Conditional dissociation, i.e. the option to leave an interacting partner in response to his behaviour, is a mechanism that has been shown to promote cooperation in several settings, but the fundamental features that make conditional dissociation work in this way are not yet fully understood. This paper identifies some of the key conditions that make conditional dissociation lead to high levels of cooperation, explains how this mechanism can support the evolutionary coexistence of cooperative and non-cooperative behaviour typically observed in nature, and provides an analytical formula to estimate the expected degree of cooperation thus achieved. Our model involves a population of individuals who are paired to play an iterated prisoner's dilemma. All individuals share the same capacity to react to the action previously chosen by their partner and, without any other a priori constraint or exclusion, they may use any behavioural rule that is compatible with this capacity. The dynamic evolution of the population eventually enters either a non-cooperative or a partially cooperative regime, depending mainly on the expected lifetime of individuals. Whenever the partially cooperative regime materializes, the cornerstone of its long-run stability is the coexistence of defectors and "Out-for-Tat" strategists, the latter being those who start cooperating and respond to defection by merely leaving. We find, therefore, that conditional dissociation is the essential disciplinary device supporting cooperation, whilst other conditional strategies (such as Tit-for-Tat) remain present only in small population shares. These conclusions are obtained both by extensive numerical simulations and through analytical mean-field methods that approximate the stochastic simulation dynamics and deliver accurate predictions for general parameter configurations.

No abstract

We study population game dynamics under which each revising agent tests each of his strategies a fixed number of times, with each play of each strategy being against a newly drawn opponent, and chooses the strategy whose total payoff was highest. In the centipede game, these best experienced payoff dynamics lead to cooperative play. When strategies are tested once, play at the almost globally stable state is concentrated on the last few nodes of the game, with the proportions of agents playing each strategy being largely independent of the length of the game. Testing strategies many times leads to cyclical play.

The market effects of quality variability and uncertainty have classically been studied in the particular context of asymmetric information, focusing on the sellers' expected behavior and the phenomenon of adverse selection. Looking instead at the consumers' expected behavior, this paper uses an agent-based model to illustrate how quality uncertainty by itself can lead to market failure, even in the absence of asymmetric information. Assuming that buyers estimate the quality of the product they buy on the basis of their experience from previous purchases, and considering quality estimation rules which are individually sensible and unbiased, this paper shows that market interaction with quality uncertainty generally produces underestimation of product quality as well as systematic drops in prices and losses of market efficiency. This study also shows that the spread of information through social networks can greatly mitigate this market failure.

We study a family of population game dynamics under which each revising agent randomly selects a set of strategies according to a given test-set rule; tests each strategy in this set a fixed number of times, with each play of each strategy being against a newly drawn opponent; and chooses the strategy whose total payoff was highest, breaking ties according to a given tie-breaking rule. These dynamics need not respect dominance and related properties except as the number of trials become large. Strict Nash equilibria are rest points but need not be stable. We provide a variety of sufficient conditions for stability and for instability, and illustrate their use through a range of applications from the literature. JEL classification numbers: C72, C73.