write-introduction

name	write-introduction
description	Write the thesis Introduction (Chapter 1) using a research-informed, multi-pass workflow grounded in academic writing best practices

Write Introduction Skill

Write Chapter 1 (Introduction) for the thesis "Automation and Political Ideology: Task Displacement and the Conservative Shift Among Exposed Workers."

This skill implements a structured, multi-pass writing workflow distilled from leading academic writing guides and adapted for a Master's thesis.

Arguments

• draft — Run the full pipeline: research, draft, style-match, verify (default)
• outline — Produce only a detailed outline with paragraph-level annotations
• revise — Revise an existing draft of Ch.1 (assumes 01_introduction.tex already has content)

Academic Writing Foundations

This skill is built on best practices from:

• Keith Head ("The Introduction Formula"): Hook → Question → Antecedents → Value-Added → Roadmap
• Jesse Shapiro ("Four Steps to an Applied Micro Paper"): Distinguish aspirational paragraphs ("I show that X causes Y") from contractual paragraphs ("Section 2 does A, Section 3 does B"). Aspirational paragraphs are the heart of the introduction; contractual paragraphs are its skeleton.
• Claudia Goldin ("Writing Tips"): Lead with your strongest result. The paper is persuasion, not a mystery novel. No generic section headings.
• Marc Bellemare ("How to Write Applied Papers"): In a thesis, the introduction is longer than in a journal article — it should be self-contained enough that a reader who skips everything else still understands the contribution.
• Gallea's one-shot template: Hook → Gap → Contribution → Results → Roadmap — compressed for a journal paper but expandable for a thesis.

Key Principles

1. No mystery novels (Goldin): State the main finding in the first 2 paragraphs. The reader should know "automation exposure → conservative shift" before paragraph 3.
2. Every sentence earns its place (Shapiro): No filler. No "it is worth noting." No "interestingly."
3. Aspirational before contractual (Shapiro): Tell the reader what you found before telling them how the chapter is organized.
4. Self-contained (Bellemare): A busy committee member who reads only Ch.1 should still understand the research question, method, key findings, and contribution.
5. Bridge, don't duplicate (Head): The Antecedents section is a 3-4 paragraph map of the literature, NOT a mini literature review. Point to Ch.2 for depth.

---

Target Structure (~450-550 lines, ~8-10 pages)

Chapter 1: Introduction
├── 1. Hook (2-3 paragraphs)
│   Opening with the contemporary phenomenon: populist/conservative surge
│   + economic disruption from automation. Ground in a specific fact or
│   striking observation. Then pivot to the scholarly puzzle.
│
├── 2. Research Question (1 paragraph)
│   Clear, precise statement: "I ask whether workers in occupations more
│   exposed to task-displacing automation shift toward conservative
│   political ideology."
│
├── 3. This Thesis: Approach & Findings (3-4 paragraphs)
│   [ASPIRATIONAL — Shapiro]
│   What I do: task displacement measure from BEA-KLEMS + ACS + O*NET,
│   IRT ideology from CCES, pseudo-panel 2006-2022, TWFE estimation.
│   What I find: β > 0 (automation → conservative shift), heterogeneous
│   effects, lag structure.
│   How I identify: shift-share logic, pre-determination of shares.
│
├── 4. Antecedents (3-4 paragraphs)
│   [Head's "Antecedents" — brief, not Ch.2-depth]
│   a. Task-based framework (ALM 2003) → automation measurement
│   b. Economic shocks → politics (ADH 2013, Colantone-Stanig, Frey et al.)
│   c. Cultural backlash vs economic insecurity debate
│   Each paragraph should end with what the prior work DOESN'T do that
│   this thesis addresses.
│
├── 5. Contribution (3-4 paragraphs)
│   [Moved from Ch.2 §2.4 — Head's "Value-Added"]
│   a. Comprehensive automation measure (task displacement, not just RTI/robots)
│   b. Continuous ideology outcome (IRT estimates, not binary vote choice)
│   c. Pseudo-panel spanning 2006-2022 (not single cross-section)
│   d. Mechanism adjudication (economic insecurity vs cultural backlash)
│   For each: why it matters, what it enables that prior work couldn't do.
│
├── 6. Roadmap (1 paragraph)
│   [CONTRACTUAL — Shapiro]
│   Chapter-by-chapter preview using \Cref{ch:...} for all chapters.
│   Brief, functional, no more than 1 sentence per chapter.
│
└── \Cref connections to all subsequent chapters

---

PHASE 1: RESEARCH (read existing chapters)

Before writing a single word, you MUST read the following files to understand what the introduction needs to reference and not duplicate:

