The new ghosts in the machine: 'Pragmatist' AI and the conceptual perils of anthropomorphic description

Algorithms are becoming interwoven with increasingly many aspects of our affairs.
That process of interweaving has brought with it a language laden with anthropomorphic
descriptions of the technologies involved, which variously hint at ‘humanesque’
or ‘conscious-like’ activity occurring within or behind their operations. Indeed,
the term ‘Artificial Intelligence’ (AI) seems to refer to a quality that is thought
to be largely human; namely, intelligence. However, while anthropomorphic descriptions
may be useful or harmless, when taken at face value they generate a false
picture of algorithms as well as of our own thinking and reasoning practices by
treating them as analogues of one another rather than as distinct. Focusing on the
algorithm, and what it is misleadingly said to be and to be like, in this article we
outline three ‘perspicuous representations’ (Wittgenstein 1953: §122) of AI in specific
contexts. Drawing on Wes Sharrock’s ethnomethodological and Wittgensteinian
work, our aim is to demonstrate that by attending to the particular, occasioned
and locally accountable, not to say highly specified, usages of language that accompany
the ‘New AI’ in particular, we can avoid being haunted by the new task performing
ghosts currently being discursively conjured up in our algorithmic machines.