Human group coordination in a sensorimotor task with neuron-like decision-making

Abstract

the formation of cooperative groups of agents with limited information-processing capabilities to solve complex problems together is a fundamental building principle that cuts through multiple scales in biology from groups of cells to groups of humans. Here, we study an experimental paradigm where a group of humans is joined together to solve a common sensorimotor task that cannot be achieved by a single agent but relies on the cooperation of the group. in particular, each human acts as a neuron-like binary decision-maker that determines in each moment of time whether to be active or not. inspired by the population vector method for movement decoding, each neuron-like decisionmaker is assigned a preferred movement direction that the decision-maker is ignorant about. from the population vector reflecting the group activity, the movement of a cursor is determined, and the task for the group is to steer the cursor into a predefined target. As the preferred movement directions are unknown and players are not allowed to communicate, the group has to learn a control strategy on the fly from the shared visual feedback. Performance is analyzed by learning speed and accuracy, action synchronization, and group coherence. We study four different computational models of the observed behavior, including a perceptron model, a reinforcement learning model, a Bayesian inference model and a Thompson sampling model that efficiently approximates Bayes optimal behavior. The Bayes and especially the thompson model excel in predicting the human group behavior compared to the other models, suggesting that internal models are crucial for adaptive coordination. We discuss benefits and limitations of our paradigm regarding a better understanding of distributed information processing.

Experimental Evaluation of a Game-Theoretic Human Driver Steering Control Model

“…A high degree of agreement was seen between model output and measured player arm movements. Quite recently, Schmid and Braun [17] revisited their work on human sensorimotor interactions [16,[18][19] and remarked that human players in simple sensorimotor games acted in accordance with simulation results derived from noncooperative game theory. In separate studies, Chackochan and Sanguineti [20] developed a noncooperative-game-theoretic model to predict human players' arm forces in an arm reaching game, and validated it against experiments.…”

“…In (17), the AFS angle 2 () k  now depends on the driver angle determined at time step 1 k − . Such an amendment causes little change to 2 () k  , given that the sampling time T of the discrete system is sufficiently small.…”

Nash equilibria in human sensorimotor interactions explained by Q-learning with intrinsic costs

“…The role of reinforcement learning in the context of implicit learning has been previously examined in a number of studies on motor learning 68 – 70 , including physical robot-human interactions 71 , 72 and group coordination 73 , however, not in the context of game theory and haptic interactions. While our task was simple enough to allow for model-free reinforcement learning, a challenge for the future remains to study reinforcement learning in more complex environments with multiple agents.…”

Nash equilibria in human sensorimotor interactions explained by Q-Learning

“…Instead, we have focused on haptic interactions with forces that require implicit learning. The role of reinforcement learning in the context of implicit sensorimotor learning has also been previously examined [61][62][63][64] , including physical robot-human interactions 65,66 , however, not in the context of game theory and haptic interactions. Thus, our study adds to a growing body of research harnessing the power of reinforcement learning models to understand human interactions.…”