Least Effort and Alignment in Task-Oriented Communication

Conversational partners align the meanings of their words over the course of interactionto coordinate and communicate. One process of alignment is lexical entrainment, wherebypartners mirror and abbreviate their word usage to converge on shared terms for referentsrelevant to the conservation. However, lexical entrainment may result in inefficient mimicry thatdoes not add new information, suggesting that task-oriented communication may favor alignmentthrough other means. The present study investigates the process of alignment in Danishconversations in which dyads learned to categorize unfamiliar “aliens” using trial-and-errorfeedback. Performance improved as dyad communication became less verbose, measured as adecrease in the entropy of word usage. Word usage also diverged between partners as measuredby Jensen-Shannon Divergence, which indicates that alignment was not achieved through lexicalentrainment. A computational model of dyadic communication is shown to account for the aliengame results in terms of joint least effort. The model shows that alignment of partner referentscan increase as a result of minimizing both the joint entropy of dyadic word usage, and theentropy of word usage conditioned on the separate partner referent distributions. We concludethat the principle of least effort, originally proposed to shape language evolution, may alsosupport alignment in task-oriented communication.