On aggregating information in actor networks

Abstract

This paper provides a way to think formally about the aggregation processes that take place in networks where individual actors (whether sensors, robots, or people) possess data whose value may decay over time. The various actors use data to make decisions: the larger the value, the better (i.e. more informed) the decision. At every moment, individual actors have the choice of making a decision or else to defer the decision to a later time. However, the longer they wait, the lower the value of the data they hold. To counter-balance the effect of time discounting, we define an algebraic operation that we call aggregation, whereby two or more actors integrate their data in the hope of increasing its value. Our main contribution is a formal look at the value of time-discounted information and at the algebra of its aggregation. We allow aggregation of time-discounted information to proceed in an arbitrary, not necessarily pairwise, manner. Our model relates aggregation decisions to the ensuing value of information and suggests natural thresholding strategies for the aggregation of the information collected by sets of network actors. Extensive simulations have confirmed the accuracy of our theoretical predictions.