Is CHI/ACL a storytelling festival? From community gripe to measurable construct¶
🌐 Language / 语言：中文 · English
Provenance（来源与元数据）
| id | marginalia-004 |
| title | Is CHI/ACL a storytelling festival? From community gripe to measurable construct |
| date | 2026-08-15 |
| published | 2026-08-15 |
| kind | survey + proposal |
| issue | 7 |
Is CHI/ACL a storytelling festival?¶
A community gripe, a pile of quantitative studies nobody had connected, and a proposal that splits narrativism into what the text says and what the figures do.
The claim, and where it comes from¶
"CHI/ACL 是故事会" — top-venue papers succeed by storytelling rather than substance. The gripe has named carriers: a Zhihu thread "ACL 为什么叫故事汇" (2023), Wobbrock's critique of CHI's rejection culture, Nacke openly teaching the CHI "narrative arc". What it has never had: a measurement. That gap is the musing.
The evidence that already exists (once you squint)¶
- Hillier et al. 2016 (PLOS ONE): 732 climate-science abstracts, crowdsourced scores on 6 narrative elements → composite index; PCA first component carries 76.5% of variance; 4 elements correlate with citations (confounded with impact factor, R² = 0.62).
- Qiu et al. 2024 (JAMA): 11,535 grant applications; promotional language predicts funding, OR = 1.47.
- Peng 2024 (PNAS): high-promotion abstracts ≈ 2× the odds of funding.
- Stavrova 2025 (HSSC): 130k abstracts; hype predicts citations — with higher returns for men.
- Counter-evidence: Vincent-Lamarre & Larivière 2021 (QSS) — accepted AI-conference papers are less readable.
- Context drift: Vinkers 2015 — positive words in abstracts up ×9 over decades; Kobak 2025 — ≥13.5% of 2024 biomedical abstracts carry LLM traces.
The proposal: explicit vs. implicit narrativism¶
- Explicit — the linguistic surface (v1 five dimensions: hook, first person, sensory language, appeal, connectives, hype words).
- Implicit — the structural-visual layer: impersonal CARS moves, master narratives, figure devices (teaser figure, hero chart, caption micro-narratives), narrative denial (denial gap = visual-rhetoric intensity − explicit explanation markers), and the artifact layer (benchmark casting, metrics-as-plot, baseline character tables, the ablation morality play). L3: field ideology.
- Unified construct: the narrative depth gradient — the deeper the narrative hides, the higher the return.
Method sketch (v2)¶
RQ1–9: explicit/implicit separation · rhetorical drift · payoff decomposition · CHI-vs-ACL field contrast · LLM breakpoint · interviews · denial premium · artifact narrativity · stigma boundary. Hypotheses include: explicit narrativity peaks in introductions, implicit in results; the denial gap grows with technicality and carries an acceptance premium.
Dual-track: multimodal large-sample census + critical ethnography, with reflexivity. The figure pipeline is deliberately local: Docling + GROBID extraction → ArXivCap (6.4M figures) warm-up → ACL (~50k papers) + OpenReview + CHI; three-tier annotation (SigLIP figure typing → Qwen2.5-VL-7B structured pass → semantic VL pass) with a 500-figure gold set; DuckDB throughout; ~250k figures ≈ 3 GPU-days. Plus arXiv v1↔camera-ready diffs and community discourse as complementary signals.
People worth remembering¶
Sophie Qiu = Huilian Sophie Qiu (邱惠莲) — CMU PhD 2022, now at Northwestern working with Brian Uzzi (JAMA 2024, IC2S2 2025). Theoretical anchor: Dourish & Gómez Cruz 2018 — yes, the same Dourish whose prose is dissected in musing 002; small field. Also in the net: Birhane 2022; Espeland & Sauder 2007 (reactivity); Segel & Heer 2010 and Hullman & Diakopoulos 2011 (narrative visualization).
Next¶
M1–M5: coding scheme + 500-figure gold set → 250k-figure census → payoff decomposition → interviews → writing. Three exits: P1 measurement (IC2S2 / QSS / Metascience), P2 main paper (CHI / CSCW), P3 critical (BD&S). Read DiagramBank to draw competitor boundaries.
Provenance¶
The original source documents (verbatim, in Chinese) are published alongside this note in docs/ — the full survey and the proposal v2, not just this summary.
| field | value |
|---|---|
| Data | CHI_ACL_故事会量化_调研.md; AI论文显隐叙事主义_提案v2.md; external chain as listed in frontmatter |
| Initial prompt | "Community wisdom says CHI/ACL papers are 'storytelling festivals'. Quantify it: has anyone measured narrative in papers? Then turn it into a research proposal." |
| Time | survey + proposal 2026-08-15 · note published 2026-08-15 |
| Agent / model | ZCode CLI · GLM (Zhipu) |
| Issue | #7 |
🌐 阅读中文版