HCD.4:5.1 - Constructed Positive Case: ImpactReview-17
ImpactReview-17 is a constructed pedagogical case. ImpactReview-17-Domain-Oracle-v1 supplies engineering correctness and criticality criteria for PumpWorks-Config-Baseline-42 and AI-Assist-M4. The observations and decision conditions below are stipulated case data.
The human holder is Engineer-P14. The later-Work demand is independent review of AI-proposed change-impact claims under a forty-five-minute case limit, current source/configuration access, auditable evidence, and bounded safety and maintenance specialist returns. The work permits a routine reminder to construct and test an alternative hypothesis, while the engineer must supply the case-specific rival and its test. The explicit probe prompt in HCD.3:5.1 names that operation without supplying its answer. Self-initiated recall remains a separate question. HCD.4 keeps four claim rows:
| Claim in the constructed profile | Current evidence | Target and trade-off disposition |
|---|---|---|
| source/configuration trace | 8/8 under the stipulated configuration | Preserve; do not trade traceability for speed. |
| independent AI-output challenge | 2/4 planted-error cases; both misses are unfamiliar variants | Priority gap; target 4/4 and zero accepted seeded safety-critical omissions, while the forty-five-minute limit remains a visible cost. |
| bounded specialist return | 2/2 relevant cases | Preserve the authority boundary; do not internalize specialist judgement. |
| auditable evidence return | 8/8 | Preserve; no aggregate may hide a missing critical-error record. |
The resulting comparison says: preserve the three supported claims, investigate independent AI-output challenge as the limiting gap, keep the prompt-design rival live, and do not select a programme. The current profile is not one score; the target is not proof of mastery or transformation.
Engineer-P14 faces one person-level self-development-focus choice within a separately pre-authorized twelve-hour allocation. The constructed decision record captures the chooser’s inputs and application of C.11:
| Decision value | Constructed case value |
|---|---|
DecisionSubject | Engineer-P14. |
DecisionSubjectGranularity | Person-level; Engineer-P14 participates directly in the comparison. |
OptionSet | O1 preserve the profile without selecting a change; O2 make independent AI-output challenge the priority for later programme comparison while preserving the three supported claims; O3 defer a human profile-change priority and return the live prompt-design question to the Method/tool owner. |
EvaluativeMeasure | Prefer fewer accepted seeded safety-critical omissions while preserving source/configuration trace, specialist authority, auditable evidence, the forty-five-minute case limit, and reversibility. |
BeliefState | Two unfamiliar misses, a small sample, and a live prompt-design rival remain. |
OutcomeModel | A two-case prompt comparison can still change which of O2 or O3 survives, while differences between the cases remain unresolved. This supplies evidential input rather than an isolated causal prompt effect. O1 leaves the observed critical-error gap unchanged. |
| dependence layer | Evidential baseline only; no causal intervention or capability-change claim is used. |
ChoiceRule | Choose a profile-change priority now only if it survives the live nonhuman rivals and no feasible decision-changing probe is worth its cost; otherwise emit probe again for the smallest feasible probe that can reverse O2 versus O3 within the allocation and protected conditions. |
ProbeActionSet | Two further unfamiliar seeded-error cases, with one first attempt per case: one uses a neutral prompt, the other an explicit counter-hypothesis prompt. Observe whether Engineer-P14 generates and tests a counter-hypothesis, and whether they reject the wrong AI claim. Keep the same sources, tools, specialist routes, and forty-five-minute limit per case. |
ProbeBudget | Two hours remaining for this pre-choice probe inside the twelve-hour allocation. |
CostToProbe | Up to ninety minutes for the two attempts, plus thirty minutes reserved for preparation and comparison: two hours in total. |
| value of information | A tested counter-hypothesis and rejection of the wrong claim only with the permitted reminder would strengthen O3: the observed human contribution works under that assistance condition. It would not establish self-initiated recall. Misses under both prompts would strengthen O2. Interpret either result with the small sample and unresolved case differences. The seeded critical-error boundary makes this possible change in choice worth the bounded cost. |
| probe-worthiness | Feasible and worth its cost. |
ChoiceResult | probe again with this two-case comparison, one attempt per prompt condition. No profile option is selected yet. |
| next question and authority | Return the new observations to HCD.3 and the updated comparison to HCD.4. The accountable manager owns time allocation, Work assignment, and employment consequences; safety and release owners retain their authority. |
Applying the stated ChoiceRule returns probe again: the probe fits the two-hour budget, preserves the protected conditions, and can still change which of O2 or O3 survives. Engineer-P14 participates in the comparison. The pre-authorized time supplies resources for the bounded probe; safety, release, and employment decisions remain with their authorized owners.