Rethinking Knowledge Brokerage: A Case Study of a Large Language Model in R&D

The work of knowledge brokers comprises the transfer, translation, and transformation of knowledge between individuals who are unlikely to interact efficiently because of knowledge boundaries. In an extension of this theory, algorithmic brokers are defined as individuals performing these practices with artificial intelligence (AI) output to enable a community to leverage this output in their work. However, with the introduction of large language models (LLMs), we argue this brokerage role is shifting and that LLMs have the potential to broker knowledge between humans. We conducted a case study with domain experts in a Research and Development (R&D) department of a large multinational science and technology company who regularly use a recently developed domain-specific R&D-LLM. Our preliminary findings show that the R&D-LLM is reshaping interactions between human experts through three knowledge brokerage practices of varying complexity, assisting in simple knowledge recall, enabling the approach to experts and being a simulated counterpart.