Class Of Coordination Games Research Articles

Evolutionary dynamics in bilingual games

In two-strategy coordination games with distinct payoff- and risk-dominant equilibria, existing results show that the inefficient risk-dominant equilibrium is uniquely selected under many evolutionary dynamics. In the above class of coordination games, we study the effect of introducing a bilingual strategy that is compatible with both of the existing strategies. An agent playing the bilingual strategy incurs an additional adoption cost but never miscoordinates with any other agent. We show that if the adoption cost of the bilingual strategy is low, then the efficient payoff-dominant equilibrium can be uniquely selected under many evolutionary dynamics.

Restricted and free-form cheap-talk and the scope for efficient coordination

This paper is the first to compare the efficiency-enhancing capacity of two communication protocols in experimental stag-hunt games. The traditional restricted protocol that allows communicating intentions only fails to improve efficiency over the no-communication conditions. When players are allowed to send any messages (free-form), the majority of them send messages that underscore a reason for which players should choose the efficient action. To explore further whether the absence of such a richer message in the traditional restricted communication protocol can be the cause of its ineffectiveness, we ran an additional treatment that included the intention-based message as well as a reason-based message. The data show that the richer restricted communication is as effective as free-form communication, and that reason-based messages are an effective efficiency-enhancing device in this class of coordination games regardless of whether the protocol is restricted or free-form.

Full disclosure in decentralized organizations

We consider a class of coordination games with asymmetric information. Agents have different ideal actions and disclose private information strategically. Complete information is the Pareto dominant information structure. We provide conditions on information certifiability for the complete information outcome to be implementable under private and public communication.

Absence of pure Nash equilibria in a class of co-ordination games

We prove that for a certain class of co-ordination games with increasing payoffs for co-ordination there are not necessarily pure Nash equilibria. This is achieved by introducing a particular model for the payoffs. This result is in contrast to the congestion game with decreasing payoff. We discuss a generalization of the model introduced which is of independent interest.

Absence of pure Nash equilibria in a class of co-ordination games

We prove that for a certain class of co-ordination games with increasing payoffs for co-ordination there are not necessarily pure Nash equilibria. This is achieved by introducing a particular model for the payoffs. This result is in contrast to the congestion game with decreasing payoff. We discuss a generalization of the model introduced which is of independent interest.

Generalized multiagent learning with performance bound

We present new Multiagent learning (MAL) algorithms with the general philosophy of policy convergence against some classes of opponents but otherwise ensuring high payoffs. We consider a 3-class breakdown of opponent types: (eventually) stationary, self-play and "other" (see Definition 4) agents. We start with ReDVaLeR that can satisfy policy convergence against the first two types and no-regret against the third, but it needs to know the type of the opponents. This serves as a baseline to delineate the difficulty of achieving these goals. We show that a simple modification on ReDVaLeR yields a new algorithm, RV ?(t), that achieves no-regret payoffs in all games, and convergence to Nash equilibria in self-play (and to best response against eventually stationary opponents--a corollary of no-regret) simultaneously, without knowing the opponent types, but in a smaller class of games than ReDVaLeR . RV ?(t) effectively ensures the performance of a learner during the process of learning, as opposed to the performance of a learned behavior. We show that the expression for regret of RV ?(t) can have a slightly better form than those of other comparable algorithms like GIGA and GIGA-WoLF though, contrastingly, our analysis is in continuous time. Moreover, experiments show that RV ?(t) can converge to an equilibrium in some cases where GIGA, GIGA-WoLF would fail, and to better equilibria where GIGA, GIGA-WoLF converge to undesirable equilibria (coordination games). This important class of coordination games also highlights the key desirability of policy convergence as a criterion for MAL in self-play instead of high average payoffs. To our knowledge, this is also the first successful (guaranteed) attempt at policy convergence of a no-regret algorithm in the Shapley game.

Equilibrium selection in coordination games with simultaneous play

We apply the dynamic stochastic framework proposed in recent evolutionary literature to a class of coordination games played simultaneously by the entire population. In these games payoffs, and hence best replies, are determined by a summary statistic of the population strategy profile. We demonstrate that with simultaneous play, the equilibrium selection depends crucially on how best responses to the summary statistic remain piece-wise constant. In fact, all the strict Nash equilibria in the underlying stage game can be made stochastically stable depending on how the best response mapping generates piece-wise constant best responses.