SPILLOVER: Measuring Cyberbullying NormPropagation on Social Media
2026-07-23T08:52:13Z•379bc6cf10238d46f81abbd0122f961543363742548a786206984c1deb733f11
content-moderationcyberbullyingdataset-releasedatasetsdisinformationhate-speechinstagramonline-harmreal-time-detectionredditsocial-contagionsocial-media
What happened
This collection highlights social-media research with direct security and platform-moderation implications. The lead paper (SPILLOVER) measures how cyberbullying (CB) propagates comment-to-comment: a CB comment substantially increases the odds the next comment is CB, driven mainly by cross-user spread, CB→CB pairs are more textually similar than NoCB→CB pairs, and non-aggressive replies become more negative as prior CB severity increases. Findings replicate across Instagram, Reddit, Wikipedia Detox, and SOCC. A simple binary feature (whether the prior comment was CB) improves real-time CB-pred
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_si
- Record identifier
- 379bc6cf10238d46f81abbd0122f961543363742548a786206984c1deb733f11
- Enrichment time
- 2026-07-23T08:52:13Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.