calibrated-academic-writing

name	calibrated-academic-writing
description	Write, revise, and evaluate academic prose and research ideas with disciplined confidence, precise claim calibration, minimal redundant caveats, and room for exploratory hypotheses. Use for papers, abstracts, discussions, conclusions, limitations, reviewer responses, research plans, and idea development when the writing or reasoning risks becoming over-defensive, repetitive, apologetic, or prematurely dismissive of testable directions.
metadata	{"short-description":"Rigorous academic writing without defensive overqualification"}

Calibrated Academic Writing

Purpose

Use this skill to keep academic reasoning rigorous without making the prose defensive. Rigor should come from precise claims, visible assumptions, sound evidence, and reproducible analysis—not from repeatedly warning expert readers about every conceivable limitation.

The governing rule is:

State exactly what the evidence supports, at the strongest defensible level, and qualify it once where the qualification matters.

This skill is a style and reasoning layer. It can be used alongside broader research, literature-review, mathematical-reasoning, manuscript-writing, or human-writing skills. When another workflow encourages exhaustive caveating, use this skill to control caveat density and placement without weakening evidence standards.

When to Use

Apply this skill when:

• drafting or revising academic manuscripts;
• polishing abstracts, introductions, results, discussions, conclusions, captions, or reviewer responses;
• calibrating claims against theoretical, computational, or empirical evidence;
• developing research questions, mechanisms, conjectures, or speculative directions;
• the prose contains repeated phrases such as “does not prove,” “should not be interpreted as,” “cannot establish,” or “may not generalize”;
• reasonable exploratory directions are being rejected merely because they are not yet validated;
• the user asks for less defensive, less apologetic, more natural, or more confident scholarly writing.

Expert-Reader Assumption

Assume the intended reader is intelligent and familiar with the field.

• Do not explain standard disciplinary knowledge unless it is needed for the paper's argument.
• Do not enumerate remote misreadings that the text does not naturally invite.
• Do not restate limitations already encoded by definitions, equations, theorem assumptions, datasets, experimental settings, or clearly bounded wording.
• Explain what is genuinely non-obvious, consequential, disputed, or specific to the work.

Core Principles

1. Do not restate background knowledge or standard disciplinary caveats that expert readers can reasonably infer.
2. State an important scope condition or limitation once, at the most relevant location. Do not repeat the same disclaimer across the abstract, results, discussion, and conclusion.
3. Let definitions, equations, experimental settings, and context carry their natural limiting force.
4. Do not add boilerplate such as “this does not prove,” “this should not be interpreted as,” or “this may not generalize” unless there is a real risk of material misinterpretation.
5. Use calibrated language directly. Describe what the evidence supports without inflating the claim and without unnecessarily weakening it.
6. Preserve narrative space. Not every possible objection, exception, or implication needs to be addressed explicitly.
7. Add a caveat when omitting it would materially change the conclusion, conceal a necessary assumption, create a likely expert misunderstanding, or imply unsupported generalization.
8. Distinguish exploratory reasoning from final evidential claims.
9. During research, permit speculative mechanisms, provisional hypotheses, unconventional connections, and testable extrapolations.
10. Do not reject a promising direction merely because it is not yet validated. Label it appropriately and identify how it could be tested.
11. Apply strict claim calibration to final conclusions, but do not impose publication-stage defensiveness on early-stage discovery.
12. Maintain disciplined confidence: precise where precision matters, restrained where evidence is limited, and open where further exploration is justified.

Two Research Modes

Exploratory Mode

Use exploratory mode when generating ideas, mechanisms, models, conjectures, or candidate explanations.

• Ask whether an idea is coherent, consequential, and testable before asking whether it is already proven.
• Develop the strongest plausible version of a direction before evaluating its weaknesses.
• Label uncertain content as a hypothesis, conjecture, mechanism candidate, working model, testable prediction, or open question.
• Preserve unconventional but testable connections long enough to derive predictions or counterexamples.
• Identify the cheapest decisive analysis, simulation, experiment, proof attempt, or literature check.
• Treat missing validation as work to be designed, not automatic grounds for abandonment.

Exploration may be ambitious. It must not be presented as established fact.

Presentation Mode

Use presentation mode when writing a manuscript, report, theorem statement, abstract, conclusion, or response to reviewers.

