fpf.tools
Library / First Principles Framework (FPF) - Core Conceptual Specification
Jump to passage
In this reading
Source · lines 95922–95932

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 14:36:52 UTC · snapshot created 2026-10-03 14:38:14 UTC · last check 2026-10-03 14:50:08 UTC

E.17.EFP:5.7 - Human-authored and generated task replay against the simpler alternative

This is a qualitative task replay for local architecture choice, not an empirical performance study. Each case compares EFP with the least-cost source-linked note on comprehension, semantic preservation, author/check time, and prevention of overread.

Task and credible simpler alternativeComprehensionSemantic preservationAuthor/check timeOverread preventionNon-dominated result
Human-authored shift note. An engineer writes two sentences that repeat inspection note N-14 without changing its claims. Simpler alternative: Reader orientation; source N-14; not an operating procedure.The simple sentence is as easy to understand as an EFP class note.The source locator and unchanged wording preserve the needed tether.The simple note is shorter to write and check.not an operating procedure blocks the only credible overread.The simpler note dominates. Do not apply EFP; use the source/publication pattern and stop.
Generated incident explanation. A generated paragraph restates one observed recovery and adds therefore the design is robust. Simpler alternative: attach a source link and label the paragraph AI summary.Both versions are readable.The simple label misses the widened robustness claim; EFP’s identity-and-bounded-sufficiency screen detects another target claim and prevents source identity from being inherited.EFP adds one focused claim comparison; no full metadata block is needed.EFP blocks reliance on the widened claim until its target episteme and source-to-target relation exist.EFP is non-dominated when the generated text will be reviewed, reused, disputed, or relied on. Keep the identity screen, class only after identity, bounded/blocked use, and reopen; add trace or evidence only for the named reliance.

The human-authored case is the ordinary non-use boundary. The generated case is the source-grounded branch supported by XAI/NLP/generated-explanation literature. A human-authored case may still use EFP when a real source-pinned/reconstructive/didactic/speculative ambiguity changes the next action, but authorship alone never triggers the profile.