Mandatory Reads

1. thesis/CLAUDE.md                                → conventions, commands, bib keys
2. .claude/reference/literature-evidence.md        → concrete findings from 20+ papers (use for Antecedents)
3. thesis/chapters/02_literature.tex               → what Ch.2 covers (avoid duplication)
4. thesis/chapters/03_data.tex                     → data sources, sample sizes, match rates
5. thesis/chapters/04_methodology.tex              → formulas, specifications, identification
6. thesis/chapters/05_results.tex                  → main coefficients, key tables/figures
7. thesis/references.bib                           → available citation keys
8. CLAUDE.md (root)                                → project context, measure pipeline, findings

What to Extract

Source	Extract for Introduction
Ch.3	Data sources (CCES 2006-2022, ACS, BEA-KLEMS, O*NET), sample size, IRT ideology variable
Ch.4	Task displacement formula (briefly), 18 demographic groups, pseudo-panel design, TWFE spec
Ch.5	Main coefficient (β=0.291, p=0.06), lag structure (k=1 strongest), heterogeneous effects
Ch.2 §2.4	Contribution text to migrate to Ch.1 (shorten §2.4, expand Ch.1 §5)
references.bib	Available citation keys — NEVER cite a paper not in the bib file

---

PHASE 1.5: BUILD ARGUMENTATIVE STRUCTURE

Before drafting, use the narrative-architect agent to design the argumentative flow for the Antecedents and Contribution sections. The agent reads .claude/reference/literature-evidence.md and produces paragraph-by-paragraph outlines with specific paper citations.

How to Use

Dispatch the narrative-architect agent with:

• Goal: "Build the Antecedents section of the Introduction: briefly position the thesis in three literatures (task-based framework, economic shocks → politics, mechanisms) and identify what prior work does NOT do."
• Target section: "Chapter 1, §4 (Antecedents)"
• Constraints: "3-4 paragraphs max. Must cite ALM 2003, ADHM 2020, Colantone & Stanig 2018, and Gallego & Kurer 2022. Each paragraph must end with a gap statement."

The architect will return a structured outline. Use it as the backbone for Phase 2 drafting.

Argumentative Flow Templates (from literature-evidence.md §7)

The evidence base includes pre-built flow templates for:

• §7.1: The Introduction Paradox (automation → rightward, not leftward)
• §7.5: The 4 Contributions structure

Use these as starting points. The narrative-architect agent refines them into paragraph-level outlines with specific citations.

---

PHASE 2: DRAFT

Writing Rules (MANDATORY)

1. Voice: First person singular. "I define", "I construct", "I estimate", "I find". Never "we" or "the author."
2. Tense: Present tense for methodology ("I define 18 groups"), past tense for data ("The CCES was administered..."), present for findings ("The results indicate...").
3. Citations: \citet{key} when author is subject, \citep{key} for parenthetical. NEVER write "Author (Year)" manually.
4. Cross-references: \Cref{ch:data} at sentence start, \cref{ch:data} mid-sentence. Use liberally — the introduction is a navigation hub.
5. Numbers: Cite exact figures inline ("517,736 employed individuals", "83.4% match rate"). Never round when precise numbers are available.
6. No fabrication: Every coefficient, p-value, and sample size must come from Ch.5 or the analysis outputs. If a number is not in the existing chapters, leave a % TODO: verify [number] comment.
7. Equations: At most 1-2 display equations in the introduction (the TD formula and maybe the TWFE spec). Keep them simple; full derivations are in Ch.4.
8. Hedging: "consistent with", "suggesting", "indicating". Never "clearly proves" or "undoubtedly."
9. Paragraph structure: Topic sentence → 2-4 supporting sentences → transition sentence. Every paragraph.
10. No em-dashes: Do NOT use --- in prose. Use commas, parentheses, colons, or semicolons for asides and elaborations. Match Chapters 3-5 style.

Anti-Patterns (NEVER do these)

• Start with "In recent years..." or "Automation has become increasingly important..."
• Write more than 2 sentences of abstract methodology without grounding in a concrete example
• Use the word "interesting", "important", or "noteworthy" as filler
• Write a paragraph that is just a list of citations: "Author1 (Year) found X. Author2 (Year) found Y. Author3 (Year) found Z."
• Include a subsection structure — the introduction is ONE continuous flow
• Duplicate the literature review — keep Antecedents to 3-4 paragraphs max
• Use rhetorical questions (except very sparingly)
• Write the roadmap paragraph as anything other than the LAST paragraph
• Reference tables/figures from Ch.5 by number (they might renumber) — use \Cref{tab:label}

Hook Strategies (choose one)

A. The Paradox Hook: "Despite decades of rising returns to education and skill-biased technological change favoring cognitive work, the political response to automation has defied economic models: workers facing displacement move rightward, not toward the redistributive left."

