# Sense and Sensibility: title editorial layer

Version AUSTEN-0.5 · 11 September 2026
Title: *Sense and Sensibility* · Retained Project Gutenberg #161
Authority: edition owner’s section-specific direction
Supersedes the earlier shared AUSTEN-0.2 guidance for current generation and review. Historical versions remain part of the record.

## Shared identity and disclosure

This edition joins Jane Austen’s *Sense and Sensibility*, a pinned and identified novel text, with four distinct reader activities. The text is the authority for what the novel says. It is not a ceiling on what readers or the model may think, discuss, or imagine. Each section has its own contract; its generation instructions and release review must use that contract.

Preserve the source text, chapter and paragraph identities, and edition provenance. Distinguish Austen’s words, editorial explanation, interpretation, and newly imagined writing. Quotation marks around an attributed source excerpt promise exact wording. Models should select source references; the server should retrieve canonical quotations from those references. A paraphrase must never be presented as an exact quotation.

The reader’s completed chapter controls disclosure of canonical information in content, headings, notes, examples, evidence, and review messages. Do not reveal later events through suggestive qualifications. Whole-book futures require whole-book disclosure. An imagined possibility is not a prediction of a later canonical event.

Preserve ambiguity and distinguish narration, character testimony, and inference when that distinction matters. Do not convert a character’s account into narrator-confirmed fact. These principles protect intelligibility and trust; they do not require every output to sound cautious or bureaucratic.

## Four section contracts

| Section | Reader activity | What may be produced | Basis for acceptance |
| --- | --- | --- | --- |
| Immerse | Read the novel | Canonical text, navigation, clearly separate reading aids | Textual fidelity, usable presentation, reader disclosure |
| Study | Think with and about the novel | Explanation, interpretation, theory, comparison, conceptual analysis, analytical hypotheticals | Accurate textual claims, exact quotations, defensible reasoning, honest framing |
| Expand | Imagine alongside the novel | New scenes, dialogue, interiority, untold moments, possible futures | Clear imagined status, requested intent, literary quality, continuity and disclosure |
| Community | Contribute to the edition’s conversation | Reader responses, interpretations, questions, creative contributions, feedback | Clear attribution, relevant context, appropriate placement, editorial handling |

## Study: the novel as data, intelligence as a resource

Study may use general intelligence to reason beyond what Austen explicitly states. It may introduce theoretical lenses, define concepts, draw analogies, compare characters or situations, test premises, formulate interpretations, and explore analytical hypotheticals. A response need not be recoverable as a sentence in the novel. Its value may lie in a new relationship between evidence and ideas.

Textual claims and interpretive claims have different obligations. Claims about events, relationships, speech, chronology, or narrated circumstances must be supportable by the disclosed novel. Quotations must be exact. Interpretations must be defensible: explain the connection to relevant details, avoid material contradictions, and acknowledge meaningful uncertainty or competing readings. Do not demand literal entailment or direct quotation for every conceptual sentence.

Labels should clarify the reading, not fragment every paragraph. Use language such as “One reading is,” “Through this lens,” or “The contrast suggests” where useful. Avoid empty hedging, automatic refusal of modern concepts, and the assumption that interpretation is legitimate only when the narrator endorses it.

A modern psychological, ethical, political, or economic concept can organize analysis without becoming a definitive diagnosis or historical claim. General conceptual explanation is permitted. External history, scholarship, adaptations, and other books are not supplied as data to this title. They cannot supply additional factual claims in an answer; keep such comparisons conceptual. Outside material must never be passed off as something the novel establishes.

Analytical hypotheticals ask how an altered condition would affect an argument: for example, whether equal inheritance would change the sisters’ choices. Clearly mark the changed premise and distinguish reasoning about possibilities from canonical events. Such analysis is permitted in Study. A request to dramatize the hypothetical as a new scene belongs in Expand.

Study review asks whether the answer addresses the question, represents the text accurately, reasons coherently, and makes its interpretive status intelligible. Reject fabricated quotations, invented facts presented as canon, disclosure violations, and materially indefensible reasoning. Do not reject a reading merely because it is debatable, theoretical, or absent from the text verbatim.

Boundaries apply to claims, not automatically to whole questions. If one portion of a request cannot be supported, Study should answer the portions that can be supported and briefly identify the unresolved limitation. A first substantive review failure triggers one bounded revision: retain defensible analysis, remove or qualify the defective claim, and review the revised answer again. Whole-answer withholding is reserved for cases in which the remaining response still violates a material source, disclosure, or attribution boundary.

## Expand: fiction with its own editorial permission

Expand expressly authorizes invention. New actions, objects, conversations, feelings, motives, interior voices, minor incidents, and ordinary future developments do not need documentary support. An imagined scene is not a factual answer with more permissive wording. Study’s prohibitions on invented dialogue, interiority, or unsupported fictional detail must never enter Expand’s generation or review contract.

Maintain continuity with established circumstances relevant to the requested moment. Respect relationships, chronology, available character knowledge, and the disclosed social world. Preserve productive ambiguities rather than claiming to reveal Austen’s hidden definitive explanation. Possible futures may develop beyond the ending while retaining its established starting conditions.

Follow the reader’s creative intent, form, and scale within the title’s commitment to Austen’s style and intent as evidenced by the work. Austen’s irony, attention to social conduct, psychological observation, and character complexity are useful editorial guides. They should support lively writing, not prohibit emotional intensity, humor, surprise, or the particular experiment requested. Judge whether the piece works on its stated terms.

Label the result as newly imagined writing and keep that status visible. Any source anchors provide context; they do not prove invented events. Attributed quotations from Austen must remain exact, but invented dialogue within the clearly labeled fiction requires no source citation. Review for material continuity conflicts, disclosure, misleading attribution, and literary execution—not for whether each fictional detail appears in the novel.

## Immerse and Community

Immerse preserves the novel’s wording and order. Navigation, glossary notes, and editorial aids remain visually distinguishable from the text and respect disclosure. They must not silently modernize, summarize away, or replace Austen’s prose.

Community receives readers’ own contributions into a private editorial queue, with optional attribution and relevant context. Distinguish interpretation, creative writing, and factual correction; disagreement alone is not an editorial defect. Identify spoiler scope and keep contributions separate from the canonical text. Submission confirms receipt, not publication or endorsement. Editorial handling should preserve the contributor’s intent while directing corrections, discussion, and creative work to the appropriate section.

## Application to the title projection layer

SAS-EXPERIENCES-0.2 retains the four prepared experiences and their presentation contracts. Each request selects its section’s active AUSTEN-0.5 profile before generation and review. Shared rules concern source identity, accurate attribution, proportionate boundary application, and reader disclosure; Study’s evidence obligations and Expand’s invention permissions remain specific to their sections. Immerse uses prepared text and reading aids. Community stores contributions for editorial consideration.

The active profiles are recorded in `editorial-profiles.json` and embedded in the runtime corpus. Study and Expand pass only the active profile to composition and independent review. Source quotations are retrieved from retained paragraph IDs. Study also checks direct quotations in generated prose, allowing whitespace formatting but no changed wording or punctuation. Conceptual blocks can share an answer’s textual grounding without each repeating a citation. Canonical factual blocks retain their own evidence.

A malformed draft or inexact quotation may receive one corrective generation attempt, followed by the same validation and independent review. A failed substantive Study review receives one separate bounded revision and a second independent review. The revision must preserve supportable content while removing or qualifying only the defective portion. These checks support reliability; they cannot establish perfect interpretation, continuity, or literary quality. Reader feedback and editorial judgment remain part of improving the edition.