A new paper identifies the "projection problem" in stance detection, where annotators struggle to compress complex, multi-dimensional attitudes into single labels. This leads to disagreements that stem from different weighting of dimensions rather than confusion. The study found that dimensional agreement among annotators consistently exceeded standard label agreement, especially for complex targets like school closures. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Highlights limitations in current NLP annotation methods for complex social attitudes, potentially impacting downstream AI applications.
RANK_REASON This is a research paper published on arXiv detailing a new problem identified in stance detection.