Turn-Taking Mechanisms in Imitative Interaction: Robotic Social Interaction Based on the Free Energy Principle

Area of Science:

• Robotics
• Computational Neuroscience
• Artificial Intelligence

Background:

• Leader-follower dynamics and turn-taking are crucial for social interaction.
• Prior work established a meta-prior parameter (w) to define initial leader/follower roles in robotic interactions based on the free energy principle.
• The free energy principle offers a unified framework for perception and action, incorporating concepts like sensory attenuation.

Purpose of the Study:

• To investigate if the leader-follower relationship and turn-taking patterns can dynamically shift during dyadic imitative interactions.
• To explore the influence of the meta-prior parameter (w) on emergent coordination behaviors in robots.
• To analyze information flow between agents during different coordination regimes.

Main Methods:

• Conducting extensive robotic simulation experiments based on the free energy principle.
• Systematically varying the meta-prior parameter (w) for both agents during interaction phases.
• Utilizing transfer entropy analysis to quantify information flow between interacting agents.

Main Results:

• Identified three distinct types of behavioral coordination based on the phase space of the meta-prior parameter (w).
• Observed distinct coordination patterns: mutual intention ignoring (high ws), leader-follower (mixed ws), and spontaneous turn-taking (low/intermediate ws).
• Demonstrated dynamic role switching and turn-taking when ws oscillated anti-phase, accompanied by shifts in information flow direction.

Conclusions:

• The meta-prior parameter (w) is a critical factor in dynamically shaping leader-follower relationships and turn-taking behaviors in dyadic robotic interactions.
• Robotic simulations can replicate complex social coordination phenomena, offering insights into biological systems.
• Understanding these dynamics provides a foundation for developing more sophisticated human-robot collaboration and social robotics.