# Companion essay adaptation record

September 25, 2026, second testing drafts. Both essays adapt *The Irreducible Officer* directly and keep its thesis: the goal is learners who direct AI toward a purpose they own, rely on it selectively, and defend the result. Each edition's headings and depth follow its own argument, so the section counts differ. The [original](../the-irreducible-officer.md) and [canonical claim map](../claims.md) remain unchanged. The [revised spine](../tasks/companion-essay-spine.md) records the decisions behind this pass.

Read [Judgment in Higher Education](he.md) or [Learning to Exercise Judgment](k12.md).

## What changed from the first drafts

The first drafts made foundations the destination and left critique of AI output as the only student AI activity. Their examples, a campus survey and two surface-temperature readings, each had one right answer reachable with a textbook rule. This pass:

- restores the original's contrast, with the AI-directing student as the model and the capable unaided student as the harder case;
- restores the five-step pilot, including the step where students direct AI against their own frame;
- adds the directing-AI progression from the original's Section V, set to each level;
- replaces both examples with cases that pass the five tests in [frame-first assignment design](../artifacts/frame-first-assignment-design.md);
- keeps only the failure modes each example shows;
- removes repeated caveats and product or facilitator text, which belongs in the audience guides and facilitator.

## Examples

| Edition | Case | Frame flaw in the AI answer | Changed case |
| --- | --- | --- | --- |
| HE | Research memo for a firm deciding on return to office | Accurate study summaries; "productivity" silently means short-run output averaged across workers | An experienced call-center workforce, where different evidence decides |
| High school | "Was the New Deal a success?" in US History | Accurate figures and programs; success silently means recovery and durability, never "for whom" | Whether the school's phone policy is a success |

## Claim check against claims.md

| Canonical claim | HE | High school |
| --- | --- | --- |
| 1. AI changes the performance standard | I: the bar went up; unaided student as harder case | I: same contrast in a history class |
| 2. Finished artifacts carry less evidentiary weight | III, VIII | VII |
| 3. Purpose is expressed through frame | IV: student decides what productivity means | III: teacher sets lesson purpose, student sets the argument's standard |
| 4. Appropriate reliance is teachable | V: levels of directing AI; accept study summaries, refuse the verdict | IV: perspectives checked against the packet |
| 5. Developmental friction builds judgment | VI: foundations inside the loop | V: sourcing taught inside the loop; Bastani |
| 6. Accountability is structural | VII: whoever sets the standard answers, at every role | VI: same, scaled to students |
| 7. Assessment makes ownership visible | VIII: frame, reliance, changed case; grade the defense | VII: defended-argument rubric and phone-policy case |
| 8. Faculty judgment and AI fluency accumulate | IX–X: shared misframed-answer library, visible assumptions | VIII–IX: teacher team library; younger grades need their own design |

## Evidence

The Social Security exclusion account follows Dubin (2024), which reviews both sides: Lieberman (1998), Quadagno (1988, 1994), and Katznelson (2005) on racial politics; Davies and Derthick (1997) and DeWitt (2010) on administrative feasibility. DeWitt's article itself returned an access error and was read through Dubin's quotations and the search abstract. The HE history and engineering translation cases rely on standard secondary works named in the HE references and have not yet had discipline review. New case facts were checked on September 25, 2026 against Bloom et al. (2015, *QJE*), Bloom, Han and Liang (2024, *Nature*), Emanuel, Harrington and Pallais (2023, NBER), DeWitt (2010, *Social Security Bulletin*), and the BLS/Lebergott unemployment series. Bastani et al. (2025) and EEF are scoped as before. The original's aviation and command sources stay in the PME essay.

## Voice review

Mechanical: pass (script), `writing-as-jack/scripts/jack_eval.py` on both scratch drafts; repo files match the scanned content.

Second pass fixed the first-pass judgment fails: unsupported claims (removed or sourced), interpretive morals after examples (replaced with concrete statements), and mirrored closings (the high-school close now differs). Judgment checks stay advisory; Jack remains the judge of voice.

## Still needed

A high school social studies teacher should review the New Deal case, especially the Social Security passage, before it reaches students. A history instructor and an engineering instructor should review the HE translation cases. An educator in each setting should also review the case, the documents or reading list it assumes, and the rubric. The audience guides, interactive context, site build, and workbench profiles still carry the old examples and need updating once these drafts are accepted.