Can Crowdsourcing Survive the LLM Era? A Community Survey on Human Data Collection
A survey of 155 NLP researchers explored how large language models (LLMs) affect crowdsourced data collection. Findings show 44% observe LLM usage, with concerns about data homogeneity and validity. Researchers use detection methods like style patterns, but current mitigation efforts are insufficient, suggesting crowdsourcing needs a rethink.