A defensible revision: “In the Folded Card team's invented one-table record, the three-card sensor value was 18 rather than the exact-half prediction of 12; the method and viewing conditions were not validated.” Variants naming team, comparison and limitation earn credit. “Proved reliable” does not.
4. Accept an operational request such as fixed sensor distance and background reading, sensor calibration, or independent repeat at three cards; “test more” alone needs precision.

**Response branch:** If 96×2^-3 becomes −12, use eight equal shares. If learner reports 18 as the prediction, draw separate source/model columns. If the rewrite only swaps “improvement” for “change”, ask who did what, in which invented record, and with what limit. Do not alter a first-response score because of later feedback.

## Check B · first response only

1. `P(3)=120×2^-3=120/8=15`. Three-sheet invented values are 23,21,22; ordered 21,22,23, so middle is **22**. Difference `22−15=+7` arbitrary units. Range 21–23 is also acceptable as supplementary description.
2. Three values cluster for one reported “same desk” setup; this is within-record similarity. Missing calibration, materials, ambient light and independent observers prevent claims about another setup, method validity, mechanism, safety or real reproducibility.
3. Good question: “With a specified paper batch and fixed lamp/sensor geometry, how do sensor readings change as sheet count increases from 0 to 3?” Controls must be operational: e.g., mark lamp and sensor positions at specified fixed separation, use the same named lamp setting/background condition, use equal-size paper from one batch, calibrate or zero the sensor consistently. Merely writing “keep light same” needs a setting. A written plan does **not** prove any actual retest took place.
4. Accept “Require a documented supervised retest before a decision about a trial: the team's fictional three-sheet middle 22 is 7 above the exact-half prediction of 15; calibration and background are missing; repeat with fixed geometry, background readings and an independent observer.” Pausing or approving only a *small supervised non-public retest* is equally defensible when reasoned. Public use or safety claims are unsupported.

**Response branch:** If median is mistaken for average, order the three values physically; both happen to be 22 here, so ask learner to name their method. If clustered values are called universal proof, ask what a different observer/setting could change. If a method plan is described as completed, mark `not observed` and separate future plan from completed action.

## Desk scoring note

Use four dimensions rather than a single permanent label: exponent/model, source/data separation, method/uncertainty, audience/wording. For each, note **independent**, **with access support**, or **needs next teaching step**. Score the first response; document help used. A different communication route carries the same evidence demand. No claim of actual classroom efficacy has been established by this desk review.