We use simple learning models to track the behavior observed in experiments concerning three extensive form games with similar perfect equilibria. In only two of the games does observed behavior approach the perfect equilibrium as players gain experience. We examine a family of learning models which possess some of the robust properties of learning noted in the psychology literature. The intermediate term predictions of these models track well the observed behavior in all three games, even though the models considered differ in their very long term predictions. We argue that for predicting observed behavior the intermediate term predictions of dynamic learning models may be even more important than their asymptotic properties.

Abstract not found

This paper therefore builds on [3], in which a model of consistent aspirations-based learning was introduced

Most learning models assume players are adaptive (i.e., they respond only to their own previous experience and ignore others ’ payoff information) and behavior is not sensitive to the way in which players are matched. Empirical evidence suggests otherwise. In this paper, we extend our adaptive experience-weighted attraction (EWA) learning model to capture sophisticated learning and strategic teaching in repeated games. The generalized model assumes there is a mixture of adaptive learners and sophisticated players. An adaptive learner adjusts his behavior the EWA way. A sophisticated player rationally best-responds to her forecasts of all other behaviors. A sophisticated player can be either myopic or farsighted. A farsighted player develops multiple-period rather than single-period forecasts of others ’ behaviors and chooses to ‘‘teach’ ’ the other players by choosing a strategy scenario that gives her the highest discounted net present value. We estimate the model using data from p-beauty contests and repeated trust games with incomplete information. The generalized model is better than the adaptive EWA model in describing and predicting behavior. Including teaching also allows an empirical

Most learning models assume players are adaptive (i.e., they respond only to their own previous experience and ignore others ' payo ® information) and behavior is not sensitive to the way in which players are matched. Empirical evidence suggests otherwise. In this paper, we extend our adaptive experienceweighted attraction (EWA) learning model to capture sophisticated learning and strategic teaching in repeated games. The generalized model assumes there is a mixture of adaptive learners and sophisticated players. An adaptive learner adjusts his behavior the EWA way. A sophisticated player rationally best-responds to her forecasts of all other behaviors. A sophisticated player can be either myopic or farsighted. A farsighted player develops multiple-period rather than single-period forecasts of others ' behaviors and chooses to `teach ' the other players by choosing a strategy scenario that gives her the highest discounted net present value. We estimate the model using data from p-beauty contests and repeated trust games with incomplete information. The generalized model is better than the

This paper explores ramifications of quasirational behavior in capital markets. Despite the academic literature asserting that financial markets are efficient and humans are rational agents, there is a widespread discontentment among practitioners and a growing body of researchers that neither of those premises is valid. The increasing amount of empirical evidence against perfectly efficient capital market processes has been examined in great depth. However, academic research is just beginning to explore the causes of that imperfect rationality. A diverse group of fields has taken a keen interest in how humans make decisions. The foundation of this paper uses methods drawn from such disparate disciplines as cognitive psychology and philosophy, biology, genetics, economics, computer science, and game theory. Human decision processes in making investment decisions have a fundamental impact on capital market processes. Thus, many of the previously unexplained phenomena of the capital mar...

Abstract. In 1987 the first published research appeared which used the Genetic Algorithm as a means of seeking better strategies in playing the repeated Prisoner’s Dilemma. Since then the application of Genetic Algorithms to game-theoretical models has been used in many ways. To seek better strategies in historical oligopolistic interactions, to model economic learning, and to explore the support of cooperation in repeated interactions. This brief survey summarises related work and publications over the past thirteen years. It includes discussions of the use of gameplaying automata, co-evolution of strategies, adaptive learning, a comparison of evolutionary game theory and the Genetic Algorithm, the incorporation of historical data into evolutionary simulations, and the problems of economic simulations using real-world data. 1

We studied asymmetric competition between two (three-person) groups. Each group member received an initial endowment and had to decide whether or not to contribute it. The group with more contributions won the competition and each of its members received a reward. The members of the losing group received nothing. The asymmetry was created by randomly and publicly selecting one group beforehand to be the winning group in the case of a tie. A theoretical analysis of this asymmetric game generates two qualitatively different solutions, one in which members of the group that wins in the case of a tie are somewhat more likely to contribute than members of the group that loses, and another in which members of the group that loses in the case of a tie are much more likely to contribute than members of the group that wins. The experimental results are clearly in line with the first solution. Copyright # 2005 John Wiley & Sons, Ltd. key words intergroup competition; participation game; asymmetric game 1.

While many learning models have been proposed in the game theoretic literature to track individuals ’ behavior, surprisingly little research has focused on how well these models describe human adaptation in changing dynamic environments. This paper evaluates several learning models in light of a laboratory experiment on responsiveness in a low-information dynamic game subject to changes in its underlying structure. While history-dependent reinforcement learning models track convergence of play well in repeated games, it is shown that they are ill suited to dynamic environments, in which sastisficing models accurately predict behavior. A further objective is to determine which heuristics, or “rules of thumb, ” when incorporated into learning models, are responsible for accurately capturing responsiveness. Reference points and a particular type of experimentation are found to be important in both describing and predicting play. Implications for the design of learning models for dynamic, low-information settings such as the Internet are discussed.

Abstract not found