ME.22 - Compare Method Descriptions by Content and Representation
ME.22:1 - Problem frame
Use this when. A Method description appears more useful after revision, but its content and its form both changed. You need to decide which revision to retain without crediting new instructions to a new layout, or treating a readable description as evidence that the described Method works.
The working reader is a Method engineer comparing descriptions, pattern languages or other inspectable representations for a named reader and action. Start by asking: what answer or action should improve, and what changed in the material available to that reader?
The first useful result is a bounded comparison: the earlier and proposed answers, the content and representation differences, what the evidence supports, and the next repair or selection. One local contrast can be enough. If the problem is simply choosing a representation for one action, use C.37; if several unlike Method uses need complementary representations, use ME.9. Use ME.14 when the question is the practical worth of the Method itself.
ME.22:2 - Problem
A new pattern language often adds examples, stops and source returns while changing headings and navigation. Readers then answer better, and the gain is attributed entirely to the language form. Conversely, a terse table can look worse because essential content was removed, not because tables are unsuitable.
The same confusion occurs with a checklist, diagram, executable description or AI-supported presentation. A comparison that changes subject, content, tools and task together can support a useful bundle choice, but it cannot isolate the contribution of one change.
ME.22:3 - Forces
| Force | Tension |
|---|---|
| Useful revision and explained gain | The team needs a decision now, while attributing a gain requires a comparison that can distinguish its sources. |
| Equivalent information and different affordances | A form can expose the same claims differently, while an added edge or omitted warning can change the content. |
| Fresh use and learning | A reader needs enough domain knowledge, while seeing another answer can teach the tested result. |
| Proportional evidence and consequential claims | A small desk comparison can settle a local correction; population, transfer and causal claims require more. |
| Better description and better Method | Understanding guidance is valuable, while enactment and domain performance remain separate observations. |
ME.22:4 - Solution
Compare the description change at the receiving use, separate the available contrasts, and report only the difference that the observations warrant.
ME.22:4.1 - Fix the comparison question
Name the reader or consuming system, the Method or candidate account, the intended action or decision, and the result and stop the description should let the reader recover. Keep the task facts, relevant source conditions, tools and opportunity to act comparable.
State the content that must survive: the relied-on action claims, relations, applicability, uncertainty, result, stop and source-return information. Then state the representation difference, such as notation, arrangement, salience, navigation or interaction. Do not call a new condition or an unsupported diagram edge a form-only change.
Use A.6.3.RT for a changed representation scheme, or the applicable same-regime wording, narrative or coarsening result. Its preservation and loss answer helps establish the comparison inputs; it does not demonstrate improved performance. If the Method’s reusable semantics also changed, retain the ME.15 identity/variant question and bound the description comparison accordingly.
ME.22:4.2 - Choose the smallest informative contrasts
When only content changed in a fixed form, compare those two versions. When only form changed and the relevant content is preserved, compare those two forms.
When both changed and the receiving decision needs their separate contributions, use:
| Variant | Content | Form | What its comparison can address |
|---|---|---|---|
| A | Prior | Prior | The earlier usable answer and its limits. |
| B | Added or revised | Prior | B against A: the content difference in the prior form. |
| C | Same revised content as B | Proposed | C against B: the form difference at the revised content. |
| D, conditional | Prior | Proposed | Complete the crossed comparison if content/form interaction or transfer of the form effect across contents matters. |
The three-variant arrangement gives two local contrasts. It does not identify an interaction. Four cells make that question inspectable but supply neither replication nor causal identification by themselves.
If a cell cannot be constructed without changing another load-bearing feature, say so. A bundle comparison may still support a local choice. Do not describe it as an isolated content or form result.
ME.22:4.3 - Observe the receiving use
Ask the reader to obtain the first useful result, not to rate which document looks better. Recover:
- the applicable action or decision and the answer actually given;
- the result, stop or missing condition the reader identifies;
- unsupported inferences or lost distinctions that change the answer;
- the source returns needed to support or correct it; and
- the observed burden when time, source openings, attempts or other relevant effort were actually recorded.
