On Defining Erasure Harms for NLP
A new paper published on arXiv proposes a structured definition for "erasure harms" in Natural Language Processing (NLP) systems. The authors aim to provide a clearer conceptual foundation for identifying and measuring this specific type of harm, which has been a growing concern with the deployment of NLP technologies. The proposed definition is intended to be adaptable across various settings, addressing the limitations of existing broad or highly specific conceptualizations. AI