Contagion and Diversity of Preference Saori Iwanaga Dept. of Maritime Safety Technology Japan Coast Guard Academy Wakaba 5-1, Kure, Japan s-iwanaga@jcga.ac.jp Akira Namatame Dept. of Computer Science National Defense Academy Hashirimizu 1-10-20, Yokosuka, Japan nama@jcga.ac.jp Abstract— There are growing interests for studying collective behavior including the dynamics of markets, the emergence of social norms and conventions, and collective phenomena in daily life such as traffic congestion. In this paper, we deal with collective behavior in population with several social networks. We showed that collective behavior in cooperative relationships is affected in the structure of the social network and under the heterogeneity of preference. We showed that collective behavior is affected by clustering coefficient. In regular network and small world network for a homogeneous preferences population, minority choice has a chance to spread throughout in the population. This is when the standardized clustering coefficient is low. We showed that collective behavior is affected by standard deviation. In population of uniform distribution and two- sided distribution, minority choice has a chance to spread throughout in the population. This is when standardized deviation is high. Then, we found that clustering coefficient affects on collective behavior in any preferences. On the other hands, collective behavior in diverse population is hard to forecast. Keywords- agent; social network; contagion; collective behavior; preference I. INTRODUCTION There are growing interests for studying collective behavior including the dynamics of markets, the emergence of social norms and conventions, and collective phenomena in daily life such as traffic congestion. Many researchers have pointed out that an equilibrium analysis does not resolve the question of how peoples behave in a particular interdependent decision situation. It is often argued "It is hard to see what can advance the discussion short of assembling a collection of people, putting them in the situation of interest, and observing what they do"[6]. In examining collective behavior, we shall draw heavily on the interactions of individuals. We also need to describe on two different levels: the microscopic level, where the decisions of the individual agents occur, and the macroscopic level where collective behavior can be observed [17]. The greatest promise lies in analysis of linking microscopic behavior to macroscopic behavior [16]. What makes collective behavior interesting and difficult is that the entire aggregate outcome is what has to be evaluated, not merely how each person does within the constraints of her own environment. The performance of the collective system depends crucially on the type of interaction as well as the heterogeneity in preference of agents [8]. There is a growing literature on the approach of bounded rationality, and the hypotheses employed in these researches reflect the ability of each agent to receive partial information from other agents in the course of their interaction [14]. Our model can be interpreted in like manner; however, we intend to combine the hypotheses of adaptation and local interactions in modeling evolutionary processes. The first hypothesis reflects limited ability (on the agent's part) to receive, decide, and act upon information they get in the course of interactions. The second interpretation is that agents perform optimization calculations. We formalize these ideas in a model with a finite population of agents in which agents are repeatedly matched within a period to play a game and we consider to describe coordination (or common interest) game. Here are many parameters to be considered such as payoff structure, localization, the shadow of the future, the number of agents and so on. Among these parameters, we examine parameters: payoff structure and localization. In previous works in the area of collective behavior or coordination behavior, the standard assumption was that agents use the same kind of adaptive rule [4]. In this paper, we departed from this assumption by considering a model heterogeneous agent with respect to their payoff structures. We showed how agents interact with the others and their aggregate. We used the term emergent to denote stable macroscopic patterns arising from the idiosyncratic rules of agents. From these simulation results, we can induce that knowing preferences, motives, or beliefs of agents can only provide a necessary but not a sufficient condition for the explanation of outcomes of the collective behavior. The interaction of many agents produces some kind of coherent, systematic behavior. The surprise consists precisely in the emergence of macrostructure from the bottom up, which is from simple rules that outwardly appear quite remote from the collective phenomena they generate. On the other hands, threshold model [15] has been postulated as one explanation for the contagion. Contagion is said to occur if one behavior can spread from a finite set of players to the whole population. When can behavior that is initially played by only an infinite set of players spread to the whole population? Morris [12] shows that maximal contagion occurs when local interaction is sufficiently uniform and there is low neighbor growth, i.e., the number of players who can be SCIS-ISIS 2012, Kobe, Japan, November 20-24, 2012 978-1-4673-2743-5/12/$31.00 ©2012 IEEE 935