Sociolinguistic variation and artificial agents

Most philosophical work on language produced or interpreted by artificial systems has focused on questions about the sense in which those systems might be able to produce or grasp linguistic meanings. For example, if Grice is right that a string is meaningful when it is produced by a speaker who intends for a listener to recognize that a certain communicative intention was involved in its production, questions loom about whether stings produced by LLMs could be meaningful.

In this talk, my goal will be to frame a quite different set of questions artificial systems raise about language. After briefly introducing the phenomenon of sociolinguistic variation, I’ll show how important it will be for us to think seriously about the forms AI speech is allowed to take, and about the ways in which AI interpreters are sensitive to the forms of our speech.