CONCEPT

Felicity Conditions

J.L. Austin’s doctrine that speech acts succeed or fail not by being true or false but by satisfying a structured set of requirements—appropriate persons, accepted procedures, correct execution, and genuine inner backing—whose systematic absence in AI speech is the most precise account of why machine utterances so often feel hollow even when they are accurate.

A speech act cannot be false, because it describes nothing. But it can fail in a dozen other ways, and Austin named them with unusual precision. For a performative to be “happy”—to come off as the act it purports to be—there must exist an accepted conventional procedure with a conventional effect, the particular persons and circumstances must be appropriate for invoking that procedure, the procedure must be executed correctly and completely, and—where the procedure is designed for persons with certain thoughts or feelings—those persons must actually have them and must conduct themselves accordingly afterward. Austin called violations of the first two pairs misfires: the act simply does not occur, as when someone without judicial authority “sentences” a defendant. He called violations of the latter pair abuses: the act comes off but hollowly, as when a promise is made by someone

Keep reading with YOU ON AI