--- title: "Examining Gender and Power on Wikipedia through Face and" description: "The paper argues that NLP work on politeness has mostly used Brown and Levinson's theory only for surface markers of… 4 references." authors: ["Adil Soubki", "Shyne E. Choi", "Owen Rambow"] year: 2024 venue: "Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue" source: "https://aclanthology.org/2024.sigdial-1.4/" doi: "10.18653/v1/2024.sigdial-1.4" arxiv: "2408.02798" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/soubki-2024-examining-gender-power-wikipedia-through/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/soubki-2024-examining-gender-power-wikipedia-through/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2024](https://md.research.elamur.ai/free/en/papers/2024/page/16/) / Examining Gender and Power on Wikipedia through Face and Politeness # Examining Gender and Power on Wikipedia through Face and Politeness **Authors:** [Adil Soubki](https://md.research.elamur.ai/free/en/authors/adil-soubki/), Shyne E. Choi and [Owen Rambow](https://md.research.elamur.ai/free/en/authors/owen-rambow/) · **Year:** 2024 · **Venue:** [Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue](https://md.research.elamur.ai/free/en/venues/sigdial/) ([SIGDIAL 2024](https://md.research.elamur.ai/free/en/venues/sigdial/2024/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2024.sigdial-1.4/) · [PDF](https://aclanthology.org/2024.sigdial-1.4.pdf) · [DOI 10.18653/v1/2024.sigdial-1.4](https://doi.org/10.18653/v1/2024.sigdial-1.4) · [arXiv 2408.02798](https://arxiv.org/abs/2408.02798) · [OpenAlex](https://openalex.org/W4402715046) **Citations:** 0 citations (OpenAlex, 2026-10-14) ## Summary The paper argues that NLP work on politeness has mostly used Brown and Levinson's theory only for surface markers of politeness. It ignores the underlying face acts, meaning what an utterance does to the speaker's or hearer's public self-image. The authors treat face acts and perceived politeness as two linked layers of analysis. They annotate utterances from the Wikipedia Talk Pages Corpus with one of eight face acts and fine-tune LLMs to build a face act tagger. They run this tagger alongside existing politeness predictors over roughly 1.3 million talk-page sentences, then compare editors by self-reported gender, admin status and experience. Findings: - Women are more polite overall, and they produce notably more self-humbling face acts. - Among admins, this gender gap shrinks sharply. - Experience lowers politeness for men but not for women. The work matters because it moves politeness research beyond word-matching cues. Instead, it grounds the analysis in what speakers do to each other's face, at a much larger scale than earlier face-act corpora. ## Key points - **Label scheme:** eight face acts, built from three binary choices: speaker vs. hearer, positive vs. negative face, and threat (−) vs. raise (+). Each gets a mnemonic name: - hearer acts: IMPOSITION, DISAGREEMENT, PERMISSIVENESS, AGREEMENT - speaker acts: INDEBTEDNESS, APOLOGIES, AUTONOMY, CONFIDENCE - **Data source:** the Wikipedia Talk Pages Corpus, with 125,292 exchanges among 38,462 editors and 391,294 posts in total. The authors annotate it for face acts and release the data, code and a pretrained tagger at github.com/cogstates/wikiface. - **Prior work and scope:** the annotation follows the CMU Face Acts Corpus of Dutt et al. (2020), whose BERT-based classifier reached 69% F (0.60 macro). Like Dutt et al., the authors assign one label per utterance, since only about 2% of utterances there carried multiple labels. - **Experience effect:** experienced men drop INDEBTEDNESS (thanking, commitments) from 12.3% to 9.8% of their utterances. Women's rate stays flat, which reverses which gender uses this act more. - **Statistics:** differences were tested with the Mann-Whitney U test. The authors caution that effect sizes are generally small. ## Interesting ideas - Keep the intended social act (face act) and the perceived tone (politeness) as separate labels, since the same IMPOSITION can read as a polite request or a rude challenge. - Build the label