press-conference-revision-evidence-bound

Evidence-Bound Press-Conference Revision

中文说明

Contribution-forward academic revision without overclaiming.

Co-created on RedNote by @Lensback's Lab and @Naruto AI.

press-conference-revision-evidence-bound is an academic revision skill for diagnosing and reducing defensive writing while preserving the evidentiary limits that make scholarly claims credible.

It is designed for manuscripts whose contribution is already present and supported, but is obscured by repeated caveats, self-defence, work-log narration, excessive qualification, unnecessary comparisons, or contribution statements buried beneath anticipatory responses to hypothetical reviewers.

The skill does not make a paper more assertive by inflating its claims. Instead, it helps a manuscript present its strongest supported contribution clearly while retaining the scope conditions, source-status distinctions, rival explanations, contradictions, methodological limits, and citation functions that determine what the evidence actually permits.

Core principle: remove rhetorical defensiveness, not scholarly caution.

---

Why “press-conference revision”?

Academic manuscripts often accumulate defensive prose during drafting, co-author revision, peer review, and repeated attempts to anticipate criticism.

The result can be a paper that reads less like an argument and more like a record of everything the authors worried about while producing it:

• the contribution appears only after several qualifications;
• limitations are repeated far beyond the point at which they constrain interpretation;
• paragraphs begin by explaining what the paper does not claim;
• methodological caution becomes self-deprecation;
• failed analytical routes remain visible as work-log narration;
• comparisons with stronger cases or literatures are volunteered even when they do not advance the argument;
• the paper answers objections that its own claims never created.

The press-conference metaphor asks a simple question:

If the evidence is fixed, what should the paper say first about what it has actually established?

A good revision therefore does not conceal weaknesses or manufacture certainty. It reorganizes the manuscript so that the supported proposition leads, while the necessary qualification appears where it performs real analytical work.

The goal is:

contribution → evidence → qualification

rather than:

qualification → self-defence → more qualification → eventual contribution

---

What this skill protects

The method is intentionally evidence-bound.

Every revision must preserve, or narrow, the original manuscript's epistemic commitments. In particular, the skill protects:

Claim ceiling

A revision cannot increase the certainty, causality, novelty, generality, representativeness, or portability of a claim beyond what the existing evidence supports.

may shape cannot silently become determines.

the interviews suggest cannot become the evidence demonstrates.

in this case cannot become cities generally.

Evidence status

The workflow keeps analytically different kinds of evidence separate, including:

• proposal;
• authorization;
• reported activity;
• direct observation;
• implementation;
• outcome;
• interpretation;
• causal inference.

Fluency is never a sufficient reason to collapse these distinctions.

Scope conditions

Case, temporal, population, institutional, geographical, and comparative boundaries remain visible wherever they are needed to interpret a claim correctly.

Rival explanations and negative evidence

Contradiction, delay, non-adoption, decoupling, reversal, failed implementation, and competing explanations are not treated as embarrassing residue when they are part of the empirical record.

Citation and quotation roles

A stylistic revision must not silently change what a citation, quotation, date, number, source-specific term, table, figure, or footnote is being asked to support.

---

What the skill does

The skill supports two distinct modes.

1. Read-only diagnostic

The manuscript is audited without changing its prose.

The diagnostic produces:

• a contribution-and-scope contract;
• a manuscript-wide candidate register;
• explicit KEEP decisions for necessary caution;
• a section-level concentration map;
• an exception and query register;
• a bounded recommendation for any later revision.

This mode is useful when the author wants to understand whether defensive writing is actually a manuscript-level problem before authorizing changes.

2. Authorized revision

Once revision authority is explicit, the skill can perform a concentrated revision while preserving the evidence ceiling.

Typical outputs include:

• a tracked manuscript;
• a clean manuscript;
• a change summary organized by rhetorical pattern;
• a claim/evidence regression report;
• a record of deliberate non-edits;
• unresolved author decisions requiring human judgment.

The distinction is important: diagnosis does not automatically authorize rewriting.

---

Workflow

flowchart LR
    A["Phase 0<br/>Contribution & Scope Contract"] --> B["Phase 1<br/>Batch Diagnostic"]
    B --> C{"Author authority<br/>to revise?"}
    C -- No --> D["Diagnostic Register<br/>+ Queries + Recommendation"]
    C -- Yes --> E["Phase 2<br/>Concentrated Revision"]
    E --> F["Phase 3<br/>Whole-Manuscript Regression"]
    F --> G["Tracked + Clean Manuscript<br/>+ Integrity Report"]

Phase 0 — Contribution-and-scope contract

Before editing begins, the skill establishes a compact contract containing:

If the contribution hierarchy, evidence ceiling, or editing authority cannot be established, the workflow stops at diagnosis rather than improvising a solution.

---

Phase 1 — Batch diagnostic

The manuscript is first read as a whole.

The skill does not perform a blind search-and-replace operation. Lexical cues are treated only as signals for contextual inspection.

Candidates are classified using the following taxonomy: