The (Ir)relevance of Ethnomethodology and Conversation Analysis for Technology Companies: Incommensurability in Action

Abstract To what extent does articulating the practical logic of humans coping with a situation at a pre-reflexive level contribute to designing rule-based agents that exhibit some form of skilled expertise in “similar” situations? Can a micro-analytic approach help program rule-based conversational AIs, as opposed to fully relying on designers’ and engineers’ commonsense and intuitive knowledge? Conducted over several years within a technology company specialized in the creation of social robots, this ethnography documents concrete disjunctions that occurred between, on the one hand, pre-existing methods, or theoretical backgrounds prevalent among industry professionals and, on the other hand, practices or concepts (sequentiality, locality, etc.) ordinarily connected with an ethnomethodological and conversation analytic approach. Although incommensurability is both well-theorized and regularly experienced by ethnomethodologists, this study describes incommensurability in action , as it emerged during meetings, data sessions, and methodological debates with engineers and designers, as part of the daily work of designing social robots. In particular, this work details how practical knowledge of human participants (drawn upon as a source of inspiration) unavoidably “leaked out” as it was integrated into “user experience” reports. In doing so, this ethnographic study finds that (superficially straightforward) methodological disagreements were often connected to more fundamental phenomenological and praxeological stances regarding the very possibility of (sufficiently) articulating practical action. It concludes that many of the recurring challenges examined in this work stem from long-documented flawed assumptions about the efficacy of deriving rules from a micro-analytic study of human conduct to, then, embed those rules into conversational agents.