geb-writing-style

Writing Style (geb-writing-style)

When to trigger

• The introduction reaches the theorem only after pages of setup and notation
• The abstract is over 250 words, or states the model but not the result
• The proof idea is invisible behind algebra
• Prose is dense even for a specialist game-theory reader

GEB house style: precise, idea-first, specialist-but-broad

GEB is read across the game-theory community — economics, political science, biology, computer science, mathematics, and psychology — so the writing must be rigorous yet legible beyond one sub-field. Lead with the idea and the result; push heavy algebra into proofs and appendices. State theorems precisely but motivate them in words first. The reader is a game theorist, so you can assume the vocabulary of equilibrium and mechanisms, but not the details of your specific construction.

Format facts that shape the writing:

• Abstract must not exceed 250 words — concise and factual; state the contribution and the main result, not just the topic.
• References: no strict style is imposed at submission as long as it is consistent; the published style is elsarticle-num. (Avoid mixing styles.)
• Generative-AI use must be disclosed in a section titled "Declaration of generative AI and AI-assisted technologies..." (待核实 — Elsevier uses both "...in the writing process" and "...in the manuscript preparation process"; copy the exact heading from the live Guide for Authors), placed before the reference list (basic grammar/spell tools exempt) — keep the prose your own and the declaration accurate.
• This is a late-stage skill: polish only once theorems, assumptions, and (if any) experiments are settled.

The introduction arc (GEB template)

1. The question / phenomenon — the strategic problem, in plain terms.
2. Why it was open — what prior solution concepts or results could not deliver.
3. The result — state the main theorem/finding early, in words, then formally.
4. The key idea — the one move (construction, mechanism, design) that makes it work.
5. Scope & relation to the literature — generality and the delta against the nearest result.
6. Roadmap — brief.

Abstract: state the result (≤250 words)

• Open with the setting and the question, then state the main result — the theorem, the mechanism, or the experimental finding.
• Name the advance ("we characterize", "we prove existence", "we show equilibrium play converges to ...").
• No throat-clearing; a game theorist should know what you proved from the abstract alone. Keep it under the 250-word cap.

Two prose lanes: theory and experiment

GEB publishes both proofs and lab evidence, and each lane has its own exposition discipline:

• Theory lane — theorems are the spine. Each formal environment gets a paragraph of interpretation before the formalism; each proof gets a one-sentence "idea of proof" before the argument; every assumption a theorem invokes is stated, not buried in a footnote — GEB referees audit assumption-to-result traceability.
• Experimental lane — hypotheses come from the game's equilibrium (or behavioral) predictions and are stated before the design; the design section reads chronologically (subjects, matching, rounds, payment); results sentences pair the effect's direction with the theoretical benchmark ("play converges toward the risk-dominant profile"), not statistics alone.
• Papers combining both lanes should let the theory section stand alone: the prediction must be graspable without the data section, and vice versa.

Micro-rewrite: an opening sentence

• Before: "In this paper we consider a model in which n agents interact repeatedly under incomplete information and study its equilibria." — announces a model, promises nothing.
• After: "We show that in repeated interactions with incomplete information, reputation effects select a unique equilibrium payoff — resolving a multiplicity that folk-theorem arguments cannot break." — states the result, names the obstacle, and tells a game theorist why to read on.

Sentence-level craft

• Define notation once; do not force the reader to hold many symbols to parse a sentence.
• Active voice; short declaratives for theorem statements and key claims.
• Motivate before formalizing; give the proof's idea before its steps.
• Calibrated claims — say exactly how general the result is, no more.

Anti-patterns

• An abstract over 250 words, or one that states the model but never the result
• Opening the intro with notation and assumptions instead of the question and result
• A proof with no stated idea, only symbols
• Mixed/inconsistent reference styles
• Forgetting the generative-AI declaration when AI tools were used

Output format

【Abstract】states the result + ≤250 words? [Y/N] — fix: ...
【Intro arc】question / why-open / result / key idea / scope present? [Y/N each]
【Result stated early】in words then formally? [Y/N]
【References】consistent (elsarticle-num target)? [Y/N]
【AI declaration】needed? present + placed before references? [Y/N / NA]
【Next step】geb-replication-and-data-policy or geb-review-process