Linguistic Alignment at the Conceptual Level: Comparisons between Human-Human and Human-Computer Interaction

Abstract

People prefer to use basic names over superordinate ones (e.g., apple versus food) when referencing objects. Nonetheless, superordinate name usage does occur sporadically, yet the mechanisms that regulate this usage remain largely unexplored. In two experiments, we examined whether superordinate name usage may come about due to people copying the linguistic behavior (i.e., linguistic alignment; in this case the use of superordinate names) of their interlocutor, and if this alignment (if any) differs in strength in human-human compared to human-computer dialogue contexts. Participants engaged in a picture matching and naming game with an interlocutor, who was either a human or a computer using strictly basic or superordinate names (Experiment 1), or a mixture of both (Experiment 2). We discovered alignment of superordinate name usage when interacting with an interlocutor using only superordinate names or an interlocutor mixing superordinate names and basic names, with this alignment progressively intensifying over time with both interlocutors. The level of alignment remained consistent across human and computer interlocutors. These findings signify the existence of such linguistic alignment and demonstrate that people respond similarly to humans and computers regarding linguistic behavior, suggesting that superordinate name usage alignment is likely driven by an automatic, non-interlocutor specific mechanism.

Keywords: Dialogue, human-computer interaction, alignment, superordinate concepts, language production