Human–AI Conflict Discourse Dataset

Human–AI Conflict and Constraint Discourse Dataset (YouTube, 2010–2024) is a large-scale corpus of 209,228 English-language YouTube comments related to artificial intelligence, collected via the YouTube Data API from videos identified using the term “Artificial Intelligence.” The dataset includes cleaned textual content, publication metadata, sentiment classifications generated using a pretrained DistilBERT model, and topic cluster assignments derived from Latent Dirichlet Allocation, mapping public discourse to key AI constraint mechanisms (parametric reductionism, agency transference, and regulated expression). The corpus captures naturalistic public perceptions of AI across personal and institutional contexts and is intended to support research on human–AI interaction, AI ethics, and computational analysis of digital discourse.