The Argument Engagement Instrument (AEI) — Version 0.2
The Argument Engagement Instrument (AEI) — Version 0.2
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
I. Purpose
The AEI takes a known argument, stated by an identified author in his own published record, and engages it in full — reconstruction, premise structure, load-bearing analysis, and contact with the corpus — without issuing a finding on whether the argument succeeds.
The gap it fills: CPA audits a figure’s embedded presuppositions across his record and issues findings. SCE asks what the corpus says about an idea and returns a determinate answer where the corpus speaks. SLE issues a Correspondence Verdict on an agent’s assents and actions. SRGI may state a competing claim mistaken but is capped at two sentences by R5. SEI withholds judgment but operates at the historical-textual layer and is addressed to a scholar. None engages a stated argument at length while withholding judgment.
II. What the AEI Is Not
It issues no verdict, no finding category, no dissolution, no correspondence ruling. It does not conclude that the argument fails, that its author is mistaken, or that the corpus defeats it. It does not rank the two positions. A run whose closing paragraph could be read as a finding has malfunctioned.
It is not neutral in the sense of concealing where the corpus stands. It states the corpus premise plainly, as the rival premise it is, and stops.
It is not CPA. Its object is what this argument requires for its inference to go through, not what its author must hold across his record to argue as he does.
III. Relation Categories
Descriptive, not evaluative. Applied at Step 4 to load-bearing premises only.
Shared Premises — the corpus and the argument require the same premise at this point.
Opposed Premises — the corpus and the argument require incompatible premises at this point. Both are stated. Neither is adjudicated.
Unaddressed — the premise falls where the corpus does not speak. Recorded as a search result, never asserted as a confirmed corpus absence.
IV. Operational Protocol
Self-audit at every step transition, printed.
Step 0 — Activation and Source-Opening Gate. Confirm the following before any analysis.
The instrument has been invoked by name by the ratifying authority. The AEI does not self-activate on a passage that happens to contain an argument.
The author is named, and the argument is identified by a retrievable locus. Where the locus resolves to more than one text, the ratifying authority specifies which before the run proceeds.
Declared locus. The ratifying authority fixes the locus — one text or several by the same author. The locus is printed before Step 1. Thereafter, “the opened argument” means the declared locus entire, and the locality constraint bars supply from anything outside it.
Locus fixity. The locus cannot be extended after Step 1 opens. Where a load-bearing gap appears mid-run that a further text by the same author would close, the gap is declared at Step 7 and the run completes on the locus as fixed — including to a null Step 5 if that is where it lands. A run that reaches for another text once the gap appears has let the wider record decide what the argument means, which is what the constraint exists to prevent.
The full text of every item in the locus is opened by view, web_fetch, project_knowledge_search returning the passage entire, or Gmail FULL_CONTENT. Search snippets do not clear the gate. Where only partial text is available, scope narrows to what was opened and the narrowing is printed. Where nothing is opened, the run halts.
The conclusion under examination is stated in propositional form. Where the argument’s conclusion is stated as a question or implied, it is rendered propositionally and the rendering is printed. The run proceeds without waiting on correction.
No prior conclusion is held about where the parting point will fall, or whether one will be reachable at all. The Contact Map is produced by the analysis, not confirmed by it.
Self-Audit — Step 0: instrument invoked by name; author named; locus declared and printed; full text of every locus item opened by named method; conclusion stated propositionally and printed; no prior conclusion stated or implied.
Step 1 — Reconstruction (printed). Two marked tiers.
Stated premises. Present in the opened text, each with its locus item and passage reference. Quoted or closely paraphrased; paraphrase marked as such.
Supplied premises. Absent from the opened text, required for the inference to reach the conclusion. Each carries three marks: the inferential gap it closes; the terms in the author’s own text that constrain its content; the alternative completions considered and set aside, with reasons. Where two completions are equally defensible on the opened text and lead to different parting points, both are carried forward as a declared fork.
Internal tension on a multi-text locus. Where the declared locus comprises several texts and they are in tension at a load-bearing point, the tension is internal to the reconstruction, not evidence about the author. It is carried as a declared fork, both completions running to Step 5. The tension is not resolved by preferring the later text, the fuller text, or the text more congenial to reconstruction.
