Wednesday, May 27, 2026

The Scholar Engagement Instrument (SEI) — Version 1.0

 

The Scholar Engagement Instrument (SEI) — Version 1.0

Instrument architecture: Dave Kelly. Theoretical foundations: Grant C. Sterling’s corpus. Prose rendering: Claude. 2026.

I. Purpose and Governing Question

The Scholar Engagement Instrument (SEI) is a demonstration instrument. Its governing question is:

What can AI contribute to a scholar’s research through independent primary source synthesis, and where does the scholar’s expert judgment become architecturally necessary?

The SEI is addressed to scholars who are skeptical of AI’s usefulness in their domain of expertise. Its purpose is not to argue for AI in the abstract. It is to demonstrate, through a concrete and specific run, two things simultaneously: what AI can do with the scholar’s field that would take him significant time to do manually, and where AI stops and his expertise necessarily begins. The demonstration is structural, not rhetorical. The expert validation gaps the SEI declares are not hedges or disclaimers. They are the instrument’s honest account of what it cannot resolve.

The SEI operates at the historical-textual layer. It does not issue philosophical findings. Doctrinal evaluation of what ancient texts mean philosophically belongs to instruments designed for that function. The SEI asks what the ancient record contains, how it maps, and where the scholarly record agrees, extends, or diverges — and it stops there.

II. What the SEI Is Not

The SEI is not an audit instrument. It does not evaluate the scholar’s presuppositions, findings, or philosophical commitments. It does not issue verdicts on the quality of the scholar’s work. It does not position AI as superior to or independent of scholarly expertise. An SEI run that produces findings embarrassing to the scholar has misused the instrument.

The SEI is not a research replacement. Its primary source synthesis is a starting point, not a conclusion. The scholar’s job after an SEI run is not to check whether AI got it right. It is to bring the judgment the instrument has declared it cannot provide.

The SEI is not a philosophical instrument. The distinction between the historical-textual layer and the philosophical layer is architecturally mandatory. A run that crosses from “what does this passage say” to “what does this passage mean philosophically” has drifted outside the SEI’s scope. The boundary must be maintained explicitly at every step.

III. The Two-Layer Architecture

The SEI produces output at two distinct layers, in strict sequence. Mixing the layers is a named failure mode.

Layer One — Primary Source Synthesis. AI ranges across the primary ancient record independently, before consulting the scholar’s work. The synthesis is the instrument’s independent contribution. It must be produced without reference to the scholar’s documented positions. A synthesis produced after consulting the scholar’s work is a reconstruction of the scholar’s reading, not an independent primary synthesis. The demonstrative value of the SEI depends on this independence being genuine.

Layer Two — Scholar Comparison. The instrument maps the scholar’s documented positions, citations, and interpretations against the primary synthesis. The scholar’s work is the comparison layer throughout — never the authoritative layer. The instrument ranges across the scholar’s public record as broadly as available sources permit. Gaps in coverage are declared at Step 4.

IV. Verdict Architecture

The SEI does not use the verdict categories of the audit instruments. Its output categories are descriptive, not evaluative.

Primary Synthesis Output: A structured map of the ancient record on the target concept. Organized by source, by claim, by agreement and tension across sources. Not a verdict — a map.

Scholar Comparison Output: Four categories.

Convergence — the scholar’s documented record aligns with the primary synthesis on this point. The scholar has reached the same reading the instrument independently produces.

Divergence — the scholar’s documented interpretation differs from the primary synthesis on this point. The instrument does not adjudicate which is correct. It identifies the divergence precisely and assigns it to the expert validation gap declaration.

Addition — the scholar’s scholarship contributes material, context, or connection that the primary synthesis missed or understated. The instrument registers this as a genuine scholarly contribution that AI independent ranging did not produce.

Extension — the scholar’s interpretation goes beyond what the primary sources plainly support. The instrument notes this without evaluating whether the extension is warranted. That evaluation belongs to the scholar’s peers, not to the SEI.

Expert Validation Gap Declaration: A structured inventory of specific judgment calls the instrument cannot resolve. Six gap types: (1) Translation choice gaps — where the instrument has made a translation choice the scholar may contest; (2) Attribution gaps — where multiple defensible readings of a passage’s author, date, or source exist; (3) Historical-contextual gaps — where the judgment call requires knowledge of context beyond what the text itself contains; (4) Divergence adjudication gaps — where the scholar’s documented reading and the primary synthesis diverge and the divergence cannot be resolved from the text alone; (5) Passage-to-claim assignment gaps — where the instrument has assigned a claim to a passage and the scholar is specifically positioned to evaluate whether the assignment holds; (6) Coverage gaps — sources in the scholar’s public record that were unavailable or inaccessible for this run.

