This paper studies the population dynamics of preference traits in a model of intergenerational cultural transmission. Parents socialize and transmit their preferences to their offspring, motivated by a form of paternalistic altruism (``imperfect empathy''). In such a setting we study the long run stationary state pattern of preferences in the population, according to various socialization mechanisms and institutions, and identify sufficient conditions for the global stability of an heterogenous stationary distribution of the preference traits. We show that cultural transmission mechanisms have very different implications than evolutionary selection mechanisms with respect to the dynamics of the distribution of the traits in the population, and we study mechanisms which interact evolutionary selection and cultural transmission. Journal of Economic Literature

Learning to act in a multiagent environment is a difficult problem since the normal definition of an optimal policy no longer applies. The optimal policy at any moment depends on the policies of the other agents and so creates a situation of learning a moving target. Previous learning algorithms have one of two shortcomings depending on their approach. They either converge to a policy that may not be optimal against the specific opponents' policies, or they may not converge at all. In this article we examine this learning problem in the framework of stochastic games. We look at a number of previous learning algorithms showing how they fail at one of the above criteria. We then contribute a new reinforcement learning technique using a variable learning rate to overcome these shortcomings. Specifically, we introduce the WoLF principle, "Win or Learn Fast", for varying the learning rate. We examine this technique theoretically, proving convergence in self-play on a restricted class of iterated matrix games. We also present empirical results on a variety of more general stochastic games, in situations of self-play and otherwise, demonstrating the wide applicability of this method.

this article, please send e-mail to: tpami@computer.org, and reference IEEECS Log Number 108453

This paper presents an overview of multi-agent system models of land-use/cover change (MAS/LUCC models). This special class of LUCC models combines a cellular landscape model with agent-based representations of decisionmaking, integrating the two components through specification of interdependencies and feedbacks between agents and their environment. The authors review alternative LUCC modeling techniques and discuss the ways in which MAS/LUCC models may overcome some important limitations of existing techniques. We briefly review ongoing MAS/LUCC modeling efforts in four research areas. We discuss the potential strengths of MAS/LUCC models and suggest that these strengths guide researchers in assessing the appropriate choice of model for their particular research question. We find that MAS/LUCC models are particularly well suited for representing complex spatial interactions under heterogeneous conditions and for modeling decentralized, autonomous decision making. We discuss a range of possible roles for MAS/LUCC models, from abstract models designed to derive stylized hypotheses to empirically detailed simulation models appropriate for scenario and policy analysis. We also discuss the challenge of validation and verification for MAS/LUCC models. Finally, we outline important challenges and open research questions in this new field. We conclude that, while significant challenges exist, these models offer a promising new tool for researchers whose goal is to create fine-scale models of LUCC phenomena that focus on human-environment interactions.

Game theory is one of the key paradigms behind many scientific disciplines from biology to behavioral sciences to economics. In its evolutionary form and especially when the interacting agents are linked in a specific social network the underlying solution concepts and methods are very similar to those applied in non-equilibrium statistical physics. This review gives a tutorial-type overview of the field for physicists. The first four sections introduce the necessary background in classical and evolutionary game theory from the basic definitions to the most important results. The fifth section surveys the topological complications implied by non-mean-field-type social network structures in general. The next three sections discuss in detail the dynamic behavior of three prominent classes of models: the Prisoner’s Dilemma, the Rock–Scissors–Paper game, and Competing Associations. The major theme of the review is in what sense and how the graph structure of interactions can modify and enrich the picture of long term behavioral patterns emerging in evolutionary games.

The purpose of this chapter is to describe a menu of experimental games that are useful for measuring aspects of social norms and social preferences. Economists use the term “preferences ” to refer to the choices people make, and particularly to tradeoffs between different collections (“bundles”) of things they value—food, money, time, prestige, and

Abstract not found

In the field of economics, perhaps the most important break with the past—one that leaves open huge areas for future work—lies in the economics of information. It is now recognized that information is imperfect, obtaining informa-tion can be costly, there are important asymmetries of information, and the extent of information asymmetries is affected by actions of firms and individuals. This recognition deeply affects the understanding of wisdom inherited from the past, such as the fundamental welfare theorem and some of the basic characterization of a market economy, and provides explanations of economic and social phenomena that otherwise would be hard to understand. I.

