What makes intelligence artificial?

Debates about whether artificial systems are intelligent usually assume a contrast with a natural intelligence located in organisms. This paper questions that assumption. It argues that intelligence is artifactual: built, like a tool or a notation system, through the displacement of cognitive capacities onto external supports, and stabilized in the practices of a community of reasoners. Drawing on wide computationalism, the paper takes the coupled system of agents, tools and practices as the unit of analysis. It traces how organisms deposit the results of their interactions in groups, constructed niches and tools, and argues that this process can be described computationally without appeal to internal representation, with interaction itself as a form of computation. Game semantics, computability logic and Girard's Ludics give the process a formal shape, and Sellars's account of the space of reasons shows how its products become norms: interaction deposits patterns, and the logical structure of the game of giving and asking for reasons turns some of them into norms.