Erev, I., Roth, A.: Predicting how people play games: Reinforcement learning in games with unique strategy equilibrium. American Economic Review 88 (1998) 848--881  Home/Search   Document Not in Database   Summary   Related Articles   Check

....automated systems with strategic interactions, adaptive, learning artificial agents can be used to discover e#ective strategies and to test proposed system designs. See also [48] for an application in strategy formation. And there is much literature on design of game playing agents, e.g. [24, 51, 25, 64] 5. Extraction Just as expert systems have long been used to encode information, neural nets have long been used to extract information from, and to obtain classifications of complex phenomena. Such applications continue, along with underlying innovations that make the use of neural nets more ....

Ido Erev and Alvin E. Roth. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. The American Economic Review, 88(4):848--881, 1998.

....may hard to be satis ed with these arguments and accept the MG as a model of a market. For this reason, we give Our reference framework is not Game Theory, but rather agent based models. Agents are adaptive and their learning dynamics shall be de ned in terms of reinforcements or attractions [20 23]. Then u i (t) should better be called perceived payo (specially because we are going to assume later that agents consider counter factually also the outcomes of strategies which they did not actually play) We shall borrow some terminology from game theory and speak simply of payo s. Also, ....

....approach. The interested reader is referred to Ref. 9] for more detailed account of the results. Agents learn from past experience which action a i (t) is the best one. The learning dynamics is the one used in general in minority games and it is well rooted in the economic literature [20 23]. The past experience of agent i is stored in the score i (t) i (t) 0 means that the action a i = 1 is (perceived as) more successful than a i = 1 and vice versa. Agents use the information accumulated in i (t) to take decisions : Probfa i (t) 1g p i (t) e (8) and ....

Erev I. and Roth A. E., (1998) Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. Am. Ec. Rev. 88, 848

....that the two coincide in a certain limit. Perhaps the greatest impulse to this direction of study was the widely cited 1995 paper of Roth and Erev [RE95] They proposed a multi agent reinforcement model based on Herrnstein s linear reinforement and response. Here and in subsequent publications [ER98, BE98], they show a good t with a wide range of empirical data. Limiting behavior in the basic model has recently been studied by Beggs [Beg02] and by Ianni [Ian02] In [SP00] both basic and discounted versions of Roth Erev learning are applied to social network formation. Individuals begin with ....

Erev, I. and Roth, A. (1998). Predicting how people play games: reinforcement learning in experimental games with unique mixed-strategy equilibria. American Economic Review 88, 848-881. 12

....our results to general stochastic games. 1 Introduction Reinforcement learning in the context of multi agent interaction has attracted the attention of researchers in cognitive psychology, experimental economics, machine learning, arti cial intelligence, and related elds for quite some time [8, 4]. Much of this work uses repeated games [3, 5] and stochastic games [10, 9, 7, 1] as models of such interactions. The literature on learning in games in game theory [5] is mainly concerned with the understanding of learning procedures that if adopted by the di erent agents will converge at end to ....

I. Erev and A.E. Roth. Predicting how people play games: Reinforcement learning in games with unique strategy equilibrium. American Economic Review, 88:848-881, 1998.

....results to general sum stochastic games. 1 Introduction Reinforcement learning in the context of multi agent interaction has attracted the attention of researchers in cognitive psychology, experimental economics, machine learning, arti cial intelligence, and related elds for quite some time [13, 6]. Much of this work uses repeated games [5, 8] and stochastic games [16, 15, 12, 2] as models of such interactions. The literature on learning in games in game theory [8] is mainly concerned with the understanding of learning procedures that if adopted by the di erent agents will converge at the ....

....into one of the following two paradigms: 1. The study of learning rules that will lead to a Nash equilibrium (or other solution concept) of a game [8] 12 2. The study of learning rules that will predict human behavior in non cooperative interactions, such as the ones modeled in repeated games [6]. While the approach taken in (2) has signi cant merit for descriptive purposes, a normative approach to learning should go beyond recommending behavior that will eventually lead to some desired solution. The major issues one needs to face are: 1. The learning algorithms of the agents should be ....

I. Erev and A.E. Roth. Predicting how people play games: Reinforcement learning in games with unique strategy equilibrium. American Economic Review, 88:848-881, 1998.

....choosing plausible subsets thereof as predicted outcomes. These difficulties with the standard theory of games have led to the development of a field known as cognitive game theory [4] in which models of learning play a central role in explaining and predicting strategic behaviour. Erev and Roth [17] show how simulations of agents equipped with a simple reinforcement learning algorithm can explain and predict the experimental data observed when human agents play a diverse range of trading games. Such multi agent reinforcement learning models form the basis of our solution concept for ....

....agents . Rather, we are attempting to predict how boundedly rational agents, who have no prior knowledge of an equilibrium solution nor the means to calculate one, might actually play against the mechanism we are (automatically) designing. For this reason, we chose to use the Roth Erev algorithm[17] , since it forms the basis of In other words Nash equilibrium strategies a cognitive model of how people actually behave in strategic environments. In particular it models two important principles of learning psychology: Thorndike s law of effect choices that have led to good outcomes ....

