What kind of documents can AI language model outputs be? The concept of artificially blended testimony
Sascha Donner · Journal of Documentation · 2025
Purpose The development of large language models (LLMs) has significantly enhanced capabilities in AI-powered text generation. The impact of this new technology, which is expected to significantly influence our work and private lives, on document creation is still largely unknown. This article is inspired by the article “What Kind of Science Can Information Science Be” by Buckland (2012). Design/methodology/approach Buckland’s (2012) considerations about the human’s central role in information science are applied to the question of how the human’s central role in documentation could be affected by the devolvement of LLMs. The Model of Documentation Activity (MoDA) (Donner, 2023) is used as a framework to evaluate the influence of LLM outputs as part of the documentation activity. LLM outputs are placed within the model after an analysis of their potential to be a document from conventional, functional and semiotic points of view. Findings An advanced and more detailed version of the MoDA, the MoDA2, is presented, which is intended to clarify the potential implications of LLMs on the documentation activity. Originality/value This article coins the term “artificially blended testimony” for LLM output as novel data provider along nature and testimony and demonstrates the value of the MoDA2 for exploring the impact of technological advances such as LLMs on the documentation process.