Key points for an ethnography of AI: an approach towards crucial data

Recent years have seen an increase in calls for ethnography as a method to study Artificial Intelligence (AI). Scholars from diverse backgrounds have been encouraged to move beyond quantitative methods and embrace qualitative methods, particularly ethnography. As anthropologists of data and AI, we appreciate the growing recognition of qualitative methods. However, we emphasize the importance of grounding ethnography in specific ways of engaging with one's field site for this method to be valuable. Without this grounding, research outcomes on AI may become distorted. In this commentary, we highlight three key aspects of the ethnographic method that require special attention to conduct robust ethnographic studies of AI: committed fieldwork (even if the fieldwork period is short), trusting relationships between researchers and participants, and, importantly, attentiveness to subtle, ambiguous, or absent-present data. This last aspect is often overlooked but is crucial in ethnography. By sharing examples from our own and other researchers' ethnographic fieldwork, we showcase the significance of conducting ethnography with careful attention to such data and shed light on the challenges one might encounter in AI research.