A. E. Roth and I. Erev. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):848--881, 1998.

....the Internet infrastructure. Furthermore, play is highly asynchronous, because agents adapt their strategies at different rates. New definitions of learning and new solution concepts that attempt to capture these aspects of Internetbased, repeated games have been proposed by, e.g. Erev and Roth [17] and Friedman and Shenker [24] Which, if any, of these solution concepts and game theoretic definitions of learning will play a central role in the emerging theory of DAMD is a wide open question. Moreover, we have yet to understand the relationship between the solution concept and the network ....

L. Erev and A. Roth, "Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria," American Economic Review 88 (1998), pages 848--881.

....provides an excellent discussion of the use and misuse of genetic algorithms, genetic programming, and other forms of evolutionary learning representations in the modeling of social processes. Additional types of learning algorithms that have been used include reinforcement learning algorithms [33, 88], Qlearning [92, 109] classifier systems [41] and various forms of learning algorithms that have been adapted for use in automated markets [38, 91] Many of these learning algorithms were originally developed with optimality objectives in mind, so caution must be used in applying them to social ....

Erev, I., and Roth, A. (1998). Predicting how people play games: Reinforcement learning in experimental games with unique mixed strategy equilibria. American Economic Review 8, 848--881.

....significant explanatory factors. 2 1. Introduction. In recent years, the field of learning in games has evolved considerably, with recent models able to make robust predictions in a variety of interesting games. Recent examples in the reinforcement learning arena include Roth and Erev (1995) Erev and Roth (1998), and Sarin and Vahid (1999) In belief learning, Fudenberg and Levine (1998) and Cheung and Friedman (1997, 1998) provide some comprehensive reviews. In hybrid models, Camerer and Ho (1997, 1998, 1999) have led the field. These models, though different in structure, are based on players ....

Erev, I. and A. Roth (1998), "Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique Mixed Strategy Equilibria," American Economic Review 88, 848-881.

....model with far fewer parameters. We will address this pertinent empirical question. We focus on the class of rule learning models of Stahl (1996, 1997a,b, 1999) hereafter S96, S97a,b, and S99) This is a rich class of learning models that encompasses action reinforcement (Roth and Erev, 1995; Erev and Roth, 1998), fictitious play (Brown, 1951) and belief updating (Mookherjee and Sopher, 1994; Camerer and Ho, 1997, 1999) 4 Briefly, a rule is a mapping from the game and history of play to a mixed strategy. For example, a noisy best response to the recent past is a Cournot like rule that describes much ....

Erev, I., Roth, A., 1998. Predicting how people play games: reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review 88, 848--881.

....#RL# model is a model of #rote learning, in which actions whichdowell in the past are more likely to be repeated in the future. Learning models in this spirit have a long history in biology and psychology. Their systematic application in experimental economics starts from Roth and Erev #1995#. Erev and Roth #1998# show that the RL model tracks the data well across a wide variety of experimental games with unique, mixed strategy equilibria. Note that the amount of information in the two experiments is exactly the same as that required by the RL model, which bases its predictions solely on the individual ....

.... cost sharing games we set R j #0# = A j #0# = # # j #0# = u j #0# = 200 for all players in RL, EWA and PA respectively, since the average #rst round payo#s was around 200 which also result in a probability predictions around the centroid, 6 This approach has been used in Roth and Erev #1995#, Erev and Roth #1998#, etc. Its applications and limitations are discussed in Erev and Haruvy #2000#. 7 See Haruvy and Stahl #2000# for a discussion of various criteria for gauging goodness of #t. 8 This yields a statistical accuracy of 1#. 9 #1=12; ###; 1=12#, for the #rst round. For the same reason, in the ....

Erev, Ido and Alvin Roth. #Predicting HowPeople Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria." American Economic Review 88 #September 1998#: 848-881.

No context found.

Erev, I., Roth, A.: Predicting how people play games: Reinforcement learning in games with unique strategy equilibrium. American Economic Review 88 (1998) 848--881

No context found.

I. Erev and A. E. Roth. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):84881, September 1998.

No context found.

I. Erev and A. E. Roth. Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria. American Economic Review, 88(4):848--81, September 1998.

No context found.

Ido Erev and Alvin E. Roth. Predicting how people play games: reinforcement leaning in experimental games with unique, mixed strategy equilibria. The American Economic Review, 88(4):848--881, September 1998.

No context found.

A. E. Roth and I. Erev. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):848-881, 1998.

No context found.

I. Erev and A.E. Roth. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. The American Economic Review, 88(4):848--881, September 1998.

No context found.

I. Erev and A. E. Roth. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):848--81, September 1998.

No context found.

I. Erev and A. E. Roth. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):848--881, 1998.

No context found.

A. E. Roth and I. Erev. Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. American Economic Review, 88(4):848-881, 1998.

No context found.

Erev, I., and Roth, A. (1998). Predicting how people play games: Reinforcement learning in experimental games with unique mixed strategy equilibria. American Economic Review 8, 848--881.

No context found.

Erev, I. and Roth, A., Predicting How People Play Games: Reinforcement Learning in Experimental Games With Unique Mixed Strategies American Economic Review, vol. 88, (1998), pp. 848-881.

No context found.

Erev, I. and A. Roth, (1998), Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria, Am. Econ. Review, 88, 848-881.

No context found.

Erev, Ido and Roth, Alvin E., \Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria," American Economic Review, 1998, 88(4), pp. 848-81.

No context found.

Erev, I. and A.E. Roth (1998), "Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria," American Economic Review, 88, pp. 848--881.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC