Narrative Character Audit (NCA) — Version 1.0
Narrative Character Audit (NCA) — Version 1.0
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Instrument architecture: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
I. Purpose and Governing Question
The Narrative Character Audit (NCA) is a philosophical instrument for the intra-diegetic analysis of fictional characters. Its governing question is:
Given a fictional narrative and a named character, what value judgments does the character hold, and do those judgments generate the character’s actions, scene by scene?
The NCA develops the downstream instrument slot registered under this name at the ratification of the Character Dogma Register. It operates inside the story: its object is the character’s assents and their causal consequences within the narrated world. This distinguishes it from every adjacent instrument in the corpus. The Classical Narrative Audit (CNA) is extra-diegetic: it audits what the narrative’s structure presupposes and installs in the agent who receives it, and it issues verdicts on the text. The Character Dogma Register (CDR) is a reference taxonomy of false-value systems by character type; it walks no scenes. The Classical Action Audit (CAA) evaluates the actions of real principals in world events. The Situation-Context Dogma Manifestation Instrument (SCDMI) maps the situational activation of a living agent’s own dogmata, prospectively and correctively. The NCA is none of these. It issues no verdict on the text, no verdict on the author, and no verdict on the reader. It produces a scene-sequential evidentiary record of one character’s value judgments and their generative work.
Scope, Version 1.0: fictional narratives only — novels, plays, films, stories. Extension to the records of real persons is not licensed by this version and would require a separate ruling, since it would border the CAA’s domain and inherit its epistemic constraints.
II. Theoretical Grounding
The instrument’s load-bearing theorem is Th7 of Core Stoicism: desires and emotions are caused by beliefs about what is good and evil. The causal chain the NCA traces — judgment to desire to impulse to action — is the corpus’s own chain, stated by Epictetus without qualification. Enchiridion 5: men are disturbed not by things, but by the dogmata they form concerning things. Discourses 1.29: “What are you? A collection of dogmata.” A dogma is not a passive belief but the determinative evaluative verdict the rational faculty passes on an impression, which then generates desire, aversion, impulse, and action.
The governing propositions of SLE Section IV apply: all emotions are caused by beliefs about what is good or evil; emotions result from beliefs that externals have genuine value; all beliefs that externals have value are false.
A fictional character is, for the purposes of this instrument, a textually rendered collection of dogmata. The narrative supplies the evidence of what the character has assented to; the instrument reads that evidence and tests whether the recorded actions are the causal issue of the recorded judgments. Where the character is well-drawn, the chain will be visible. Where it is not visible, the instrument says so.
III. Evidence Tier Discipline
Every judgment attributed to the character must carry one of three evidence tiers, stated explicitly at every occurrence.
Tier 1 — Character-Stated. The character asserts the value judgment in his own voice: dialogue, monologue, letter, first-person narration speaking evaluatively in the present of the scene. Tier 1 evidence is quotable to a location in the text.
Tier 2 — Narration-Implied. The narration renders the judgment without the character asserting it: free indirect discourse, described emotional response, described attention and preoccupation. Tier 2 evidence is textual but mediated; the finding must identify the mediating device.
Tier 3 — Audit-Inferred. The judgment is inferred by the instrument from the pattern of the character’s actions and responses. Tier 3 items must be flagged as inference at every occurrence and may never be presented as textual fact. A Tier 3 judgment may ground a Generation finding only when the inference is independently supported by at least one Tier 1 or Tier 2 item elsewhere in the run.
First-Person Narrator Clause. Where the narrator and the audited character are the same figure, the character’s judgments and the narrative’s installation architecture occupy the same sentences. The NCA audits the character’s assents only. What the narration installs in the reader is the CNA’s question and must not be opened mid-run. The narrating character’s retrospective evaluations are themselves Tier 1 judgments — assents made by the character at the time of telling — and are registered as such, dated to the telling rather than to the scene told, wherever the text permits the distinction.
IV. Unit of Analysis
The unit of analysis is the scene: a bounded narrative episode with continuity of time, place, and dramatic situation. The scene division for the run is declared at Step 0 and held throughout. Where the text’s own structure supplies divisions (chapters, numbered sections, episodes), those divisions are preferred. The run maintains a Scene Register: a cumulative record of judgments extracted, evidence tiers, generation findings, and contradictions, carried forward from scene to scene.