inventory as the product of a few binary dimensions (whose face, which face, raise or threaten), and give each cell a human-readable mnemonic. - Find interesting cases by crossing two independent annotations and studying where they disagree, rather than evaluating each label alone. - Justify a simplifying assumption (one label per utterance) with a measured rate in existing data, and list it openly as a limitation. - Report significance together with effect size, and warn when significant differences are tiny at corpus scale. ## For Mova How these ideas could serve Mova, our small open language model — its files, stages and tests: - **I1** (frames): crate prag could attach a face-act slot next to each speech act, so irony or rudeness is explained as a mismatch between act and wording, not inferred from words alone. - **I2** (frames): a factorized face-act label is easy to derive from frame roles (speaker/addressee, who is affected) and stays explainable to users. - **I3** (learn): we could pair the pragmatic layer with an external politeness or sentiment witness, and route disagreements to an LLM teacher or human review as high-value training cases. - **I4** (learn): when the pipeline collapses ambiguity (one frame or one sense per clause), we should log how often that choice discards alternatives, so the simplification stays auditable. - **I5** (learn): evaluation reports over large silver corpora should show effect sizes beside p-values, so trivial gains are not mistaken for real improvements. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [Stony Brook University](https://md.research.elamur.ai/free/en/universities/stony-brook-university/) — University, Stony Brook, United States · 46 works · 3,024 citations 📚 Shelves (5): [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,954 works · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 541 works · [🤝 Gender Studies](https://md.research.elamur.ai/free/en/shelves/social-sciences/gender-studies/) · 52 works ## Cite Cite as: Adil Soubki, Shyne E. Choi and Owen Rambow. 2024. Examining Gender and Power on Wikipedia through Face and Politeness. Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{soubki2024examining, title = {{Examining Gender and Power on Wikipedia through Face and Politeness}}, author = {Adil Soubki, Shyne E. Choi and Owen Rambow}, booktitle = {Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue}, year = {2024}, doi = {10.18653/v1/2024.sigdial-1.4}, eprint = {2408.02798}, archivePrefix = {arXiv}, url = {https://doi.org/10.18653/v1/2024.sigdial-1.4} } ``` - [← previous Evaluating the Effectiveness of the Foundational Models for Q&A Classification in Mental Health care](https://md.research.elamur.ai/free/en/papers/alhuzali-2024-evaluating-effectiveness-foundational-models-q/) - [next → Experiences from Creating a Benchmark for Sentiment Classification for Varieties of English](https://md.research.elamur.ai/free/en/papers/srirag-2024-experiences-creating-benchmark-sentiment-classification/) - [↑ back to papers of 2024](https://md.research.elamur.ai/free/en/papers/2024/page/16/) ## References 4 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. AI@Meta. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). 2. Hutheifa Y. Al-Duleimi, Sabariah Hj Md Rashid, and Ain Nadzimah Abdullah. 2016. A critical review of prominent theories of politeness. Advances in Language and Literary Studies, 7:262–270. [find](https://scholar.google.com/scholar?q=Hutheifa+Y%2E+Al%2DDuleimi%2C+Sabariah+Hj+Md+Rashid%2C+and+Ain+Nadzimah+Abdullah%2E+2016%2E+A+critical+review+of+prominent+theories+of+politeness%2E+Advances+in+Language+and+Literary+Studies%2C+7%3A262%E2%80%93270%2E) 3. J. L. Austin. 1962. How to do things with words. Oxford University Press. [find](https://scholar.google.com/scholar?q=J%2E+L%2E+Austin%2E+1962%2E+How+to+do+things+with+words%2E+Oxford+University+Press%2E) 4. Penelope Brown and Stephen C Levinson. 1987. [Politeness](https://md.research.elamur.ai/free/en/papers/brown-1987-politeness-some-universals-language-usage/): Some Universals in Language Usage. Cambridge University Press. *Updated 2 November 2026*