When the question is how much selected structure the reader can extract, use C.2.8 U.ExtractableStructuralInformation. Identify the expressed account, its publication form and the reader, with selected relations, correctness criterion, prior knowledge, operations, help, access and budget qualified for this comparison. A qualitative account of recovered and missing relations often suffices. A number needs a declared structural scale and fixed units or denominator, not counts of words or arrows.
Keep a design estimate separate from an actual recovery observation or a mapped formal estimate. A response establishes recovery in that trial, not a maximum or population reliability. State what changed in content, what changed in expression, and what preparation, actual help or a source return supplied. Familiar relations can count; a reader’s independent correction of false or absent instruction is not credited to the expression. Missing basis is not zero. A successful changed-condition return supports its own condition rather than increasing the original amount.
These are candidate observations, not one universal score. Select those that can change this comparison. A source-opening count does not measure comprehension; shorter reading does not prove lower total effort; recognizing a prescribed action does not establish its correct enactment.
Choose readers and allocation to variants in proportion to the claim. A fresh probe withholds the other answers and the author’s preferred conclusion when they would teach the task. Repeated-reader comparisons retain learning and order effects instead of calling the reader fresh. In AI probes, keep the model, supplied context and available tools recoverable when they affect inference. Generalizing from a few AI answers to human performance is another unanswered question.
ME.22:4.4 - When testing a changed-condition return
Use this probe when the comparison needs to establish how a reader responds to a new fact. First retain the reader’s initial answer. Then disclose the changed source, missing fact or altered receiving use. Ask what must change and what remains usable.
A good return revisits the dependent claim rather than repeating the whole task or preserving an invalid assumption. Compare the actual revised answer, its source returns and its stop. If the changed condition changes the comparison question or available action set, identify that new question; do not present the two answers as observations under identical conditions.
ME.22:4.5 - Return a bounded revision result
State the inspected variants, the content/form contrasts they actually support, observations, losses, uncertainty and next decision. Distinguish a demonstrated local difference, no distinguishing result in this probe, an unavailable contrast and a confounded comparison.
Compare extraction effort and receiving value separately from structural amount through C.11.CRC and C.11. Retain the sufficient incumbent when the proposed change supplies no worthwhile gain.
Return a content correction to ME.8, a contradicted maintained claim to ME.12, and selection evidence to C.37 or ME.9. Use ME.11 for representative enactment evidence and ME.14 for a Method-worth decision. A causal claim needs its own justified research design and inference basis; neither this table nor an attractive explanation supplies it.
An ordinary comparison can remain short prose. When another decision will rely on it, retain the candidate versions, receiving use, observations and inference limits as an identifiable account, with the source conditions that reopen it.
ME.22:5 - Archetypal Grounding
ME.22:5.1 - Added stop information versus a table
This is a constructed comparison, with explicit text rather than an assumed reader study.
The receiving action is deciding whether an earlier review may support a current description revision.
A, earlier prose: “Reuse the previous review for a display-only change.”
B, revised prose: “For a display-only change, reuse the previous review only while the reviewed claims, question and qualification window still match. If a required source argument cannot be recovered for a new recheck, stop that recheck and return to its source; preserve unrelated earlier results.”
C, the same revised content in a table:
| Condition | Action |
|---|---|
| The change is display-only, and reviewed claims, question and qualification window still match | Reuse the previous review only under this full condition. |
| A required source argument cannot be recovered for a new recheck | Stop that recheck and return to its source. |
| An earlier result is unrelated to the missing argument | Preserve that result. |
Initially, the facts say that the change is display-only, the claims and question are unchanged, and the qualification window is current. All variants can support reuse. The later condition says that a new recheck needs an argument outside the available excerpts and that the retained dossier contains only its locator.
B and C explicitly support the local stop and preservation answer; A does not contain that distinction. This establishes a content-coverage difference in the constructed texts. It does not show that an actual reader preferred or used C more effectively than B. A reader comparison would still be needed for that form claim.
The useful authoring decision is to retain the missing stop information and withhold the table-superiority claim.
