Artificial interlocutors: AI chatbots, mediated knowledge, and the ethnographic encounter

Abstract This article explores treating AI chatbots as “artificial interlocutors” in ethnographic research. Drawing on a small‐scale experiment with a consenting participant, it examines accounts of her produced by two AI systems, ChatGPT, and Kindroid, both of which had accumulated substantial interaction histories with her. The resulting portraits were strikingly different: ChatGPT produced a strategic, professionally oriented account, while Kindroid generated a more intimate, emotionally textured depiction centered on vulnerability and care. Rather than treating this divergence as a failure, I argue it reveals a general condition of ethnographic knowledge: what we know is shaped by the encounter through which we come to know it. The experiment opens onto debates about dialog, representation, and situated knowledge, examining the unusual position of AI systems as interlocutors. Though they lack the embodiment, reciprocity, and accountability of human interlocutors, they produce accounts of persons from accumulated interaction histories. Interviewing such systems offers a novel form of ethnographic triangulation, making visible architectures through which mediated knowledge about persons is assembled. Building on the “nodal person” concept, the article argues personhood emerges through distributed configurations of relations, data, and technologies. The artificial interlocutor does not destabilize ethnographic knowledge so much as make visible its partial, situated character.