Research Statement

Research Statement

Core Purpose

Write a research statement that gets the candidate an interview. The document's only job is to convince a non-specialist search-committee member, in 5 minutes of skimming, that:

1. The candidate has a clear research vision (one sentence they could repeat at lunch).
2. The candidate has the track record to execute it.
3. The candidate's next 3–5 years will produce fundable, citeable, hireable work.
4. The candidate fits this department specifically — not a generic department.

Everything else — eloquence, autobiography, exhaustive coverage — is secondary and usually counterproductive.

Final delivery: humanize the output

REQUIRED SUB-SKILL (run before returning any draft to the user). Load the humanizer skill (via the skill tool) and run the finished text through it as a final editing pass. It removes AI-writing tells (inflated significance, copula avoidance like "serves as", rule-of-three, overused AI vocabulary such as "delve"/"testament"/"landscape", filler phrases, sycophancy, and em/en-dash overuse) without inventing any fact, name, number, date, or citation. Preserve the format and conventions this skill requires (length tier, section order, department-fit specifics), and if the user supplied a writing sample, pass it so humanizer matches their voice.

The Mission/Strategy/Evidence/Story Frame

Every effective research statement, regardless of discipline or region, can be checked against four questions. If any answer is missing, the draft is incomplete.

This frame is adapted from Arie van Deursen's TU Delft hiring rubric and matches what NSF CAREER, NWO Veni, and ERC StG reviewers score on. Use it as the master checklist for every draft.

Workflow

Stage 1: Diagnose the Brief (15 min, mandatory)

Before drafting anything, extract from the job ad:

1. Document name — "research statement", "research vision", "research plan", "future research statement", "5-year plan". These are not interchangeable. See references/terminology.md.
2. Length cap — 1, 2, 3, 4-5, or 7-10 pages? This dictates content allocation. See references/length_tiers.md.
3. Region/venue — US R1, UK, EU/Dutch, Canadian, Australian, Asian. Voice register and fit-paragraph weighting differ. See references/regional_voice.md.
4. Discipline — CS / engineering / life sciences / math / social sciences / humanities have different expectations for past-work density, future-work specificity, and the role of grants/equipment/archives. See references/discipline_variants.md.
5. Special instructions — words like "from the heart", "intrinsic motivation", "vision", "fit with our research lines", or "potential for funding" each carry specific obligations the brief is testing for. Treat each as a checkbox.

For any US faculty or postdoc application, also load references/us_school_guidance.md. US university guidance is not a single R1 template: Cornell / CMU / Penn / Yale often describe 1–4 page committee-readable statements, MIT / Harvard / Caltech often expect STEM statements with 3–5 pages and multi-horizon planning, and Berkeley Math expects a nontechnical executive summary before technical depth. The job ad and discipline still win.

If the brief uses "from the heart" / "personal motivation" / "what drives you" language, also load references/personal_voice.md — these phrases are common in Dutch and Northern European applications and require a calibrated register that is neither US-emotional nor sterile-CV.

Stage 2: Draft Inside-Out (5 short sessions)

Drafting in the order the document is read produces bad statements. Drafting inside-out produces good ones:

1. North Star — one sentence stating the animating question. Iterate 5+ versions before moving on.
2. Future research lines — 2–3 (rarely 4) lines, each shaped to be NSF-CAREER / Veni / ERC-StG / NIH-K fundable. See references/future_research.md for the 6-element shape.
3. Past-work themes — 2–3 themes, each told as Problem → Approach → Result → Impact (PARI). Organize by theme, not chronology. See references/past_work.md.
4. Fit paragraph + closing credo — name 2–3 specific groups, 1–2 specific faculty (with one specific paper or project each), 1 cross-cutting initiative, and one realistic funding pathway. See references/department_fit.md.
5. Opening paragraph — written LAST. Must do four jobs in 4–6 sentences: name the problem domain, name the gap, state the vision, hint at method. See references/openings.md.

For US statements, run the brand/story overlay from references/us_school_guidance.md after the North Star: state the candidate's research identity in one sentence, then test each past theme and future line against CMU's Sequence / Novelty / Problem moves and Caltech's ABT (And–But–Therefore) story logic.

Stage 3: Red-Team & Score (mandatory before delivery)

A draft is not done until it has passed two checks:

1. Score against the rubric in references/rubric.md. Any criterion below threshold → revise that section before continuing.
2. Run the red-flag scan in references/red_flags.md. Every flagged phrase must be removed or replaced.

If revising for someone else, always perform both checks and report scores explicitly — do not just say "looks good".

Universal Structural Skeleton

This is the load-bearing structure every effective research statement implements, scaled to length tier.

[1] Opening / Vision         — North Star + framing  (≈10–15% of length)
[2] Past Work                — 2–3 themes, PARI shape (≈30–40%)
[3] Future Research          — 2–3 fundable lines    (≈35–45%)
[4] Department Fit           — specific groups/people (≈8–12%)
[5] Closing / Credo          — personal commitment   (≈3–5%)

For 1-page UK statements, [4] often collapses into the cover letter. For 7-10 page humanities statements, [2] expands into a "book project" arc. For most CS / engineering positions, this allocation holds. See references/length_tiers.md for tier-specific allocations.

