Coordination dynamics of multi-agent interaction in a musical ensemble

Area of Science:

• Human social dynamics
• Dynamical systems analysis
• Acoustic signal processing

Background:

• Human interaction occurs across various timescales and social contexts.
• Coordination patterns differ between genuine interaction and mere co-existence.
• Local coordination dynamics can propagate, influencing global group behavior.

Purpose of the Study:

• To investigate multi-agent human interaction dynamics using Recurrence Quantification Analysis (RQA).
• To analyze group-level coordination patterns solely from audio data during a musical performance.
• To simulate the transition from uncoordinated to coordinated group behavior.

Main Methods:

• Applied Recurrence Quantification Analysis (RQA) to audio recordings of a wind orchestra performance.
• Analyzed group-level acoustic data, avoiding individual video or audio recordings.
• Examined changes in recurrence and stability measures during coordinated vs. uncoordinated musical sections.

Main Results:

• Recurrence and stability measures increased significantly when musicians performed as a coordinated group.
• Increased variability in recurrence measures indicated a greater range of explored behaviors in the interacting ensemble.
• The study provides baseline distributions for coordination patterns in orchestrated settings.

Conclusions:

• Audio-based RQA effectively captures group-level coordination dynamics in human interactions.
• Coordinated ensembles exhibit distinct dynamical signatures compared to non-interacting individuals.
• This research offers a framework for understanding emergent coordination through non-emergent examples.