Institutional Anthropomorphic Error

Abstract Sakr proposes institutional anthropomorphic error (the projection of role-based authority onto AI systems that cannot bear a role’s constitutive obligations) as a third kind of error to be added to my distinction between metaphysical and pragmatic anthropomorphic error. I argue that the proposal is better understood as belonging to a second category of anthropomorphism than a third kind of error within an existing one. To ‘anthropomorphise’ a system may mean either seeing it as human-like (anthropomorphic cognition) or presenting it as human-like (anthropomorphic presentation). Metaphysical and pragmatic error are two kinds of error in anthropomorphic cognition; institutional anthropomorphic error, as Sakr defines it, is an error of presentation. A cognitive counterpart can be formulated, but it resolves into a distinctive form of either metaphysical or pragmatic error. Sakr’s contribution is valuable, but it does not require revision of the two-part account.