Linguistic alignment between people and computers

There is strong evidence that when two people talk to each other, they tend to converge, or align, on common ways of speaking (e.g., Pickering and Garrod, 2004). In this paper, we discuss possible mechanisms that might lead to linguistic alignment, contrasting mechanisms that are encapsulated within the language processing system, and so unmediated by beliefs about the interlocutor, with mechanisms that are mediated by beliefs about the interlocutor and that are concerned with considerations of either communicative success or social affect. We consider how these mechanisms might be implicated in human–computer interaction (HCI), and then review recent empirical studies that investigated linguistic alignment in HCI. We argue that there is strong evidence that alignment occurs in HCI, but that it differs in important ways from that found in interactions between humans: It is generally stronger and has a larger mediated component that is concerned with enhancing communicative success.