| This study focuses on the social organization of knowledge production in the area of Artificial Intelligence (AI) science. It draws on ethnographic data to theorize the open-ended processes whereby members of AI labs imagine, design, and account for the innovative simulation technologies they make. I draw on extensive observational, interview, and archival data collected over a three-year ethnography of two AI labs, one working in the AI subfield of knowledge representation and the other in the area of information retrieval. The main argument is that the production of knowledge in AI does not involve the progressive rationalization of human-level intelligence so much as a creative balance between the intrigue of ambiguous problems and the precision required for computational formalisms. This tension between ambiguity and precision cannot be resolved. Rather, it is managed in ways that are particular to an organizational setting and its institutional environment. Three main areas of lab practice are compared across the two labs: organizational culture, standards for verifying legitimate science, and the design process. I also unpack the different forms of ambiguity and uncertainty that each lab manages. The project contributes primarily to organizational theory, science and technology studies, and interpretive social psychology. |