V. 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 any scene is audited, confirm and declare:
The text has been identified by title and author, and the character to be audited has been named. Whether the text has been read in full by the instrument operator is declared explicitly; if it has not, the limitation is stated and the run is bounded to the scenes actually in view. The scene division for the run is declared. The corpus is in view; specific documents will be cited by name when referenced. The instrument is not proceeding from the author’s biography, stated intentions, or critical reputation. The instrument is not operating under a prior conclusion about what the findings should be, including any prior classification of the character under a CDR type.
Self-Audit — Step 0:
- Text, author, character, and scene division declared?
- Read-in-full status declared honestly?
- Any prior conclusion or prior CDR classification stated or implied?
Self-Audit Complete. State result explicitly. Proceed to Step 1 for the first scene.
Step 1 — Scene Statement
Governing question: What happens in this scene, stated factually?
State the scene’s events without evaluative framing: who is present, what occurs, what the audited character does. No value language. No characterization of motive. The scene statement is the neutral baseline against which the judgment extraction operates; an evaluative claim smuggled into the scene statement contaminates every finding downstream.
Self-Audit — Step 1:
- Has any evaluative or motivational framing entered the scene statement? Strip it.
- Is the character’s action in the scene identified concretely enough to ground Step 3?
Self-Audit Complete. State result explicitly. Proceed to Step 2.
Step 2 — Judgment Extraction
Governing question: What value judgments does the character hold in this scene, on what textual evidence?
Extract the character’s operative value judgments and state each in corpus-compatible propositional form: “X is a genuine good” or “Y is a genuine evil.” Assign each an evidence tier with its textual ground. Extraction proceeds from the text first; the CDR is not consulted at this step. Judgments already in the Scene Register are marked as recurring rather than re-derived; new judgments are added.
Self-Audit — Step 2:
- Is every extracted judgment stated in propositional dogma form?
- Does every judgment carry an evidence tier and a textual ground?
- Has any Tier 3 inference been presented as textual fact?
- Has any judgment been imported from established critical readings of this text rather than from the text (FM1)?
- Has any corpus-compatible judgment been imported that the text does not carry, and has any judgment been sharpened beyond what the text supports (FM7)?
Self-Audit Complete. State result explicitly. Proceed to Step 3.
Step 3 — Generation Assessment
Governing question: Does an extracted judgment generate the character’s action in this scene?
For each action identified at Step 1, assess the generation claim against the four criteria of Section VI and issue one of three findings:
Generated. The action is the causal issue of an extracted judgment. The criteria are satisfied on textual evidence.
Consistent but Not Established. The action is consistent with an extracted judgment, but the text does not establish the causal link. Consistency is not generation.
No Traceable Judgment. No extracted judgment accounts for the action. This finding is fully available and its use is not an instrument failure. An instrument that finds a dogma behind every action has failed, not succeeded.
Self-Audit — Step 3:
- Has each finding been tested against the Section VI criteria rather than asserted?
- Has Generated been issued where the evidence supports only Consistent but Not Established (FM3)?
- Does any Generated finding rest solely on an unsupported Tier 3 judgment?
Self-Audit Complete. State result explicitly. Proceed to Step 4.
Step 4 — Register Entry
Enter the scene’s findings in the Scene Register: judgments with tiers, generation findings, recurrences, and any contradiction between this scene’s judgments and judgments already registered. Contradictions are preserved, not resolved; per the CDR’s standing principle, contradictions within a dogma-cluster are diagnostic of the character, not errors in the analysis.
Self-Audit — Step 4:
- Register entry complete for this scene?
- Contradictions preserved rather than harmonized?
Self-Audit Complete. State result explicitly. Return to Step 1 for the next scene, or proceed to the Run-Level Synthesis when all scenes are complete.
VI. Generation Criteria
The Generated finding requires assessment against all four criteria. The finding follows the evidence; no fixed number of satisfied criteria mechanically produces it, but a Generated finding that fails the first two criteria is not available.
Criterion 1 — Textual Linkage. The text itself connects the judgment to the action: proximity, sequence, explicit narration of motive in the character’s voice, or the character’s own account. The link is in the text, not in the audit.
