CoChat

Exploring incel group dynamics: a computational study of hierarchy and group-boundary policing

Open access
Solopova, Veronika, Popa-Wyatt, Mihaela, Berškytė, Justina · 2025 · Journal of Computational Social Science

The DOI matches its Crossref record; 40 of the 42 references it cites were found in a registry or an index, and 2 are not checked yet; 5 of the 26 sources its claims cite were checked for support.

Abstract

Abstract Incels (involuntary celibates) are part of a broader misogynistic culture known as the manosphere. Some communities within the manosphere, including incels, promote gender-based violence through misogynistic rhetoric and ideology. Incels are men who struggle to form romantic relationships and thus seek solace in online forums to find a sense of purpose and community. The community is organised around an ideology and a hierarchical classification of members. This paper presents a computational linguistic analysis of the utterances made within the community. We analyse the linguistic patterns of six different ranks of users: paragons, high-rank users, low-rank users, greycels, banned users, and self-banned/deleted users. We conducted an analysis of the sentiments expressed between ranks and of the affective attitude of posters when discussing a variety of topics. We then we analysed linguistic features, also conditioned on the rank of the poster. These analyses reveal qualitative differences between users of different ranks in how they express themselves and their sentiments. Finally, we trained models based on the linguistic features of users’ posts and show that already based on this information it is possible to predict the rank of users. Overall, our results indicate that the incel community has developed distinctive linguistic patterns to disseminate their ideology and that these linguistic patterns vary according to the rank of an incel within the community. This methodology offers a robust framework for studying similar echo chamber-like communities.

Verified 2026-11-22 · 2 references not checked yet

Overview

CoChat Summary

## TL;DR

This paper uses a computational linguistics analysis of 2,917 posts/comments from 745 users in 96 threads on an incel forum to test whether the community is hierarchical and how members police boundaries. It finds rank-linked differences in interaction patterns, emotion, jargon use, and linguistic style, and shows that user rank can be predicted from language features, with the best binary model reaching F1 = 0.73.

## Key findings

  • The corpus covered 2,917 posts/comments from 745 unique users in 96 threads, spanning 2018–2023 and totaling almost 112k tokens; 1,341 content pieces were reply pairs.
  • Interaction patterns were strongly rank-structured: high-rank users were involved in 41% of all interactions, paragons in 36%, and the most frequent answer pair was high-rank to high-rank, accounting for almost 10% of all posts.
  • Incel jargon use decreased with rank: banned users had the highest term frequency at 0.00825 per token, while paragons had the lowest at 0.00025 per token.
  • Machine learning could predict rank from linguistic features: the best binary experiment (new vs old users) achieved F1 = 0.73; the worst reported binary setup had micro F1 = 0.35, macro F1 = 0.46, and weighted F1 = 0.40.
  • LIME showed higher-rank predictions were associated with incel terminology, longer sentences, negation, subordinate clauses of purpose, statistics, abstract nouns, and reports; lower-rank predictions were associated with questions, comparatives, money, relative clauses, modal verbs, proper nouns, and passive voice.

## Methods

  • Computational corpus study of a well-known English-speaking incel forum (Incel.is), using scraped posts, comments, usernames, timestamps, and thread links; usernames were anonymized.
  • Users were annotated into hierarchy-related groups based on post count and forum status: greycels, low-rank, high-rank, paragons, banned, and deleted/self-banned users.
  • Analysis combined NRC lexicon and T5-based emotion detection, BERTopic topic modeling, linguistic feature extraction, incel jargon frequency counts, Mann–Whitney U tests, and SVM classifiers with an 85/15 train-test split and F1 averaged over 5 seeds.

## Limitations

  • The authors state that with a high number of comparison pairs and a relatively small sample size, “not many can draw significant results,” and several sentiment comparisons were trends rather than statistically significant.
  • The paper states the data may be available only on reasonable request; it is not publicly available because of re-identification risk.
  • Evident from the text: some class sizes were small and imbalanced, which the authors say likely hurt classification performance for banned, deleted, and greycel users.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1007/s42001-024-00357-5

Cites

40 of 42 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Image 1
From the PDF
Image 2
From the PDF
Fig. 2 The heat map illustrates the emotion proportion expressed in the comments of specific rank (y-axis) towards other ranks (x-axis)
Fig. 2 The heat map illustrates the emotion proportion expressed in the comments of specific rank (y-axis) towards other ranks (x-axis)From the PDF
Fig. 3 The heat map illustrates emotion colouring for each frequent topic on the forum of each rank
Fig. 3 The heat map illustrates emotion colouring for each frequent topic on the forum of each rankFrom the PDF
Fig. 4 The heat map illustrates z-score normalised averages of each linguistic feature distribution per rank
Fig. 4 The heat map illustrates z-score normalised averages of each linguistic feature distribution per rankFrom the PDF

Verification

Claims the paper makes

3 Valid · 8 Not valid · 12 Not checked · 110 Own results and interpretations

65 more claims

Create a free account to see all 133 claims and to verify the 2 references still open.

See all claims — free account
  • Probably not a substantial claim. 32
  • It may state two separate claims as one. 40
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 42 references

Research works printed but not linked: 1

  1. 16. Hinds, Z. (July 2022). True or false incels: An analysis of the incels.is forum. Master's thesis, The University of Chicago. Master's thesis

Cited by

See all 6

Metadata

Authors
Solopova, V., Popa-Wyatt, M., & Berškytė, J.
Affiliations
  • Technische Universität Berlin (DE)
  • University of Manchester (GB)
Corresponding author
Solopova, Veronika
Venue
Journal of Computational Social Science
Published
2025-02-06
Volume
8
Issue
2
ISSN
  • 2432-2717 (ISSN-L)
  • 2432-2725
OpenAlex
W4407229237
Cited by
6
References
30
Citations by year
2026: 4 · 2025: 2
FWCI
2.43
Type
journal article
Language
en
Open access
Licence
Locations
  • Journal of Computational Social Science (publishedVersion, cc-by)
  • Research Explorer (The University of Manchester) (publishedVersion, cc-by)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DepositOnce (submittedVersion, cc-by)
  • Technische Universität Berlin – Universitätsbibliothek (cc-by)
Funders
  • University of Manchester
  • Technische Universität Berlin
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
  • Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
Keywords
incels, manosphere, online communities, group hierarchy, misogyny, gender-based violence, sentiment analysis, affective attitudes, echo chambers, online forums, linguistic patterns

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-08, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-08 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.