Understand a collaborator before choosing the next contribution

We supplied the same brief. Why are we preparing different results? Start from the contribution that must fit the whole and the message or action that puts it in doubt. Construct the few alternative explanations that would change what you do next, then obtain evidence only when it can improve that choice enough to repay the interruption. The useful result is a better-grounded explanation, division or next action. A sufficient returned contribution can end the question without a detailed model of its author.

This applies to uncertainty about a participant’s understanding or intended continuation. If a source has actually changed, supply the current source and reopen the affected work through SYSE.44. If the next operation is already determined by an inspectable program rule, inspect that rule. With a settled division and usable returns, continue the work.

Construct alternatives from the contribution that is at risk

Recover the required meaning and the observation before diagnosing the participant. Ask which difference in their interpretation, available information or planned action could produce that observation and make the receiving contribution fail. For each plausible account, say what you would explain, request or do differently if it held. Keep accounts together when their distinction changes no available action. These hypotheses concern the present continuation. A.3.3.PI explains how to recover a consequential distinction hidden by an otherwise identical description, retain alternatives and update them from an observation.

Consider this constructed engineering case. Two participants receive the same brief, source tables and rule: compare variants A and B using the limiting fractional margin m = capacity / demand - 1; a case passes this supplied numerical rule only when m >= 0. Preserve the source of the values and check the decisive inputs directly against that source, independently of the integrator’s copied calculation. They may prepare the comparison; they have no permission to change the design. Case 17 has capacity 80 and demand 100 in the same units. Its ratio is 0.8 and its margin is -0.2, so it fails the rule.

The collaborator writes: “Case 17: reserve 0.8; I can continue checking.” One failed continuation is to assume both correct understanding and the intended source check. The later return treats 0.8 as positive reserve and repeats arithmetic from the integrator’s copied values. The comparison then has a wrong interpretation and lacks the required source check. Resending the identical brief does not establish which assumption needs repair.

Return to the initial message with that common input unchanged. The expression “reserve 0.8” is consistent with confusing a capacity/demand ratio with surplus, but it could also be an imprecise label from someone who understands the margin. The planned “check” could revisit the original source or merely recalculate the copied table. The first uncertainty changes the explanation; the second changes the division of work.

Choose an observation by the work it can change

Ask what you would expect to observe under each retained account. A useful question, small work sample or permitted observation makes those predictions differ at a point relevant to the receiving result. An assurance such as “I understand” fits both calculation accounts too easily. Repeating a person’s own words may test recall while leaving their use of those words unknown. Prefer a return that exercises the disputed meaning or exposes the proposed operation.

Here the collaborator is available, showing one row is quick, and substantial calculation remains. Request: “For case 17, show the expression and pass/fail meaning you will return. Which source will your next check use?” The expression tests current use of the convention; the source answer reports the intended continuation. Their evidence is different: one performs a small calculation, while the other says what is planned. A.15.9 helps bound that requested contribution and the reliance placed on it.

Compare the question with explaining the convention directly, specifying the source contribution explicitly, completing the remaining calculation yourself, or proceeding on the current evidence. MMP.8.SD models the information available at each choice and the continuation an observation enables. A small qualitative branch comparison is enough here: a wrong formula leads to explanation and a new application; a correct formula avoids needless reteaching; a copied-table plan changes the requested source contribution. C.11.DUA asks whether that difference warrants the question’s whole burden. The stipulated quick return and substantial dependent work make this probe useful; they are not measured savings.

If only one short calculation remains, requesting its complete, checkable return may finish the work directly. If an explicit explanation and division are cheaper and adequate under all remaining accounts, use them. When the partner is unavailable or the distinction cannot be observed in time, choose a supported continuation under the unresolved alternatives or state the missing condition. The example’s two questions are not a form that every collaboration must complete.

Revise the account and perform the changed continuation

Suppose the row says 80 / 100 = 0.8, “positive reserve, pass.” That defeats the shortened-label account for this current return. Explain why the ratio must be reduced by one, giving 0.8 - 1 = -0.2, then obtain an application to another relevant row before relying on the remaining calculations. For a supplied capacity 120 and demand 100, 120 / 100 - 1 = 0.2 passes the same numerical rule. A correct second row supports this bounded application. The receiving engineering checks still govern the finished comparison.

Suppose instead the row gives -0.2, “fails,” but the partner says the next check will use the integrator’s copied values. Keep the correct calculation; change the division by requesting the decisive values from the original source, with enough provenance to locate them. If the actual source return confirms 80 and 100 for this case, use -0.2 and the failed numerical condition in the comparison. If it supplies different applicable values, recalculate the affected case. If the source is unavailable, retain the arithmetic with its input check unresolved and continue only as far as the receiving decision permits. An announced intention to check is not that check’s completed result.

An unclear reply may leave both interpretation accounts possible. Obtain a small permitted work product, or use an explanation and division adequate under both. Do not turn ambiguity into a diagnosis. An unexpected expression can defeat both initial accounts and require another decision-relevant distinction. The question itself may also prompt the partner to notice and repair an error: their correct answer then supports present use, not proof that they understood the convention before being asked.

Use the qualified observation through A.15.7 to choose the present action, and through SYSE.44 to prepare the contribution and check its join. Preserve a supported calculation while repairing its interpretation or source contribution. Stop when the needed result is usable or all remaining accounts support the same action. A later contradictory return can reopen the question. Inferred intent supplies neither permission to act nor the commitment on which another participant’s work may depend.

Adapt the observation to the participant and the claim

With a technical agent given the same brief, inspect a returned expression, cited source or observable proposed operation. A forecast that it will reuse copied values can be revised from those returns without access to private reasoning. If the actual interface or program determines the operation, its documented behavior or a permitted test may settle the question more directly. Explaining a convention to a person and changing an agent’s input or program are different interventions; observe whether the selected intervention produces the required contribution in that realization.

The published Adaptive Theory of Mind, §5.3, reports a related contrast in two Overcooked runs: earlier actions do not distinguish partner policies; a later conflict or wait changes the adaptive agent’s next move to yielding or proceeding. Its fully cooperative model and action-space conditions limit that evidence. The engineering case above illustrates the reasoning; it does not inherit experimental validation from that game.

For a different task, recover its own disputed meaning and what a useful return could show. When both parties change their continuations, check the current dependencies and shared effects again; a forecast about the partner does not ensure that the resulting joint actions will fit.