Rumors

Rumors are stories whose truthfulness is ambiguous, unconfirmed, or never definitively verified. Unlike fabricated stories (entirely false) or hoaxes (false facts presented as legitimate), rumors capture information that may be partially true, partially false, or simply uncertain in truth value.

Rumor research focuses on three main areas: propagation dynamics (how rumors spread through networks), user perception and stance (whether users support, deny, question, or comment on rumors), and verification (determining which rumors are true, false, or unverifiable).

Key properties

• Unconfirmed truthfulness: Core distinguishing feature; truth value is ambiguous or never established
• High propagation on social media: Rumors are widely shared and discussed on Twitter, Facebook, and other platforms
• Fast initial spread: Rumors often spread faster than corrections, particularly in disaster/crisis contexts
• Linguistic markers: Rumor tweets and discussions differ in writing style, sentiment, and structure from confirmed news
• Network structure: Rumor propagation follows distinct patterns influenced by network topology and user credibility

Rumor verification pipeline

Research organizes rumor handling into sequential stages:

1. Rumor detection: Identifying tweets/posts containing rumorous content (typically via crowdsourcing or annotation)
2. Stance classification: Determining user position on the rumor (Support, Deny, Query, Comment — the SDQC framework)
3. Rumor source detection: Identifying which user/post originated the rumor
4. Veracity prediction: Classifying rumor as true, false, or unverifiable

Key papers

• The Web of False Information: Rumors, Fake News, Hoaxes, Clickbait, and Various Other Shenanigans — Typology distinguishing rumors from other false information types; surveys literature on rumor propagation, detection, and containment
• Detection and Resolution of Rumours in Social Media: A Survey — Comprehensive survey of rumor detection and verification pipeline; reviews datasets (PHEME, RumourEval), annotation schemes, approaches from feature engineering to deep learning
• SemEval-2017 Task 8: RumourEval: Determining rumour veracity and support for rumours — Foundational shared task establishing rumor verification benchmark with SDQC stance classification and veracity prediction subtasks
• RumourEval 2019: Determining Rumour Veracity and Support for Rumours — Extended shared task with Twitter and Reddit data; best systems employ BERT and pre-trained contextual embeddings

Related topics

• False information — broader umbrella category
• Propagation — how rumors spread through networks
• Rumor Verification — verification and stance classification
• Stance Detection — determining user stance on rumorous claims