Coordination problems and the role of institutions: multi-agent simulations with learning

The paper develops an agent-based coordination game in which agents seek, try out and maintain coordination in the presence of positive search and coordination costs. Agents are perfectly rational in the sense that they optimise their individual payoff through coordination. They are moreover perfectly loyal, since they do not abandon coordination unless it becomes collectively impossible to maintain it and they share equally both the net advantage of coordination and the net loss of failure. Our findings appear to confirm that in context of ontological uncertainty, learning and coordination costs, cooperation often fails in terms of process and final outcome. The absence of opportunism thus do not ensure that agents will cooperate extensively. Moreover learning on one hand optimises search costs and makes the artificial world of the experiment more realistic. On the other hand, it leads agents to more prudent and conservative behaviour. Agents’ loyalty makes superfluous regulatory institutions preventing individual opportunism, but it gives them significant scope for intervention in context of coordination failure. As players involved in the game, rather than arbitrators or regulators, catalyst institutions accelerate coordination and induce cooperation equilibria higher than those single agents autonomously are able to reach.