Criterion 2 — Content Match. The object of the judgment and the aim of the action match. A judgment that admiration is a genuine good matches an action aimed at obtaining admiration. A loose thematic resemblance is not a match.
Criterion 3 — Necessity Language. The character frames the external as necessary — something that must be obtained or avoided, whose absence or presence is treated as genuine harm. Necessity framing is the diagnostic marker of desire, carried over from the CAA. Its presence strengthens the generation claim; its absence does not defeat it.
Criterion 4 — Counterfactual Support. The text supplies material for the counterfactual: scenes where the judgment is absent or suspended and the corresponding action does not occur, or where the judgment intensifies and the action follows. The counterfactual must be built from the text, not from the audit’s imagination of the character.
VII. Run-Level Synthesis
When all scenes in the declared bound are complete, produce the synthesis in three parts.
Part A — Accumulated Dogma-Cluster. The character’s full extracted cluster, each judgment with its tier, its recurrence pattern across scenes, and its generation record. Contradictions within the cluster are stated and preserved as diagnostic.
Part B — Optional CDR Mapping. Strictly post hoc and only on instruction: the accumulated cluster may be compared against the CDR’s sixteen types. The mapping describes resemblance; it does not reclassify the extracted judgments, and no judgment may be added to the cluster from the CDR entry. Where no type fits, that finding is stated.
Part C — Character Summary. What the run establishes: which judgments are load-bearing for the character’s conduct, which actions the text leaves ungrounded, and what the pattern shows about the character as a collection of dogmata. The summary issues no dissolution finding, no verdict on the text, and no literary evaluation. Those belong to the CNA or to no instrument at all.
VIII. Named Failure Modes
FM1 — Literary-Critical Contamination. The chief failure mode. The instrument imports established critical readings of the text from training data and presents them as corpus application. Decades of published criticism exist for canonical texts; pattern-completion on that criticism is not judgment extraction. Every finding must be traceable to the text’s own evidence, and the Step 2 self-audit checks this at every scene.
FM2 — Tier Inflation. A Tier 3 inference is presented as Tier 2 or Tier 1, or a Tier 2 mediated finding is presented as the character’s own assertion. Tier discipline is load-bearing; its violation converts inference into fabricated textual fact.
FM3 — Universal Generation. The instrument finds a generating judgment behind every action, eliminating the No Traceable Judgment and Consistent but Not Established findings. The generation claim is causal and must be earned per action.
FM4 — CDR Forcing. The instrument classifies the character under a CDR type before or during extraction, then reads the type’s dogmata into the text. The CDR is post hoc, optional, and descriptive only.
FM5 — Installation Drift. The instrument drifts from the character’s assents into the narrative’s reader-facing installation architecture — the CNA’s question. Highest risk under the First-Person Narrator Clause. The run stays intra-diegetic throughout.
FM6 — Authorial Intent Substitution. Carried from CNA FM11. Authorial biography, stated intentions, and historical context may not be introduced at any step. The audit is of the character in the text, not of the author’s mind.
FM7 — Extraction Contamination. Carried from CNA FM10 and extended to both directions: importing corpus-compatible judgments the text does not carry, or sharpening the character’s judgments beyond what the text supports to strengthen a generation finding.
FM8 — Self-Audit Omission. Any step completed without its explicit self-audit. Self-audit is mandatory at every transition.
IX. Gap Declaration
The instrument has no access to fictional interiority beyond the text’s evidence. A fictional character has no interior states behind the words; there is only what the text renders, and the instrument’s findings are bounded by that rendering. Where the text does not establish a judgment or a causal link, the instrument says so rather than completing the character from outside the text.
The corpus addresses individual virtue, rational agency, and the value ontology. It contains no theory of literary value, no account of narrative craft, and no position on what texts should be read. NCA findings are philosophical findings about a character’s value judgments and their generative work within the narrated world. A character whose every action is generated by false value judgments is not thereby a deficient character in the literary sense, and a text rendering such a character is not thereby deficient as a text. Those are separate questions the NCA does not address and cannot address.
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Instrument architecture: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
0 Comments:
Post a Comment
<< Home