| Local specialist fitness and cross-specialist use | A local model should use the distinctions needed by its specialist Work; a receiving engineering decision still needs explicit correspondences and limits. |
| Minimum useful set and missing perspectives | Every additional description costs maintenance and collision checking; an omitted claim—for example, a functional, physical, temporal, affected-System, realization, or evidence claim—can make the decision unsound. |
| Stable subject identity and changing configurations | Several epistemes may concern one enduring System while referring to different particulars—for example, parts, variants, versions, effectivities, or intervals. Recover the identity relation instead of relying on a shared name. |
| Human readability and machine-supported checking | Text and diagrams can reveal meaning to people; formal expressions and automation can expose selected contradictions. Warranted reliance still requires an explicit interpretation, completeness boundary, and world-side correspondence. |
| Executable model and physical result | Simulation and generated code can shorten feedback; they can also reproduce shared assumptions and omit physical effects—for example, manufacturing, installation, environmental, or operating effects. |
| Shared model and federated descriptions | One common model can reduce translation for a bounded use; independently maintained descriptions can preserve needed expertise and authority. Choose from the receiving decision and the losses of each arrangement. |
| Current decision and future reuse | General data structures can enable later use; speculative completeness can delay the present decision and preserve obsolete distinctions. |
| Publication convenience and epistemic identity | One publication occurrence can make several epistemes available, and one episteme can appear through several publication forms. Carrier consolidation leaves claim content and editions distinct. |
| AI-assisted production and warranted reliance | AI Systems can produce candidate outputs—for example, text, code, diagrams, mappings, checks, or summaries—quickly. The outputs still need recoverable sources, subjects, error checks, and decision-specific acceptance. |
| Standards compatibility and present practice | Standards supply useful distinctions and exchange constraints. Treat conformance, official publication, and academic visibility as evidence of declared content; assess practical adoption, effectiveness, and current engineering Methods separately. |