Region & Discipline Selection

Read these references only when they apply:

If the user does not specify, ask once which discipline + region apply. Do not guess past CS/US default.

Critical Anti-Patterns (always block)

Never let a draft ship with any of these. Full list with replacement language in references/red_flags.md.

1. ❌ Childhood / "ever since I was young" opening
2. ❌ Definition of the field as opening sentence
3. ❌ "Passion", "passionate", "love", "fascinated" as standalone declarations without immediate evidence
4. ❌ "Novel", "groundbreaking", "revolutionary", "paradigm-shifting"
5. ❌ "Extending" or "adding to" advisor's work — claim independence
6. ❌ Negative positioning — "Previous work fails to..." — stay in what YOU did
7. ❌ Generic department praise ("X University is world-renowned in...")
8. ❌ Future research lines under 80 words each — too vague to fund
9. ❌ Zero named faculty / groups / initiatives in the fit paragraph
10. ❌ No funding-vector mention where the brief asks for "potential for funding"
11. ❌ Closing with thanks / hopes / "I look forward to..."
12. ❌ Over-length — more disqualifying than under-length

What "From the Heart" Actually Means

In Dutch, German, and some UK briefs, the instruction "written from the heart" / "intrinsic motivation" / "what drives you" is not a request for emotional disclosure. It is a request for a personal scholarly credo: the conviction that a specific question MUST be answered, evidenced by a sustained pattern of work directed at it. The model is closer to Hamming's You and Your Research or Dijkstra's EWD essays than to a US diversity statement.

Concretely, this means show, don't tell:

• ❌ "I am passionate about reliable software."
• ✅ "I have spent the last six years on a single question: why does software fail in ways its authors did not imagine?"

Full guidance and a phrasebook of authentic motivation patterns is in references/personal_voice.md. Load it whenever the brief uses heart/passion/motivation/vision language explicitly.

Self-Editing Protocol

Run a 5-pass revision in this order. Each pass takes 15–30 minutes. Earlier passes catch structural issues; later passes catch surface issues. Do not invert.

1. Vision pass — Can you state the North Star in one sentence? Does the opening paragraph deliver it? If a non-specialist reads only paragraph 1, can they summarize?
2. Claim-evidence pass — Every assertion of impact backed by specific evidence (paper, citation count, deployment, award)? Every future-research line has a method, an outcome, a horizon, and a funding vector?
3. Fit pass — At least 2 specific groups + 1–2 named faculty + 1 cross-cutting initiative + 1 realistic funding pathway?
4. Voice pass — First-person singular consistent? Hedging level matches region? No anti-pattern phrases (references/red_flags.md)?
5. Mechanics pass — Length within cap? Tense consistent (past for done, present for ongoing, future for planned)? No typos? Could not be pasted into a different application without major editing?

If any pass fails, fix and re-run all subsequent passes — fixes propagate.

References (load on demand)

Core references — load whenever drafting or reviewing:

• references/rubric.md — Self-grading scorecard with 8 criteria, 0–5 scale, total /40. Use to score every draft before delivery.
• references/red_flags.md — Annotated list of phrases and patterns that mark weak statements, with concrete replacements.
• references/openings.md — 10 successful opening paragraphs verbatim, decomposed; opening anti-patterns; the four-job opening checklist.
• references/past_work.md — PARI structure, theme selection, advisor-independence rules, quantification guidance.
• references/future_research.md — The 6-element fundable-line shape; horizon planning; high-risk vs. safer line balance; funding-vector vocabulary.
• references/department_fit.md — How to research a target department; the 2-groups-2-faculty-1-initiative formula; the gap-not-praise rule.

Conditional references — load only when applicable:

• references/personal_voice.md — When brief contains "heart", "passion", "intrinsic motivation", "drives you", or "vision". Heavy emphasis on Dutch / Northern European register.
• references/us_school_guidance.md — For any US faculty/postdoc statement or when the user cites Cornell, CMU, MIT, Caltech, Penn, Yale, Harvard, UConn, Notre Dame, Delaware, Berkeley, or similar US guidance. Source-synthesized overlay for audience, length, story, future planning, institutional fit, funding, design, and feedback.
• references/regional_voice.md — When the target region or register matters. Sections for US, UK, Dutch/EU, Continental EU, Canada/Australia/Asia.
• references/discipline_variants.md — Per-discipline rubric variations and example structures. Sections for CS/Eng, life sciences, math, social sciences, humanities.
• references/length_tiers.md — Content allocation tables for 1, 2, 3, 4–5, and 7–10 page caps; what gets cut at each tier.
• references/terminology.md — Disambiguates "research statement" vs. "research vision" vs. "research plan" vs. "5-year plan" vs. "research pitch" vs. "future research statement".
• references/examples.md — 10 dissected real opening paragraphs from successful CS faculty applications, with structural breakdowns. Patterns extend to other disciplines.
• references/bad_examples.md — 5 annotated bad-example excerpts with diagnosis-and-fix.