ME.22:5.2 - A graph that adds a claim
A proposed graph draws an unconditional edge from “review complete” to “publish”. The prose requires a separate publication decision. The graph has added a consequential claim; it is not another form of the same content. Repair the edge and its condition before testing a pure form contrast, or compare the changed bundle with that difference explicit.
ME.22:6 - Bias-Annotation
Authors tend to know the omitted condition and may answer correctly from memory. Readers who see all variants can learn it too. Separate text inspection, fresh reading, repeated reading and actual enactment evidence.
The constructed examples expose content and inference boundaries. They contain no measured time saving, error rate or Method-effectiveness result.
ME.22:7 - Conformance Checklist
- The receiving reader, subject, action, result and stop are fixed sufficiently for the comparison.
- Content changes and representation changes are stated, including added or lost relations.
- Every claimed contrast is supported by the available variants and conditions.
- Reader learning, tool differences and other material confounds remain visible.
- When a changed-condition return is tested, an initial answer precedes disclosure of that condition.
- Observed burden is distinguished from estimated or unmeasured burden.
- The result returns to the proper content, selection, trial or worth decision.
ME.22:8 - Common Anti-Patterns and How to Avoid Them
| Misstep | Repair |
|---|---|
| Better answers prove that the new form is better. | Check whether the new form also supplied new content or tools. |
| A four-cell table is treated as a causal study. | Establish sampling, allocation, repetition and inference separately when that claim is needed. |
| A reader sees the changed condition before giving the initial answer. | Run the initial and changed-condition probes separately. |
| A diagram is declared equivalent because its labels match. | Inspect edges, polarity, conditions, uncertainty and source return. |
| No difference in one probe is called universal equivalence. | Retain the local no-distinguishing-result conclusion and its sensitivity limits. |
ME.22:9 - Consequences
A useful revision can be retained without overstating why it helped. Content defects and representation defects receive different repairs, and later selection or worth decisions receive evidence with usable limits.
The cost is preparing informative contrasts and observing the actual task. When that cost cannot change the decision, take the sufficient direct result or an honestly bounded bundle comparison.
ME.22:10 - Architectural Rationale
ME.8 asks whether a description exposes the claims a use needs. ME.9 organizes complementary representations across uses. ME.12 checks agreement and repair ownership. None of those results alone answers which difference between revised descriptions explains an observed improvement.
This pattern joins established representation-preservation and validation Methods around that professional revision question. It does not replace experimental design, produce a universal representation ranking or merge description usability with Method performance.
ME.22:11 - SoTA-Echoing
ACAP analysis (Doellken, Nelius and Matthiesen, 2024) distinguishes attention, comprehension, application and performance when investigating design-method difficulties. Adopt that separation; the evaluated sheet-metal-design setting and interpretive analysis do not justify universal metrics or causal attribution to each content element.
NIST’s factorial-design account supplies the established crossed-factor idea. Adapt it to the contrasts the Method-description decision needs; do not claim a new experimental theory. C.37, A.6.3.RT and ME.9 retain their representation results.
The serious cheaper alternative is direct claim inspection plus the existing representation selection. In the review-reuse example it already identifies the missing stop information; no reader experiment is necessary for that content defect. Use the fuller comparison only when a material form, learning, burden or interaction question remains. Reopen the result on a changed task, source condition, candidate content, reader population or hidden confound.
ME.22:12 - Relations
- ME.8 supplies the description and receiving use; A.6.3.RT supplies representation-preservation and loss.
- C.2.8 supplies the optional structural-recovery characteristic; C.11.CRC and C.11 keep extraction cost and marginal usefulness distinct.
- C.37 and ME.9 consume bounded evidence for one-action selection and cross-use representation profiles.
- ME.12 receives a contradicted-claim correction; ME.15 distinguishes description revision from Method variation.
- ME.11 supplies actual trial evidence; ME.13 and ME.14 separately judge transfer/fit and practical worth.
- ME.23 can use the comparison when choosing a pattern-language representation; ME.24 supplies reconstruction failures that may motivate another comparison.