Paper short abstract:

We propose the idea of an AI/nthropology as a new way to conduct anthropological research. We describe risks and opportunities when using AI to see 'patterns of culture'.

Paper long abstract:

Since its inception, anthropology's mantra has been to understand culture from "the native's point of view" (Malinowski1920). The goal has been to create a comprehensive reflection of the vast diversity of human social worlds in their entirety. Historically, anthropology has sought to provide an "inside" account of a new cultural reality that from the outside may have had characteristics of a "black box". However, since the field underwent a reflexive turn focusing on the cultures of the anthropologists themselves (Foley2002), anthropology has turned its critical gaze toward technologists and we have seen an increase in forensic ethnographies of algorithmic decision making (Seaver2017), algorithmic accountability (Eilish, 2019) or data collection procedures (Angèle, 2017) where machine learning engineers are being studied like "natives" (Downey, 1998) and socio-technical systems as culture (although the field of cybernetics and Science and Technology Studies (STS) is not new, Moss et al, 2019). However, instead of applying anthropological methods to the study of AI, this paper lays out how AI can be applied in new ways to gather cultural insights at scale. We describe the idea of an AI/nthropology, its potentials and risks, and what it can look like when ML and anthropology complement each other. We describe how both ML and social sciences are fundamentally inductive methods, provide examples for how ML can be applied to understand group biases, cultural practices and intersectionality, and whether computational approaches in anthropology are able to address the subjectivity/objectivity problem (Fleck, 1937; Keller, 1985).