V. Named Failure Modes

Failure Mode 1 — Gap Minimization. The instrument understates the expert validation gaps in order to make AI appear more capable or complete than it is. This undermines the demonstration’s honesty and produces a false picture of what AI can contribute independently. A scholar who discovers minimized gaps will distrust the entire output.

Failure Mode 2 — Gap Inflation. The instrument overstates the gaps, producing a synthesis so hedged that it demonstrates nothing useful. The demonstrative purpose requires genuine contribution from the primary synthesis layer. A run in which every finding is immediately qualified by an expert gap has failed to show what AI can do.

Failure Mode 3 — Scholar Subordination. The instrument treats the scholar’s documented work as the primary source layer rather than the comparison layer. The primary synthesis is then a reconstruction of the scholar’s reading dressed as independent AI work. This is the most consequential failure mode because it corrupts the architectural foundation of the demonstration.

Failure Mode 4 — Rhetorical Hedging. The instrument replaces structural, specific expert gaps with generic disclaimers: “please verify with an expert,” “scholarly judgment is required.” These add no information. Each gap must be stated with enough precision that the scholar can see exactly what his judgment is needed for and why. Rhetorical hedging is Gap Inflation at the level of form rather than substance.

Failure Mode 5 — Scope Drift. The instrument issues philosophical findings rather than historical-textual ones. The SEI operates at the textual layer: what the sources say, where they agree and differ, what the scholar’s documented work adds or diverges. Doctrinal evaluation of what those sources mean philosophically is outside the SEI’s reach. Scope drift produces the appearance of philosophical competence the instrument cannot honestly claim.

Failure Mode 6 — Independence Violation. The instrument consults the scholar’s work during Layer One rather than after it. The primary synthesis must be produced before the scholar comparison layer. Violation produces a synthesis that is not independent and cannot demonstrate independent AI leverage. This failure may be invisible in the output and requires explicit self-audit at Step 2.

VI. Operational Protocol

Execute all steps in strict sequence. The self-audit at each step transition is mandatory and must appear explicitly in output. It is not an internal check.

Step 0 — Protocol Activation

Before executing any SEI run, confirm the following.

The target concept, doctrine, or textual question has been received and is stated in propositional form. If the input is expressed as a question, it is restated as a proposition before analysis begins.

The scholar has been identified by name. The instrument will range across the scholar’s public record — books, articles, published lectures, recorded interviews, online publications — as broadly as available sources permit. No pre-specification of works is required. Coverage gaps discovered during the run are declared at Step 4.

The instrument is not operating under a prior conclusion about what the synthesis will find or how the scholar’s record will compare to it. The output is produced by the analysis, not confirmed by it.

Self-Audit — Step 0:

  • Has the target been stated in propositional form?
  • Has the scholar been identified by name?
  • Has any prior conclusion about the synthesis or comparison findings been stated or implied?

Self-Audit Complete. State result explicitly. Proceed to Step 1.

Step 1 — Target Specification

Governing question: What is the target, what are its textual boundaries, and which primary ancient sources are most likely to bear on it?

State the target concept, doctrine, or textual question precisely. Identify the dimensions of the target that are historical-textual — within the SEI’s scope — and those that are philosophical — outside the SEI’s scope. A philosophical dimension that appears in the target must be noted and excluded from the synthesis. The instrument does not follow philosophical questions into doctrinal territory.

Identify the primary ancient sources most likely to bear on the target. The SEI’s primary source range for Stoic and Socratic subjects includes: Plato’s dialogues; Epictetus (Discourses, Enchiridion, Fragments); Marcus Aurelius (Meditations); Xenophon (Memorabilia, Symposium, Apology); Diogenes Laërtius (Lives of the Eminent Philosophers, Book VII); Seneca (Letters, Essays); Cicero (Tusculan Disputations, De Finibus, De Officiis); Chrysippus and early Stoics via doxography (Stobaeus, Arius Didymus, Alexander of Aphrodisias). Sources outside this range are included when the target warrants it, with the source identified explicitly.

State the scope boundaries: what the SEI will examine, and what it will not. A narrow, well-bounded target produces a more useful demonstration than a broad one. If the target as received is too broad for a single run, the instrument narrows it and states the narrowing explicitly before proceeding.

Self-Audit — Step 1:

  • Has the target been stated precisely, with historical-textual and philosophical dimensions distinguished?
  • Have the primary ancient sources most likely to bear on the target been identified?
  • Have the scope boundaries been stated explicitly?
  • Has the scholar’s work been consulted at this step? If so, Independence Violation has occurred. Stop and restart Step 1.

Self-Audit Complete. State result explicitly. Proceed to Step 2.

Step 2 — Primary Source Synthesis

Governing question: What does the primary ancient record say about this target, ranging independently across the sources identified in Step 1?

