| Ranking layers by a single score of “progress” | Performance, diversity, integration, resource burden and attention answer different questions; hidden weights can select the direction. | Keep the relevant differences and their conditions visible. Compare explained whole possibilities under the receiving choice’s values. |
| Reading publicity as technical movement | Repeated announcements and changed tests can create apparent acceleration without a comparable result. | Recover the underlying observations and compare the technical task separately from the attention history. |
| Treating an empty cell as a stable technology | Source delay, poor coverage and unavailable evidence can produce the same blank. | Return the specific missing observation or access limit, and investigate only when distinguishing it is worthwhile. |
| Declaring the whole contribution feasible from a component gain | A missing complement or unchanged bottleneck can defeat delivery. | Follow the gain through its dependencies, resources and adverse effects to the receiving result. |
| Discovering a category and calling it a new wave | Redrawing a map moves old facts without changing their dates. | Reclassify the earlier evidence explicitly and identify any separately supported technical event. |
| Continuing the scan after it can no longer change useful work | Collection consumes interpretation capacity and delays the present adequate answer. | Return the sufficient result or stop, retaining only possibilities whose later recovery has a practical use. |