We consider a dynamic social network model in which agents play repeated games in pairings determined by a stochastically evolving social network. Individual agents begin to interact at random, with the interactions modeled as games. The game payoffs determine which interactions are reinforced, and the network structure emerges as a consequence of the dynamics of the agents' learning behavior. We study this in a variety of game-theoretic conditions and show that the behavior is complex and sometimes dissimilar to behavior in the absence of structural dynamics. We argue that modeling network structure as dynamic increases realism without rendering the problem of analysis intractable. P airs from among a population of 10 individuals interact repeatedly. Perhaps they are cooperating to hunt stags and rabbits, or coordinating on which concert to attend together; perhaps they are involved in the somewhat more antagonistic situation of bargaining to split a fixed payoff, or attempting to escape the undesirable but compelling equilibrium of a Prisoner's Dilemma. As time progresses, the players adapt their strategies, perhaps incorporating randomness in their decision rules, to suit their environment. But they may also exert control over their environment. The players may have choice over the pairings but not perfect information about the other players. They may improve their lot in two different ways. A child who is being bullied learns either to fight better or to run away. Similarly, a player who obtains unsatisfactory results may choose either to change strategies or to change associates. Regardless of whether the interactions are mostly cooperative or mostly antagonistic, it is natural and desirable to allow evolution of the social network (the propensity for each pair to interact) as well as the individuals' strategies. We build a model that incorporates both of these modes of evolution. The idea is simple. (*) Individual agents begin to interact at random. The interactions are modeled as games. The game payoffs determine which interactions are reinforced, and the social network structure emerges as a consequence of the dynamics of the agents' learning behavior. As the details of the specific game and the reinforcement dynamics vary, we then obtain a class of models. In this paper, we treat some simple reinforcement dynamics, which may serve as a base for future investigation. The idea of simultaneous evolution of strategy and social network appears to be almost completely unexplored. Indeed, the most thoroughly studied models of evolutionary game theory assume mean-field interactions, where each individual is always equally likely to interact with each other. Standard treatments of evolutionary game dynamics 1 2 operate entirely in this paradigm. This is due, to a large extent, to considerations of theoretical tractability of the model. Models have been introduced that allow the agents some control over their choice of partner (3), but the control is still exerted in a mean-field setting: one chooses between the present partner and a new pick at random from the whole population. Evolutionary biologists know that evolutionary dynamics can be affected by nonrandom encounters or population structure, as in Sewall Wright's models of assortative mating (4). Wright (5) already realized that positive correlation of encounters could provide an account of evolution of altruism. Thus, the need for social network models has been long recognized. When the social network is modeled, it is almost always static. § Interactions, for example, may be posited to occur only between players whose locations are close, according to some given spatial data. Biological models in which encounters are governed by spatial structure have become increasingly frequent in the 1990s; see, for example, the work of Durrett, Levin, and Neuhauser (7-9). A similar hypothesis of spatial structure, in a game theory context, arises in ref. 10. Here, technology from statistical mechanics is adapted to the analysis of games whose interactions take place between neighbors in a grid. A number of recent investigations by game theorists, some directly inspired by biological models, have shown that the dynamics of strategic interaction can be strikingly different if interaction is governed by some spatial structure, or more generally, some graph structure (11-13). For instance, one-shot Prisoner's Dilemma games played with neighbors on a circle or torus allows cooperation to evolve in a way that the random encounter model does not. The spatial or graph structure can be important to determine which equilibria are possible, whether repeated interactions can be expected to converge to equilibrium, and, if so, how quickly convergence takes place (14). Because the outcome of a repeated game may vary with the choice of network model, it is important to get the network model right. Further progress in the theory of games and adaptive strategies would be greatly enhanced by a theory of networks of social interaction. In particular, it would be desirable to have a framework within which models may be developed that are both tractable and plausible as a mechanism governing interactions among a population of agents seeking to improve their lot. When the network changes much more slowly than do the strategies of individuals, it is reasonable to model the social network by a structure that is fixed, though possibly random. The question of realistically modeling the randomness in such a case is taken up in a number of papers, of which a recent and well known example is the "small world" model (15). In the other extreme (16-18), evolution of social structure is modeled by agents moving on a fixed graph in the absence of strategy dynamics. In the general case, however, interaction structures are fluid and evolve in tandem with strategy. What is required here is a dynamics of interaction structure to model how social networks are formed and modified. We distinguish this structure dynamics from the strategic dynamics by which individuals change their individual behaviors or strategies. In this paper, we introduce a simple, additive model for structure dynamics, and we explore the resulting system under several conditions: with or without discounting of the past, with or without added noise, and in the presence or absence of strategic dynamics. Common to all our models is a stochastic evolution † To whom reprint requests should be addressed.

This paper explains why seemingly irrational overconfident behavior can persist. Information aggregation is poor in groups in which most individuals herd. By ignoring the herd, the actions of overconfident individuals (“entrepreneurs”) convey their private information. However, entrepreneurs make mistakes and thus die more frequently. The socially optimal proportion of entrepreneurs trades off the positive information externality against high attrition rates of entrepreneurs, and depends on the size of the group, on the degree of overconfidence, and on the accuracy of individuals ’ private information. The stationary distribution trades off the fitness of the group against the fitness of overconfident individuals. Starting any company is really hard to do, so you can’t be so smart that it occurs to you that it can’t be done.