IS.3:5 - Archetypal Grounding
IS.3:5.1 - Compare complete first-week alternatives
This constructed extension of Harbor Design uses invented quantities to show the comparison. Six analysts are available. After the local diagnosis, suppose four have adequate current capability for the selected case family and two need a supported attempt with feedback. The intended first-week contribution is eight eligible package checks, with the two additional analysts beginning to acquire the missing comparison operation. In the current continuation, the four capable analysts can provide those eight checks at 0.20 hours each, using 1.6 hours already available for checking. The two analysts’ acquisition requires additional work.
All three serious alternatives require the same access repair, taking 1.5 technical hours, and a 0.5-hour managerial contribution that makes the review interval and receiving decision effective. These two hours are counted once in each alternative.
| First-week alternative | Scarce qualified provider time | Other direct time, including ordinary checking | Complete effort for the selected work |
|---|---|---|---|
| Full six-person workshop, followed by brief individual follow-up | 2 hours preparation + 2 hours teaching + 6 × 0.25 hours follow-up = 5.5 hours | 6 × 2 hours attendance + 6 × 0.25 hours follow-up participation + 2 common hours + 1.6 hours checking = 17.1 hours | 22.6 hours |
| Targeted workplace rehearsal for the two analysts | 1 hour preparation + 2 × 0.75 hours coached rehearsal = 2.5 hours | 2 × 0.75 hours participation + 2 common hours + 1.6 hours checking = 5.1 hours | 7.6 hours |
| Qualified expert supplies the checking contribution for eight cases | 8 × 0.20 hours = 1.6 hours | 2 common hours; the expert replaces the analysts’ checking contribution | 3.6 hours |
These totals include the eight checks and the implementation work in the constructed alternatives. Other unchanged package work remains outside this comparison. Relative to the 1.6-hour checking continuation, the additional effort is 21, 6 and 2 hours respectively. Shared workshop time is counted separately for the instructor and each participant because they are different people occupying that interval; replaced checking work is not counted twice. No effect size or learning success is assumed from the arithmetic.
Suppose only three qualified support hours are available after existing commitments. The workshop is unavailable in that window. Targeted rehearsal fits the time limit; that inequality alone does not prove a workable schedule, but the two agreed 45-minute intervals and the preparation time provide one.
Expert execution supplies the immediate result at lower first-week effort. It does not supply the intended analysts’ acquired comparison operation. If the recipient requires only the eight immediate checks, that service can be selected. If developing the two analysts is part of the authorized promise, the service is not an equivalent cheaper answer.
The parties select targeted rehearsal for the bounded acquisition attempt and retain the existing qualified review condition until later performance supports another allocation. The actual rehearsal Method and assessment come from HCD and the subject provider. A.3.1.MI examines what happened; the table itself proves neither learning nor the target Method’s effectiveness.
IS.3:5.2 - Build the local instruction rather than paste a strategy name
The target requires identifying whether a source revision changes the support for a claim. The initial implementation instruction says:
Use the new review method before release.
A walkthrough shows that the analyst cannot find the source pair and that the technical lead receives only an unqualified approval flag. The local instruction is changed:
When a package cites a revised source, open the cited edition and the proposed current edition. Locate the passage supporting the claim. State whether the change affects that support, with the relevant passage. Send an unresolved or changed-support case to the technical lead before release recommendation. If either edition is unavailable, return the missing source instead of marking the claim checked.
The source owner supplies the two-edition route. The technical lead agrees the receiving response and timing. The learning provider uses representative pairs to teach the missing comparison where needed. The instruction is a local explanation of the selected operation, not evidence that anyone has performed it.
Now change the condition: a source is a continuously updated database with no recoverable historical edition. The instruction’s input condition fails. The helper returns the record-preservation and claim-support question to the source/subject owner. Inventing a timestamp from memory would not preserve the target operation.
IS.3:5.3 - Strong AI changes the obtaining way
Suppose a qualified AI service can perform the source comparison under the actual access and protection conditions, and its current evidence supports the required case family. One alternative lets it supply the comparison while the technical lead retains the receiving decision. Another uses it to prepare contrasts for the analysts’ practice.
The first can reduce demand for human comparison work without developing analyst mastery. The second can support learning only if the analysts perform the intended operation and obtain suitable feedback. The strategy choice follows the wanted result; it does not preserve a human coach merely because the earlier arrangement had one.
If the service loses access to a protected source, the alternative becomes unavailable for that case. A general capability claim about the AI does not restore the missing input.