B. The Fact Hook: "Between 2006 and 2022, the most automation-exposed demographic groups in the United States experienced a [X]-point rightward shift in ideology..."

C. The Event Hook: Start with a specific moment (2016 election, a factory closure, a policy debate) that crystallizes the tension.

D. The Literature Gap Hook: "The China Shock literature established that trade-induced labor market disruption shapes political attitudes. But automation—a structural shock that is slower, broader, and harder to attribute—has received far less systematic attention."

Opening Paragraph Template

[Hook: 2-3 sentences establishing the phenomenon]
[Pivot: 1 sentence framing the scholarly question]
I ask whether workers in occupations more exposed to task-displacing
automation shift toward conservative political ideology. Using a
pseudo-panel of [N] demographic groups tracked through the Cooperative
Congressional Election Study from 2006 to 2022, I find that a
one-standard-deviation increase in group-level task displacement is
associated with a [β]-point rightward shift in ideology, a pattern
consistent with the cultural backlash mechanism proposed by
\citet{colantone2021backlash} and \citet{manunta2025identity}.

---

PHASE 3: STYLE MATCHING

After drafting, the text must be passed through a style normalization pass to match Chapters 3-5. Either:

1. Self-check: Re-read 40+ lines each from 03_data.tex, 04_methodology.tex, and 05_results.tex. Compare sentence length, clause complexity, transition patterns. Revise the draft to match.
2. Use the style-paraphraser agent: If the draft feels "AI-generic" or doesn't match the thesis voice, dispatch to the style-paraphraser agent.

Style Checklist

• Average sentence length 25-35 words (mix short declarative + longer complex)
• No em-dashes in prose (use commas, parentheses, colons, semicolons)
• Colons before definitions/lists
• (i)/(ii)/(iii) for inline enumeration, not bullet points
• No "we", no passive where "I" works
• Topic sentence opens every paragraph
• Transition sentence closes every paragraph
• No \textbf{} in running prose
• Exact numbers, not rounded
• \emph{} on first use of technical terms

---

PHASE 4: VERIFICATION

Compile

cd /Users/alessandro/Projects/Tesi/thesis && tectonic main.tex

Check for:

• No compilation errors
• No undefined citations (Undefined citation warnings)
• No undefined references (Undefined reference warnings)
• All \Cref targets exist in Ch.2-5
• §2.4 contribution text has been shortened (not duplicated in both Ch.1 and Ch.2)

Content Verification

• Research question stated clearly within first 2 paragraphs
• Main finding previewed within first 3 paragraphs (Goldin: no mystery novel)
• At least 4 forward references to subsequent chapters
• Contribution section has 3+ distinct points
• Roadmap mentions all 5 subsequent chapters
• No number appears without a source (Ch.3-5 or analysis outputs)
• Introduction is self-contained (readable without Ch.2-5)

---

Coordination with Ch.2 Restructuring

When this skill is used, Ch.2 §2.4 (Contribution) should be shortened to a brief (1-2 paragraph) summary that says "this thesis contributes in four ways" and then points to Ch.1 for the full articulation. The detailed 150-line contribution section currently in Ch.2 should be migrated to Ch.1 §5.

After writing Ch.1, update Ch.2's opening paragraph to include a backward reference:

As outlined in \Cref{ch:introduction}, this thesis contributes to
three literatures. This chapter reviews each in turn.

---

Example: Real Introduction Openings from Related Papers

Anelli, Colantone & Stanig (2019) — "We Were The Robots"

Opens with the economic mechanism: "automation represents a source of structural change in the economy that generates aggregate gains but with winners and losers." Then pivots to political consequences through multiple channels (anti-incumbent, economic nationalism, authoritarian attitudes). Uses specific mechanisms, not generic statements.

Kurer & Palier (2024) — "The AI Shock and Socio-Political Attitudes"

Opens with the broader context: "This paper evaluates the role of digital economic shocks in affecting Americans' social, cultural, and political beliefs." Connects to existing literature on economic backlash from outsourcing/trade, then identifies the gap (AI's complementing vs displacing effects). Clear bifurcation hypothesis stated upfront.

Autor, Dorn, Hanson & Majlesi (2020) — "Importing Political Polarization"

Opens with the puzzle: import competition from China affected both labor markets AND politics. States the finding in the first paragraph: trade shock moved districts toward more extreme candidates. Methodology follows immediately.

Pattern: All strong introductions state the finding within the first 3 paragraphs.