Information integration in division of labor: Validation of earlier findings

Abstract

The present study focuses on evaluation of the underlying dynamics in two-agent, single-target pursuit within the context of division of labor paradigm. Specifically, it aims at clarifying which of the two tracking strategies viz. equipping agents with switching complementary roles versus designated specialization may result in increased information integration between agents. Although our previous findings on this topic hinted at significantly higher benefit of the first strategy, they suffered from the assumptions of linearity and independence of agents, imposed by our parametric Gaussian formulation of information integration. Here we address these shortcomings through adaptation of non-parametric multivariate information integration formalism. We show (1) that with this new formulation our previous results still hold,(2) that they also remain intact if target is taken into account for quantification of agents’ information integration and (3) that the agents’ relation with respect to the tracking task can be explained in terms of potential colinearity of their actions. We further discuss some of potential issues that require careful consideration while studying such dynamics and conclude by highlighting recent advances that can potentially help overcome such pitfalls.

Publication
Proceedings of the Joint Symposium of AROB-ISBC-SWARM 2022