Structural alignment leads to lower cognitive load in a collaborative task

Abstract One of the characteristics of dialogue is that interlocutors tend to converge on the same linguistic choices, called alignment. In this paper, we aim to investigate whether structural alignment — the tendency to use the same syntactic structures — has a positive effect on cognitive load and task completion in a task-based conversation. To do so, we engage participants in a collaborative task where they have to interact with another interlocutor (actually a bot) and inform each other about the location of landmarks on a map. In one condition the bot aligns with the participant and in the other it does not. Participants are recorded with an eye tracker during the experiment so that we can evaluate cognitive load and performance in the task. We found that when participants interact with an aligning bot, their cognitive load decreases and task completion is facilitated, but only to a certain degree. The results of the study suggest that alignment is a strategy that can be used in order to facilitate task performance.