Gazing at Social Interactions Between Foraging and Decision Theory.

Area of Science:

• Computational Neuroscience
• Cognitive Science
• Human-Computer Interaction

Background:

• Understanding social attention is crucial for designing interactions between humans and artificial agents.
• Gaze dynamics in humans provide insights into how we process socially relevant multimodal information.

Purpose of the Study:

• To computationally model the spatio-temporal dynamics of human gaze shifts during social attention.
• To frame social interaction as a foraging problem, balancing exploitation of local information with exploration of the environment.

Main Methods:

• Distilled social interactions into multimodal patches representing varying social value.
• Applied Langevin-type stochastic differential equations to describe gaze shift dynamics.
• Modeled value-based patch selection as a multi-alternative perceptual decision-making process using a race-to-threshold model.

Main Results:

• Demonstrated that gaze shift dynamics can be parsimoniously described by stochastic differential equations.
• Reduced complex value-based patch choice to a race-to-threshold decision-making mechanism.
• Showcased the effectiveness of perceptual evidence integrators in modeling patch selection.

Conclusions:

• The study provides a computational framework for understanding human social attention and gaze behavior.
• The foraging analogy offers a novel perspective on decision-making in dynamic social environments.
• Findings can inform the development of more natural and intuitive human-AI interactions.