Information parsimony in collaborative interaction

Abstract

We investigate the information processing cost associated with performing a collaborative dyadic task at a specific utility level. We build our approach on the Relevant Information formalism, which combines Shannon's Information Theory and Markov Decision Processes, for modelling a dyadic interaction scenario in which two agents with independent controllers move an object together with fully redundant control. Results show that increasing dyad's collaboration decreases the information intake and vice versa, antagonistic behavior puts a strain on the information bandwidth capacity. The key role of the particular embodiment of the environment in this trade-off is demonstrated in a series of simulations with informationally parsimonious optimal controllers.