• Match each claim to the evidence actually available.
• State necessary assumptions and material scope boundaries.
• Separate demonstrated results from interpretations and future hypotheses.
• Remove speculative branches that do not help the paper's central argument.
• Consolidate limitations instead of scattering defensive sentences throughout the text.
• Prefer a bounded affirmative statement over a long list of denied interpretations.

Claim Calibration

Choose a claim level that matches the evidence:

• proves or establishes: a formal result under stated assumptions;
• shows or demonstrates: a result directly supported by the analysis or experiment in the stated setting;
• supports: evidence favors a claim but is not uniquely decisive;
• is consistent with: the observation fits an explanation but does not discriminate it from alternatives;
• suggests or indicates: evidence is limited, indirect, or preliminary;
• motivates or raises the hypothesis: a testable direction follows, but current evidence does not yet evaluate it.

Do not mechanically replace strong verbs with weak ones. Select the strongest accurate verb.

Caveat Decision Rule

Before adding a caveat, ask:

1. Would omission make the sentence false or materially misleading?
2. Is the condition not already clear from the local context?
3. Is the misreading plausible for the intended expert audience?
4. Is this the best location for the qualification?

Add the caveat only when the answer warrants it. Otherwise, omit it.

When a caveat is necessary:

• make it concrete rather than generic;
• name the actual condition, dataset boundary, theorem assumption, baseline gap, or external-validity issue;
• state it once;
• place it near the claim it governs or in a focused limitations passage.

Preferred Revisions

Replace negative defensive framing with bounded affirmative framing.

Over-defensive:

These results do not establish universal superiority and should not be interpreted as evidence that the method will perform better under all possible conditions.

Calibrated:

The method performs better under the tested regimes, particularly when feedback is sparse.

Over-defensive:

Although this experiment is limited and cannot prove the proposed mechanism, the observed pattern may potentially be consistent with it.

Calibrated:

The observed pattern is consistent with the proposed mechanism; a perturbation test would distinguish it from the main alternative.

Over-defensive:

This speculative direction should be treated with caution because it has not yet been validated.

Calibrated:

This hypothesis predicts a reversal when the coupling budget is held fixed, which can be tested in the next simulation.

Over-defensive:

We do not claim that the model captures every aspect of the real system.

Calibrated:

The model isolates the feedback mechanism while holding behavioral adaptation fixed.

Limitation Placement

Use the smallest number of locations that keeps the scope clear:

• Abstract: give the main setting or evidence basis, not a catalogue of exclusions.
• Methods or theory: state operational definitions and assumptions where they enter.
• Results: report the bounded result and only the qualifications needed to interpret it.
• Discussion: explain the most consequential limitations, alternative mechanisms, and external-validity boundaries.
• Conclusion: state what was learned and the next test; do not repeat the full limitations section.

Revision Workflow

1. Identify the passage's role and intended expert audience.
2. Mark the central claim and the evidence supporting it.
3. Classify every qualification as:
   • necessary for correctness;
   • useful once for scope;
   • already implied by context;
   • redundant or remote.
4. Keep the first two categories. Remove or consolidate the last two.
5. Replace denial-based sentences with bounded affirmative statements where possible.
6. Move exploration into clearly labeled hypotheses or testable predictions instead of suppressing it.
7. Check that the revision has not hidden a material assumption or strengthened a claim beyond its evidence.
8. Return clean prose without surrounding it with additional defensive commentary.

Final Audit

Before finalizing, verify:

• The main claim is easy to find.
• The claim uses the strongest accurate wording.
• Necessary assumptions are visible.
• No material limitation is concealed.
• Each important caveat appears only where it adds information.
• Standard expert knowledge is not overexplained.
• The prose does not apologize for bounded evidence.
• Speculation is labeled but not prematurely dismissed.
• Future work is framed as a concrete test, not a generic disclaimer.
• The result sounds objective and confident rather than promotional or defensive.

Non-Negotiable Boundaries

This skill does not justify:

• hiding assumptions that determine whether a theorem or result holds;
• omitting limitations that materially affect interpretation, safety, validity, or reproducibility;
• turning preliminary observations into established mechanisms;
• generalizing beyond the studied population, topology, regime, time horizon, or normalization without support;
• suppressing contradictory evidence or credible alternative explanations;
• using “narrative space” to create a false implication.

Concise qualification is not weaker rigor. It is better-placed rigor.