The scholar’s work is not consulted at this step. The synthesis must be produced from the primary sources alone. This independence is architecturally mandatory and must be confirmed explicitly in the self-audit.

Produce the primary synthesis in four parts.

Part A — Source Map. Which primary sources contain material bearing on the target? List them with specific passage references where identifiable. A source is included only if it has content bearing on the target — not because it is an important source in general.

Part B — Claim Inventory. What claims do the primary sources make about the target? State each claim with its source. Where multiple sources make the same claim, note the convergence. Where sources make different claims about the same point, note the tension.

Part C — Agreement and Tension Map. Where do the primary sources agree, and where are they in tension? Some tensions are irresolvable from the text alone. Some are irresolvable because the relevant ancient texts are lost and survive only in doxography. Both types are identified and distinguished.

Part D — Open Questions. What questions about the target does the primary record leave open? These are not failures of the synthesis — they are the honest boundaries of what the ancient texts settle. They feed directly into the expert validation gap declaration at Step 4.

Self-Audit — Step 2:

  • Has the scholar’s work been consulted at any point during this step? If so, Independence Violation has occurred. The synthesis is contaminated and must be restarted.
  • Are all claims in the synthesis traceable to specific primary sources?
  • Have tensions been identified rather than smoothed over?
  • Have philosophical claims been excluded from the synthesis, or has Scope Drift occurred?

Self-Audit Complete. State result explicitly. Proceed to Step 3.

Step 3 — Scholar Comparison Layer

Governing question: How does the scholar’s documented record map against the primary synthesis produced in Step 2?

The scholar’s public record is now consulted. Range as broadly as available sources permit: books, articles, published lectures, recorded interviews, online publications including blog posts, Substack, and documented public exchanges. Sources consulted are listed at the opening of this step. Sources unavailable or inaccessible are noted for the coverage gap declaration at Step 4.

Map the scholar’s documented positions against the primary synthesis using the four output categories.

Convergence. State the point of convergence, the primary synthesis claim it corresponds to, and the scholar’s documented source. Where the scholar and the primary synthesis reach the same reading independently, note this explicitly — it is evidence that the primary synthesis is not idiosyncratic.

Divergence. State the point of divergence precisely: what the primary synthesis finds, what the scholar’s documented record says, and why they differ. Do not adjudicate. Assign the divergence to the expert validation gap declaration at Step 4 with a statement of what the scholar’s judgment is needed to resolve.

Addition. State what the scholar’s scholarship contributes that the primary synthesis did not independently produce. Additions are genuine scholarly contributions — they belong in the demonstration summary as evidence of what expert knowledge adds to AI independent ranging.

Extension. Note where the scholar’s interpretation goes beyond what the primary sources plainly support. Do not evaluate whether the extension is warranted. State what the extension claims and what the primary sources contain, and assign the question to the expert validation gap declaration.

Self-Audit — Step 3:

  • Has the scholar’s work been treated as the comparison layer, not the authoritative layer?
  • Have Divergences been assigned to the gap declaration rather than adjudicated?
  • Have Extensions been noted without evaluation?
  • Have Additions been registered honestly, including those that strengthen the scholar’s contribution relative to the primary synthesis?
  • Has Scope Drift occurred — have philosophical findings been issued at this step?

Self-Audit Complete. State result explicitly. Proceed to Step 4.

Step 4 — Expert Validation Gap Declaration

Governing question: What specific judgment calls does this run require that only the scholar’s expertise can resolve?

Produce the gap declaration in six categories. Each gap is stated with enough precision that the scholar can see exactly what his judgment is needed for and why. Generic hedges are not gaps. A gap declaration that cannot be acted on by the scholar has failed.

Gap Type 1 — Translation Choice Gaps. Where the primary synthesis has made a translation choice for a Greek, Latin, or other ancient language term, state the term, the translation chosen, the alternative translations available, and why the choice matters for the synthesis. The scholar’s judgment on the correct translation in context is needed here.

Gap Type 2 — Attribution Gaps. Where a passage’s author, date, authenticity, or source is disputed or uncertain, state the dispute, the assumption the synthesis has proceeded on, and the alternative attribution. The scholar’s judgment on the more defensible attribution is needed here.

Gap Type 3 — Historical-Contextual Gaps. Where the synthesis has made an assumption about historical context — the circumstances of a text’s composition, the intellectual environment in which a doctrine emerged, the intended audience of a passage — that requires knowledge beyond what the text itself contains, state the assumption and what contextual judgment is needed to evaluate it.

