SYSE.49:5 - Archetypal Grounding

Make a missed carry observable in human practice

The receiving engineering question is whether a proposed worksheet and sequence make carry recording and use recoverable. HCD.6 supplies a practice design for a person already able to perform the digit products: produce a correct stock total with intermediate working, first with an allowed demonstration, then correct the affected action. The arithmetic criteria are supplied. The engineer prepares the worksheet, task cards, independent answer basis and an available observer who can give the designed feedback.

Use 123 × 3 = 369 as a no-carry contrast and 127 × 3 = 381 as a carry case. The latter requires recording the 2 from 7 × 3 = 21 into the tens column and using it in 2 × 3 + 2 = 8. Capture the written relation and performed operation: an absent carry record, a recorded 2 ignored during calculation, and correct working copied into a wrong report are different observations.

If the person records 2 but writes tens 6, feedback preserves the correct units/hundreds, points to the unused carry and asks for correction of that operation. The corrected result is 381. This can be supported correction on the same case. A subsequent changed-value task is needed if the next question is whether the person recognizes and uses the carry without the answer-bearing cue. A partially worked demonstration must not be counted as that independent recognition.

Challenge the feedback itself: an answer key that approves 361 despite the visible unused carry fails the supplied arithmetic criterion. A sheet whose carry row is misplaced returns to SYSE.48/.52; a missed use despite a correct supplied layout returns to the procedure or the human learning diagnosis. The qualified material and observations can inform that repair. They do not establish that learning occurred; HCD.11–13 supplies the corresponding performance, transfer and retention questions.

Distinguish an absent effect from an absent reply

An agent sometimes repeats a configuration change after a lost reply. Reproducing the failure on the real service is costly. The interface owner supplies four predicates: the requested target reaches the requested version; one attempt has at most one state-changing effect; an acknowledgement without the effect is not completion; unresolved effect does not authorize blind replay.

The engineer constructs a fixture with target/version, attempt identity and effect count. Its operations can apply an effect, suppress the reply after application, acknowledge without applying, and return delayed state observations. Clean initialization and exercised reset come from SYSE.33.

Paired tasks vary target, initial version and reply/effect condition. Additional cases begin already complete or request an unsupported transition. A contract-derived check reads actual fixture state and effect count; it distinguishes completion, warranted restraint and unresolved effect.

To challenge shared error, deliberately make both a task generator and its generated verifier equate acknowledgement with success. An acknowledgement-only response leaves the target unchanged, so the independent state check defeats that premise. A second challenge repeats the effect: the final version looks right, but effect count exposes the forbidden duplicate.

Qualified traces return to SYSE.47 to construct the missing-effect/unknown-effect branch while the model stays fixed. A recovery wrapper under SYSE.48 or a trained policy under SYSE.45 would be a separate optional result. SYSE.46 keeps final comparison tasks outside those construction choices.

If the real service lacks the fixture’s attempt lookup, the useful-recovery claim remains unqualified. Retain valid normal-result exercises and return the mismatch to the interface/environment owner. This constructed example specifies the fixture and observations to obtain; it does not assert production reliability or an executed trial.