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

Previous joint action studies using sensorimotor games suggest that human dyads develop coordination strategies which can be interpreted as Nash equilibria. In a previous study, we argued that if players are uncertain about what their partner is doing, they develop a coordination strategy which is more robust to the actual partner actions. This suggested that humans maintain an explicit representation of what the partner will be doing - a partner model - which also accounts for their degree of confidence about it. However, the mechanisms underlying the development of a joint coordination over repeated trials remain unknown. Very much like individual sensorimotor control, dynamical models can be used to understand how joint coordination develops. Here we present a general computational model - based on game theory and Bayesian estimation - to understand the mechanisms underlying the development of a joint coordination. A joint task is modeled as a quadratic game. Each player predicts their partner's next move (partner model) by optimally combining predictions and sensory observations, and selects their actions through a stochastic optimization of its expected cost, given the partner model. We show that the model captures well the temporal evolution of performance in a previous joint action experiment, and the estimated parameters provide a comprehensive characterization of individual participants in a dyad.