Computational joint action: Dynamical models to understand the development of joint coordination

Area of Science:

• Computational Neuroscience
• Game Theory
• Human-Computer Interaction

Background:

• Human dyads develop coordination strategies, often resembling Nash equilibria in sensorimotor games.
• Uncertainty about a partner's actions leads to robust coordination strategies, suggesting probabilistic partner models.
• The mechanisms driving the development of joint coordination over repeated trials remain largely unknown.

Purpose of the Study:

• To present a general computational model for understanding the mechanisms of joint coordination development over repeated trials.
• To simulate interactive behaviors and analyze experimental data to quantify individual behaviors in joint actions.
• To explore how different partner representations influence coordination outcomes and identify factors affecting coordination development.

Main Methods:

• Modeling joint tasks as quadratic games with quadratic cost functions for each participant.
• Employing Bayesian estimation for participants to predict partner actions by combining predictions and sensory observations.
• Utilizing stochastic optimization of expected cost, given a partner model, for action selection.

Main Results:

• The model predicts different Nash equilibria based on varying partner representations in a Stag Hunt game.
• In a joint two via-point (2-VP) reaching task, the model accurately captured the temporal evolution of performance.
• Model parameters estimated from experimental data provided a detailed characterization of individual dyad participants.

Conclusions:

• Computational models offer insights into factors that facilitate or hinder joint coordination development.
• These models can be applied in clinical settings to interpret behaviors in individuals with impaired interaction capabilities.
• The research provides a theoretical foundation for developing artificial agents that enhance neuromotor recovery through coordination.