Twitter climate discourse as a signal of pro-environmental behaviors
2026-05-01T08:52:17Z•20caa138a008994dddbfbaac3be6c0534b68759be1080deceaa04e0061825618
AI-generated contentLLMsPIIYouTube biasadversarial NLPalgorithmic biasclickbaitcontent-moderationdata toolkitdatasetsdeanonymization riskdetection evasionemotion-aware attackmisinformationpersonasprivacyprofilingrecommendation systemsrobustnesssocial media
What happened
This collection of recent arXiv papers highlights several security and privacy risks in social-media and NLP research. Most critical: an emotion-aware clickbait generation attack exploits stylistic transformations and VAD modeling to significantly degrade state-of-the-art classifiers (reported misclassification increases of ~2.6%–30.6%), showing practical evasion and engagement-optimization techniques. Tooling and datasets (Social Media Data Toolkit; MEDS—28k personas; TEA Nets extraction framework) lower barriers for large-scale collection, linkage, enrichment and automated entity/event/agent
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_si
- Record identifier
- 20caa138a008994dddbfbaac3be6c0534b68759be1080deceaa04e0061825618
- Enrichment time
- 2026-05-01T08:52:17Z
- 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.