Plurification in/of language technology -- The integration of culture in next-generation AI

The paper discusses the integration of cultural considerations into Natural Language Processing (NLP) through a socio-technical model, emphasizing that merely increasing cultural data is insufficient for achieving cultural alignment in AI systems. It critiques current approaches for primarily focusing on surface-level representation of culture, while deeper issues related to power and governance remain unaddressed. The authors advocate for a plural epistemological framework in NLP design to effectively operationalize culture, highlighting the need for a reflexive approach that accommodates diverse linguistic and socio-cultural contexts.