Reposted by lynnette
Crowdsourced labels often hide bias, fatigue & noise. This #CSCW2025 paper by @quarbby.bsky.social, @feedkoko.bsky.social et. al shows how annotator metadata (speed, agreement, effort) can boost ML models of users' deception & disclosure. More:
tinyurl.com
When Labels Lie: Making Machine Learning Smarter with Annotator Metadata
Supervised machine learning is only as good as the data it’s trained on. But here’s the catch: when we rely on crowdsourced annotators to…
