The emergence of distributed multiagent systems (DMAS) for decision making in biological research studies creates stimulating inquiries. How should the dynamic properties originating from the interactions of hundreds of agents be used to model and forecast their behavior? How should the assignment of roles in DMAS that rely on local information and use modest to no communication be orchestrated?

Campbell and Wu put forward three procedures to be used by agents to achieve and retain allocation of roles in DMAS. In the first procedure, each agent takes into account its role and records the roles of its neighbors, prior to switching to the nearest role or applying a uniform probability to assume one of multiple nearby neighboring roles. In the second procedure, the agent does not consider its role to derive a new role, and retains its role when the impact of a role exchange decision is minimized in its neighborhood, in order to help minimize unnecessary updates of roles. In the third procedure, used for benchmark testing, each agent randomly selects its role with an identical probability, without communicating with other agents. To investigate the effects of synchronization in a DMAS, the authors present equations for computing various metrics: the projected proportion of agents that retain and renew their roles at a given time; the likelihood of an agent observing the role with the lowest count of neighboring agents; the probability of the permutation of the neighboring roles with the lowest count taking place, given the existing circulation of roles; the calibrator for averting excessive enumeration of agents that alter roles; and the allocation of new roles to probable agents that preserve their roles.

Campbell and Wu conduct special mock-up experiments to investigate the consequences of harmonization on the dynamics of role updates in DMAS. The results reveal large variations in the equal distribution of roles with high synchronization, due to agents over-remunerating roles with lesser attendance. The techniques agents use to update roles based on local information from neighbors impinge on the dynamics of DMAS; the technique that underplays avoidable role updates exploits the information better, and the range of attendance values is about the same for the decisions of agents at all time steps; moreover, any DMAS that uses this technique will be steadier, because agents are less likely to change roles, as roles appear.

The authors shed light on both the linear and curvilinear associations linking the synchronization of distributed decision-making procedures to the dynamics of DMAS. The concepts of DMAS presented in the paper are valuable for modeling road congestion as inadvertent emergent properties of human behavior and interactions in traffic networks [1].