Learning by imitation

Abstract

Cataloged from PDF version of article.This paper introduces a learning algorithm that allows for imitation in recursive dynamic games. The Kiyotaki-Wright model of money is a well-known example of such decision environments. In this context, learning by experience has been studied before. Here, we introduce imitation as an additional channel for learning. In numerical simulations, we observe that the presence of imitation either speeds up social convergence to the theoretical Markov-Nash equilibrium or leads every agent of the same type to the same mode of suboptimal behavior. We observe an increase in the probability of convergence to equilibrium, as the incentives for optimal play become more pronounced

Cryptocurrencies, Network Effects, and Switching Costs

“…all) parameterizations. 12 Exploring parameterizations similar to those found by Duffy (2001), where agents would do best by employing a nonsalient trading strategy, Başçi (1999) finds that allowing agents to learn by imitating the successful strategies of other agents (in addition to their own experience) increases the degree of equilibrium-consistent behavior.…”

Learning to choose a commodity-money: Carl Menger's theory of imitation and the search monetary framework

“…In particular, agents do not play strategies implying the use of a medium of exchange with the greatest storage cost, even if it is the best response for a given set of parameters of Andrés Alvarez the underlying game. Nevertheless, and this is a more interesting result, in models where agents are allowed to imitate others' strategies (like in Başçı 1999 andDuffy 2001), those equilibria may be achieved. These works show that imitation increases the speed and the likelihood of convergence towards monetary equilibria.…”

“…In the first kind of models, agents are represented as computational artificial agents following simple algorithms in order to choose the most suitable exchange strategies. All models are based on classifier systems 12 and on some genetic algorithm implementations (Marimon et al 1990, Başçı 1999, Duffy 2001and Duffy and Ochs 1999. In these models, the main results are illustrated with computer simulations.…”

“…(Menger 1976(Menger [1871 Nevertheless, imitation in ACE literature is introduced in an arbitrary manner. In fact, in Başçı (1999) agents imitate other's strategies after a random matching between two agents of the same population type, who imitate the other's strategy in a completely random way. This is interpreted like a subconscious process (Başçı 1999(Başçı : 1577, while Menger's statement shows that imitation is a more conscious action which supposes a minimal rationality of imitation, namely, to imitate successful exchange strategies.…”

“…In fact, in Başçı (1999) agents imitate other's strategies after a random matching between two agents of the same population type, who imitate the other's strategy in a completely random way. This is interpreted like a subconscious process (Başçı 1999(Başçı : 1577, while Menger's statement shows that imitation is a more conscious action which supposes a minimal rationality of imitation, namely, to imitate successful exchange strategies. This last characteristic is assumed in Wright (1995) where the evolutionary process is based on the so-called replicator dynamics.…”