On Distributed Cooperative Decision-Making in Multiarmed Bandits

Cooperative decision-making under uncertainty is ubiquitous in natural systems as well as in engineering networks. Typically in a distributed cooperative decision-making scenario, there is assimilation of information across a network followed by decision-making based on the collective information. The result is a kind of collective intelligence, which is of fundamental interest both in terms of understanding natural systems and designing efficient engineered systems. A fundamental feature of decision-making under uncertainty is the explore-exploit tradeoff. The explore-exploit tradeoff refers to the tension between learning and optimizing: the decision-making agent needs to learn the unknown system parameters (exploration), while maximizing its decision-making objective, which depends on the unknown parameters (exploitation).