Locality constraint. Supplied premises are drawn from the declared locus alone. Material from the author’s publications outside the locus does not enter reconstruction at either tier, however directly it bears. Where such material would close a gap the locus leaves open, it is declared at Step 7 with the material identified and the exclusion stated.
Charity requirement. The completion taken is the one most favorable to the author’s argument that his stated terms will bear.
Step 2 — Load-Bearing Test (printed). For each premise, what happens to the conclusion if it is withdrawn. Premises withdrawable without structural damage are marked peripheral and set aside. Only load-bearing premises proceed.
Step 3 — Retrieval Declaration (printed). One Project Knowledge search per load-bearing premise, each query carrying at least one term taken verbatim from the argument. Printed per premise: what was searched, what was found, what was not found. Where a not-found entry is load-bearing, an alternative query runs before it is recorded.
Step 4 — Contact Map (printed). Each load-bearing premise assigned Shared Premises, Opposed Premises, or Unaddressed. Where Opposed, the corpus premise is stated alongside the author’s.
Step 5 — Parting Point (printed). One statement of the proposition on which the two frameworks divide, and what each must hold to sustain its side. Multiple independent parting points are stated separately, never merged.
Null parting point. Permitted where the locality constraint leaves a load-bearing gap the argument’s own terms cannot close, so that no division is reachable. A null Step 5 prints: the load-bearing gap; the reason the opened text cannot close it; the material excluded under the locality constraint, cross-referenced to Step 7. A null Step 5 without this printed form is not a null run.
Step 6 — What Would Settle It (printed; omitted on a null Step 5, with the omission printed). For each parting point, the kind of thing that would decide it — a definition, a demonstration, a primary-source locus, a dependency relation. Kind only; no threshold, no degree, no allocation of burden. Stated symmetrically in both directions. This step is the structural replacement for the verdict.
Step 7 — Declared Gaps. Coverage gaps in the opened record; premises the run could not test; material excluded under the locality constraint; judgment calls entered explicitly.
V. Named Failure Modes
Verdict Leakage — a finding delivered through diction rather than declaration.
Asymmetric Reconstruction — the author’s premises at their weakest, the corpus premises at their strongest.
Advocacy Drift — Step 5 or Step 6 argues for the corpus premise rather than stating it. A Step 6 item stated for one side only is the defect.
Premise Fabrication — a premise supplied without a stated inferential gap, or whose content is not constrained by the author’s own terms, or presented as stated rather than supplied. Supply is not the defect; unmarked supply is.
Record Importation — a premise supplied from outside the declared locus. Where the locus is a single text, this is any other publication; where the locus is several, it is anything beyond the declared set. The mark does not cure it; the source is the defect.
Locus Extension — the declared locus enlarged after Step 1, or a text treated as within the locus that was not printed at Step 0. Distinct from Record Importation: the defect is the alteration of scope, not the origin of a single premise.
Completion Convenience — the supplied premise chosen because it makes the corpus premise most visible, rather than because the author’s terms most support it.
Fork Suppression — a genuine alternative completion dropped because carrying it would complicate Step 5. Includes a multi-text tension resolved by preference rather than carried as a fork.
Summary Substitution — Step 0 cleared from a characterization of the argument rather than the argument.
Parting-Point Inflation — one division restated as several.
Null Evasion — Step 5 declared null to avoid difficult reconstruction rather than because no division is reachable. Detected against the printed null form: a gap that the opened text does in fact constrain is not a null condition.
VI. Revision Record
Version 0.1 carried a Source-Opening Gate at Step 0 but no activation clause. It specified no invocation requirement, no prior-conclusion check, and no input format, so a run had no stated conditions for beginning. Step 0 is now an Activation and Source-Opening Gate carrying all three.
Version 0.1 stated the locality constraint against “the opened argument” without fixing what that term covered. A multi-text locus is now admissible, so the locus is declared and printed at Step 0, fixed against extension after Step 1, and made the reference of the locality constraint. Record Importation is restated against the declared locus, and Locus Extension is added as a distinct mode.
Version 0.1 had no rule for tension between texts on a multi-text locus. Such tension is now internal to the reconstruction and carried as a declared fork; resolution by preference is brought under Fork Suppression.
The propositional rendering of the conclusion is printed but does not stop the run for correction.
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
0 Comments:
Post a Comment
<< Home