EXD.6 - Compare Explanations for the Recipient’s Use and Decide Whether a Repair Is Worthwhile
EXD.6:1 - Problem frame
Use this pattern when a proposed explanation change competes with an available explanation and the comparison can change what you retain, repair or stop. The issue may be a missing relation, costly reconstruction, inappropriate help or a change after an already sufficient answer.
Begin with the recipient’s required inference or action and the usable incumbent. Select the smallest consequential difference. The first useful result is a justified repair, a bounded provisional choice, retention of the incumbent or a specific return to missing subject work.
An explanation already sufficient for the use needs no additional comparison merely because another format or technique is available.
EXD.6:2 - Problem
A longer explanation may add information the recipient does not need. A shorter one may omit a decisive condition. A clearer-looking diagram may contain new content, and improvement after a hint may be credited to the original text.
Comparison must identify what changed and what outcome the evidence can support. Correctness, structural recovery, effort, retention, transfer and usefulness answer different questions.
EXD.6:3 - Forces
| Force | Practical tension |
|---|---|
| A possible improvement and its burden | Additional structure or support may help the use while increasing authoring, reading or maintenance cost. |
| Comparable conditions and realistic use | Isolating a change aids attribution; actual explanations may need coordinated content, expression and help changes. |
| Immediate evidence and later outcomes | One response can show a current inference without establishing later independent use. |
| Decisive checking and unnecessary measurement | Some differences need observation; others can be settled by a missing claim, wrong answer or sufficient incumbent. |
EXD.6:4 - Solution
EXD.6:4.1 - Fix the receiving question and its criterion
State what the explanation should let this recipient recover or do. Identify the subject criterion and an available explanation that can serve as the incumbent. A mathematical claim needs its correct derivation; a recommendation needs its actual grounds and conditions.
For a structural comparison, use FPF C.2.8: identify the account, its expression and observer, qualified by the selected relations, correctness criterion, preparation, available operations, access, assistance and budget. Choose a qualitative description of recovered and missing relations when it is sufficient. A numeric amount needs its own declared scale and units or denominator.
Distinguish recovering what an account asserts from establishing that assertion’s subject warrant. If warranted subject structure is the selected target, an unsupported relation cannot count as a correctly supplied ground.
EXD.6:4.2 - Identify the smallest material change
Use the construction that addresses the live difficulty: clarify the question, add a missing ground, build a discriminating example, coordinate an expression or guide a recipient product. Name which of content, expression, preparation, help or target changed.
When comparing forms, keep the selected claims and receiving conditions comparable. When coordinated changes are needed, compare the whole proposed repair but limit attribution accordingly. A claim inventory can help identify what both candidates express; it is not a measure of recovered structure.
FPF C.11.CRC supplies configuration-relative contribution comparison and C.11 the marginal-choice question. For Method descriptions, use ME.22’s actual comparison. For instructional material, use HCD.19’s representative-reader work and diagnosis. Their supplied results can answer the comparison.
EXD.6:4.3 - Obtain only the evidence that can change the choice
First inspect subject correctness, missing claims and explicit contradictions. These can justify a repair before a reader trial. Use a reading or application probe when the live alternatives concern recoverability or receiving use and its possible results can change the choice.
Preserve the initial response before assistance. Identify what the recipient had available at each point and what the subsequent help contributed. Prior knowledge can enable reconstruction; independent correction of an absent or false instruction must be identified rather than credited to the expression.
If the same person sees both explanations, prior exposure or practice can affect the second response. Design the comparison for the conclusion sought or retain that attribution limit. A small local observation may suffice for a local repair; a general effect claim needs a design and evidence appropriate to that stronger question.
If useful observation is unavailable, retain a sufficient incumbent or make a reversible design choice with its intended gain stated as an expectation. The absence of a measured human effect does not prevent a qualified explanation from being used.
EXD.6:4.4 - Choose the result and its stopping condition
Relate the demonstrated or reasonably expected contribution to the burden that matters in this use. Retain the incumbent when the candidate adds no worthwhile result. Prefer a targeted repair when it supplies a missing relation or fixes a consequential difficulty. Return unsupported subject claims to their supplier.
Keep result types distinct. More recovered structure within one budget differs from less effort to recover the same structure. Immediate recovery differs from retention after a delay or transfer to another task. Agreement and willingness to act differ from understanding a recommendation.
State what is retained, why, under which conditions and what fact would reopen the choice. A larger explanation is useful only to the extent that its contribution serves the selected purpose.
EXD.6:5 - Archetypal Grounding
Compare a complete prose account of the pooled mean with the table expression in EXD.3. Both carry five selected relations:
- count multiplied by group mean reconstructs group sum;
- group sums combine into total sum;
- group counts combine into the total number of individuals;
- total sum divided by total individual count gives the pooled mean;
- averaging group means equally answers a different weighting question.
