Imitation and efficient contagion

Abstract

In this paper we study the conditions under which efficient behavior can spread from a finite initial seed group to an infinite population living on a network. We formulate conditions on payoffs and network structure under which overall contagion occurs in arbitrary regular networks. Central in this process is the communication pattern among players who are confronted with the same decision, i.e. who are at the same distance from the initial seed group. The extent to which these agents interact among themselves (rather than with players who already have faced or subsequently will face the decision problem) is critical in the Prisoner's Dilemma. In the Coordination Game the key element is the cohesion of the efficient cluster, a property which is different from the one identified in the Prisoner's Dilemma. Additional results are obtained when we distinguish the interaction and information neighborhoods. Specifically, we find that contagion tends to be favored by fast neighborhood growth if an assumption of conservative behavior is made. We discuss our findings in relation to the notions of clustering, transitivity and cohesion.

Stability and cooperative solution in stochastic games

“…Nicolas Jonard and Tristan Boyer (IPAG Business School, Paris) 62 examine the conditions needed to ensure the efficient diffusion of cooperation in an infinite population of networked individuals via imitative learning. The more the agents interact among themselves rather than with players, who are closer to or further away from the initial seed group, the easier it is for efficient contagion to take place.…”

Collaboration in networks with randomly chosen agents

“…Starting with the seminal papers of Blume (1993Blume ( , 1995 and Ellison (1993), much of the literature is concerned with local interaction games. It is assumed that agents are located in a fixed network and interact with their direct neighbors (Alós-Ferrer and Weidenholzer, 2007Weidenholzer, , 2008Weidenholzer, , 2014Boyer and Jonard, 2014;Cui, 2014;Ellison, 2000;Eshel, Samuelson andShaked, 1998 andMorris, 2000). Two prominent dynamic adjustment rules are used in these models: myopic best-response reply and imitation (for an overview, see Weidenholzer, 2010).…”