Least Effort and Alignment in Task-Oriented Communication

Conversational partners must align the meanings of their words over the course ofinteraction in order to coordinate and communicate. One process of alignment is lexicalentrainment, whereby partners mirror and abbreviate their word usage to converge on sharedterms for referents relevant to the conservation. However, lexical entrainment may be inefficientto the extent that this kind of behavioral matching does not add to communication, suggestingthat task-oriented communication may favor alignment through other means. The present studyinvestigates the process of alignment in Danish conversations in which dyads learned tocategorize unfamiliar “aliens” using trial-and-error feedback. Performance improved as dyadcommunication became less verbose, measured as a decrease in the entropy of observable wordusage. Word usage also diverged between partners as measured by Jensen-Shannon Divergence,which indicates that alignment was not achieved through lexical entrainment. A computationalmodel of dyadic communication is shown to account for the alien game results in terms of jointleast effort. Latent alignment increased, as measured by the mutual information between partnerreferent distributions, by 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.