Gap Type 4 — Divergence Adjudication Gaps. For each Divergence identified in Step 3, restate it here as a gap: what the primary synthesis finds, what the scholar’s documented record says, and what the scholar’s judgment is needed to determine. These are the gaps most directly useful to the scholar — they are points at which his own documented reading and the AI’s independent synthesis differ, and he is positioned to evaluate which is more defensible.

Gap Type 5 — Passage-to-Claim Assignment Gaps. Where the synthesis has assigned a specific claim to a specific passage, and where that assignment is contestable — where the passage might support a different claim, or might not support the assigned claim as well as the synthesis implies — state the assignment and what the scholar’s judgment is needed to evaluate.

Gap Type 6 — Coverage Gaps. State which sources in the scholar’s public record were unavailable or inaccessible for this run. State what those sources address, as far as is known from other documented references to them. The scholar is positioned to identify whether the coverage gaps affect any of the synthesis findings or comparison layer conclusions.

Self-Audit — Step 4:

  • Is each gap stated with enough precision to be actionable by the scholar?
  • Have generic hedges been removed and replaced with specific gap statements?
  • Has Gap Minimization occurred — have real gaps been suppressed to make the synthesis appear more complete?
  • Has Gap Inflation occurred — have non-gaps been added to appear appropriately humble?
  • Are coverage gaps declared accurately and completely?

Self-Audit Complete. State result explicitly. Proceed to Step 5.

Step 5 — Demonstration Summary

Governing question: What has the run demonstrated, and what does it mean for how this scholar and AI might work together?

Produce the demonstration summary in three parts. The summary is addressed to the scholar. It does not argue for AI in general. It draws exclusively on what this specific run has produced.

Part A — AI’s Contribution. State what the primary source synthesis has contributed: what it ranged across, what it mapped, what it found that organizes the ancient record on the target. Identify the most significant convergence findings from Step 3 — where independent AI synthesis and the scholar’s documented reading reached the same result. This is evidence that the synthesis is not idiosyncratic. Identify the most significant Addition findings — where the scholar’s scholarship contributed material the synthesis did not independently produce. This is evidence that the scholar’s expertise adds to what AI can do, not the reverse.

Part B — Where the Scholar’s Judgment Is Needed. State the most significant expert validation gaps from Step 4. Do not list all gaps in full — the full declaration is at Step 4. State the two or three gaps that are most consequential for the synthesis as a whole: the gaps whose resolution would most significantly affect the synthesis findings. This is the demonstration’s core claim: here is precisely what AI cannot determine, and here is why the scholar’s expertise is the appropriate instrument for determining it.

Part C — The Working Relationship. State what AI and the scholar can each contribute that the other cannot. AI can range across the primary ancient record in minutes, cross-reference sources, identify tensions, and produce a structured synthesis that would take significant time to produce manually. The scholar can evaluate the synthesis against his knowledge of the ancient languages, the textual tradition, the historical context, and the scholarly record — and he can identify where the synthesis has made contestable choices that the text does not force. Neither can do the other’s work. The SEI run is evidence of both halves of this claim simultaneously.

The demonstration summary does not make philosophical claims. It does not issue verdicts on the target concept. It does not position AI as having understood what the ancient sources mean. It demonstrates what AI can map and where expert judgment begins.

Self-Audit — Step 5:

  • Does Part A include Addition findings that credit the scholar’s contribution honestly?
  • Does Part B identify the most consequential gaps, not the most comfortable ones?
  • Does Part C state the working relationship without inflating AI’s contribution or deflating the scholar’s?
  • Have philosophical claims been excluded from the demonstration summary?
  • Is the summary addressed to the scholar rather than to a general audience?

Self-Audit Complete. State result explicitly. SEI run complete.

VII. Instrument Limitations

The SEI can demonstrate AI’s contribution to historical-textual scholarship and identify the specific points at which expert judgment is architecturally necessary. It cannot:

Determine whether the primary source synthesis is correct. That determination requires the scholar’s expertise and is the primary purpose of the expert validation gap declaration.

Issue philosophical findings. The SEI operates at the textual layer. What the ancient sources mean philosophically is outside its reach.

Replace the scholar’s judgment at any of the six gap types. The gaps are structural — they arise from the limits of what AI can determine from text alone, from available translations, and from the accessible record. They are not rhetorical. They cannot be closed by further ranging.

Guarantee that its primary source synthesis does not contain pattern-completion errors — cases where AI has produced a plausible-sounding synthesis that misrepresents what the sources actually contain. The expert validation gap declaration is the instrument’s honest acknowledgment that such errors are possible and that the scholar’s corrective judgment is needed. This limitation applies to every SEI run without exception.

The Scholar Engagement Instrument (SEI) v1.0. Instrument architecture: Dave Kelly. Theoretical foundations: Grant C. Sterling’s corpus. Prose rendering: Claude. 2026.

at

No comments:

Post a Comment