| How much reasoning effort repays its cost? | Resource-rational analysis ties computation to action value under bounded resources (Lieder & Griffiths 2020, with Russell & Wefald 1991 as lineage). | The useful-result threshold and non-dominated next move govern setup and application work. The repeated-handoff and one-off cases show the practical difference. | Adapt. FPF retains multiple separately defined dimensions rather than one universal value-of-computation scalar. |
| Should ontology/formality be complete before use? | Demand-driven incremental formalization starts from useful work and deepens when use repays it (Shipman & McCall 1999 as a historical HCI/knowledge-base precursor). | The one-apparatus path allows value before total configuration, with reopen on recurrence, failure, automation, or stronger guarantee. | Adopt as continuing lineage. No HOS tool or single formality ladder is imported. |
| Do different questions warrant different apparatus and outputs? | Current competency-question work distinguishes purposes and expected products rather than imposing one ontology checklist (Keet & Khan 2024). | Eligibility is tied to one declared use/result/guarantee; the pattern does not universalize one method or checklist. | Adapt. Question differentiation changes candidate eligibility and return. |
| How should concern and purpose constrain a model or architecture description? | ISO/IEC/IEEE 42010:2022 keeps concerns, stakeholders, viewpoints, and architecture descriptions explicit without prescribing one tool. | Declared use, guarantee, and direct result keep apparatus application purpose-bound while subject p |