The subject values and criteria are available: individual weighting gives 6.75; equal group weighting gives 5.5. The proposed table aligns count, mean and sum. Its intended contribution is correspondence under the recipient’s available preparation and budget, not an additional mathematical ground.
A representative response could reveal that the recipient cannot use the mean-to-sum relation. If the explainer then works that step and obtains a correct response, keep the original response and the supplied help separate. That sequence supports a targeted prerequisite or assistance repair. It does not show that the unassisted table outperformed the prose.
Alternatively, the recipient may already recover all five relations from the prose and use the required statistic. Without another worthwhile contribution, retain the prose. If the table is selected for a known reference task, such as locating counts and sums repeatedly, record that use and the design reason; do not report an unobserved reading-time improvement.
Now suppose the recipient changes the report to equal branch influence. Re-establish the target before comparing the explanations: the selected answer becomes 5.5. Failure to give 6.75 under that new question is not an extraction defect. The source account still supports both conditional answers.
These are constructed comparison branches. Their actual observations, if obtained, must be reported at their own conditions.
For advising, a person may recover the recommendation’s decisive premises and choose differently because of a preference. Keep the explanation if it served that understanding use; return the preference to the decision. Persuasion would require a different declared purpose and its own grounds.
EXD.6:6 - Bias-Annotation
Authors can prefer a new artifact because they invested effort in it, or prefer a fluent response because it resembles their own account. Use the sufficient incumbent, the subject criterion and the actual recipient conditions to constrain that judgement.
A human-learning result and an AI reconstruction probe have different evidential reach. Neither can silently stand in for the other.
EXD.6:7 - Conformance Checklist
Recognition concerns a choice whose outcome could change the retained explanation or next repair.
For assurance, inspect whether the incumbent and candidate have a stated use; their material differences are identifiable; correctness and recovery criteria are clear; preparation, help and initial responses are preserved at the needed grain; the evidence supports the claimed outcome; and the retained result is worthwhile under its stated burden and reopen condition.
A missing observation is not a zero score. A numerical ESI or effort value requires its declared measurement basis.
EXD.6:8 - Common Anti-Patterns and How to Avoid Them
| Failure | Repair |
|---|---|
| Crediting added grounds to layout alone | Separate content and expression, or bound the conclusion to the combined repair. |
| Crediting an assisted answer to an unassisted explanation | Keep the initial response and identify the assistance. |
| Measuring words, arrows or messages as recovered structure | Select subject relations and a correctness criterion under C.2.8. |
| Treating a revised question as a failed answer to the old one | Reset the target and preserve both conditional results. |
| Seeking an improvement after both explanations are sufficient without a live choice | Retain the usable incumbent and stop. |
| Reporting learning or persuasion from a fluent reconstruction | Name the actual outcome and obtain stronger evidence only for a claim that needs it. |
EXD.6:9 - Consequences
The author can choose a finite repair and stop work that adds no worthwhile contribution. The comparison can also reveal a missing subject input or an uncertainty about recipient conditions. Generalization beyond the observed or stipulated use remains a separate question.
EXD.6:10 - Architectural Rationale
Comparison connects constructive choices with receiving value. It uses existing characterization and choice methods while making the explanation-specific account, form, recipient and help distinctions explicit. Its result can send work to any construction pattern or preserve a sufficient answer; it is not a mandatory final stage of every exchange.
EXD.6:11 - SoTA-Echoing
FPF C.2.8 supplies the common structural characteristic; C.11.CRC/C.11 supply the contribution and marginal-choice questions. ME.22 and HCD.19 provide concrete comparison and repair practices at their subject scopes. Adopt those supplied operations and retain their distinct answer criteria.
Mengelkamp et al., 2025 distinguish comprehension and metacomprehension outcomes for informative narratives and exposition under reading-goal conditions. Their mixed results qualify any fixed genre or goal recipe. Fichtel et al., 2025 likewise distinguish more co-constructive behavior from a significant mean objective-understanding advantage.
The serious alternative is to select the most fluent, extensive or visually attractive explanation without a receiving comparison. In the worked case, the smallest account/form/help distinction can change the repair while leaving a sufficient incumbent available. Reopen when the task, subject basis, recipient, assistance, budget or relied-on evidence changes.
EXD.6:12 - Relations
EXD.1 establishes the receiving question. EXD.2, EXD.3 and EXD.5 construct candidate repairs. EXD.4 exposes changes during the exchange.
FPF C.2.8, C.11.CRC and C.11 retain their defining questions. The source and dependency account names the actual ME and HCD receiving uses.