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

Area of Science:

• Cognitive Science
• Computational Linguistics
• Human-Computer Interaction

Background:

• Shared language is typically assumed for effective communication.
• Complex Theory of Mind (ToM) reasoning is often considered necessary for communication without a shared language.
• Alternative, more parsimonious mechanisms for non-linguistic communication are underexplored.

Purpose of the Study:

• To investigate whether expectancy violations can facilitate communication comparable to Theory of Mind (ToM).
• To compare the performance of a computational model based on expectancy violations against different levels of ToM.
• To assess the efficiency and cognitive load associated with different communication mechanisms.

Main Methods:

• Development of four computational Sender models: one based on expectancy violations (Surprise model) and three based on different levels of ToM.
• Human Receivers interacted with each simulated Sender in the Tacit Communication Game.
• Measurement of goal-identification accuracy and reaction times to assess Receiver performance and cognitive load.

Main Results:

• The Surprise model achieved accuracy comparable to a high-level ToM agent.
• The Surprise model generated message patterns similar to human communication.
• Higher accuracy in communication models correlated with faster Receiver responses, suggesting reduced cognitive load.

Conclusions:

• Expectancy violations offer a parsimonious and effective mechanism for communication in unfamiliar contexts.
• The findings challenge the necessity of recursive mentalizing for efficient non-linguistic interactions.
• Expectancy violation presents a viable alternative route to successful tacit communication.