Structuring ethics for large language models in knowledge organization – philosophical foundations, norms and practical imperatives

Purpose This paper aims to develop a multi-level ethical framework for the integration of large language models (LLMs) into knowledge organization (KO) and knowledge organization systems (KOS). It addresses the ontological shift from human-curated semantic artefacts to probabilistic generated texts and the resulting ethical implications for knowledge infrastructures. Design/methodology/approach The study employs a conceptual analysis rooted in the philosophy of information and semiotics. It examines the mediation of meaning through statistical correlation versus semantic understanding to identify unique ethical risks in KOS. Findings The analysis reveals that LLMs transform the status of organized knowledge, threatening epistemic diversity and semantic authority. The paper proposes a three-level framework comprising a meta-ethical level, a high-end principles level and a guidelines level to govern LLM deployment in KO and KOS. Originality/value While general artificial intelligence ethics frameworks exist, this research provides a specialized structure for the KO community. It bridges the gap between high-level philosophical reasoning and practical implementation, specifically addressing the semiotic integrity of knowledge systems.