The Natural Logic Extraction Instrument (NLE) — Draft v0.1
Instrument: Natural Logic Extraction Instrument (NLE), draft v0.1. Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
Corpus in use: Core Stoicism, Nine Excerpts, Six Commitments, Propositional Conversion Instrument (PLCI) v1.1.
I. Purpose
The NLE exposes the logical skeleton of a Stoic philosophical text entirely in plain English — the fork it turns on, the branches it weighs, the single claim its conclusion depends on — without translating any of it into symbolic notation. The instrument exists because a passage can be logically rigorous while remaining invisible as an argument to a reader who does not read formal logic. The NLE makes the rigor visible without requiring the reader to first learn a notation.
The NLE is not a restatement of PLCI in different formatting. PLCI converts a passage’s logical structure into symbolic form — defined variables, propositions, an argument derivation. The NLE performs the same underlying analytical work — exposing what the passage’s validity actually depends on — but every step is rendered as a full natural-language sentence, checkable by a reader with no symbolic training. The two instruments may be run on the same passage to cross-check one another; neither supersedes the other.
II. Procedure
Step A — Terminus. State the conclusion the passage is actually driving toward, in one plain sentence, before analyzing anything else.
Step B — The Fork. If the passage’s argument proceeds by ruling out an alternative, state the fork explicitly: what is the prior question whose answer determines everything downstream?
Step C — Branch Tracing. For each branch of the fork, state in full sentences: the premise that defines the branch, what follows from it, and what it costs or delivers. No symbols, no compression. Each branch must be traced with equal thoroughness; an asymmetrically detailed branch is a Step C failure (see Named Failure Modes, Section IV).
Step D — The Fulcrum. Name the single claim on which the whole argument’s validity actually turns — the one classification that, if wrong, collapses the passage’s conclusion. The fulcrum must be load-bearing in fact, not merely thematically prominent.
Step E — Consistency Check. Confirm that the traced branch is the only one that permits every sentence the text actually contains, without contradiction. Test each sentence of the passage individually against the branch. If a branch cannot produce one of the text’s own sentences, it is the wrong branch, or the fork has been mis-stated at Step B.
Step F — Self-Audit. Mandatory before output is finalized. Confirm explicitly: no symbolic notation was used at any step; no premise was introduced beyond what the passage states or directly entails; no training-data philosophical content was imported to fill a gap; every distinct term in the passage was tracked as its own term throughout, with no two distinct referents collapsed under one label (see Failure Mode 1).
III. Operational Protocol
Execute Steps A through F in strict sequence. The self-audit at Step F is mandatory and must appear explicitly in output; it is not an internal check. If any item in Step F fails, the run is not complete — return to the step where the failure originated and rerun forward from that point. Do not patch the output at Step F without revising the step that produced the defect.
IV. Named Failure Modes
Failure Mode 1 — Variable Conflation. Two distinct terms in the passage are tracked under a single label because they are related or because the passage’s prose moves between them without a hard grammatical break. This is the most consequential failure mode: it can produce a false claim of inviolability or necessity for something that was never inviolable, by silently transferring a property that belongs to one term onto the other. Every noun phrase load-bearing enough to appear in the Fulcrum (Step D) must be checked individually for whether it is doing one job or two.
Failure Mode 2 — Fulcrum Misidentification. The instrument names a claim as the fulcrum because it is the passage’s most quotable or rhetorically prominent line, rather than because removing it actually collapses the conclusion. A true fulcrum passes a removal test: state the argument without it and confirm the conclusion no longer follows. A claim that merely restates or illustrates the fulcrum is not itself the fulcrum.
Failure Mode 3 — Symbolic Leakage. Formal notation, variable letters, or symbolic operators enter the output at any step. This defeats the instrument’s purpose, which is a rendering checkable without symbolic training. Any output containing a symbolic token is not an NLE run; it is a mislabeled PLCI run.
Failure Mode 4 — Premise Importation. A premise appears in the branch tracing that is not stated in the passage and does not follow from what is stated, typically supplied from general philosophical knowledge of Stoicism rather than from the specific text under analysis. Where the passage under-specifies a step needed to complete the argument, the instrument must name the gap rather than fill it silently.
Failure Mode 5 — Branch Asymmetry. One branch at Step C receives full sentence-level tracing while the alternative is dismissed in a clause. This produces the appearance of a rigorously ruled-out alternative when the alternative was never actually traced far enough to rule out. Both branches require the same depth of tracing regardless of which one the passage endorses.
Failure Mode 6 — Consistency Check Omission. Step E is asserted rather than performed — the instrument states that the branch is consistent with the text without testing individual sentences against it. A Consistency Check that does not name which specific sentence would fail under the rejected branch has not been performed.
V. Scope and Limits
The NLE exposes the logical structure a passage already has. It does not evaluate whether that structure is philosophically sound, does not issue a correspondence verdict, and does not determine action. Those are the work of the SCE, the SLE, and the SDF respectively. The NLE’s output is descriptive: this is what the passage’s argument actually depends on, stated so a reader can check it without notation. Where the passage's soundness is in question, the NLE's Consistency Check (Step E) may show that the text is self-contradictory across branches — but the instrument stops at reporting that finding; it does not adjudicate which branch the text ought to have taken.
Instrument: Natural Logic Extraction Instrument (NLE), draft v0.1. Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
No comments:
Post a Comment