Expectation violations as an effective alternative to complex mentalizing in novel communication

Communicating without a shared language is often assumed to require cognitively demanding Theory of Mind reasoning. In this study, we tested whether a more parsimonious mechanism based on expectancy violations can achieve comparable performance. We tested this in the Tacit Communication Game, where we simulated Sender behavior using four computational models: the Surprise model based on expectancy violations and three levels of Theory of Mind. Human receivers interacted with each agent; we measured goal-identification accuracy and reaction times. The results show that Receivers paired with the Surprise model matched the accuracy of the most complex second level Theory of Mind agent and produced human-like message patterns. Models yielding higher accuracy also elicited faster responses, indicating lower cognitive load. These findings question the need for recursive mentalizing in unfamiliar interactions and highlight expectancy violation as a parsimonious route to efficient communication.