Romulus-Catalin Damaceanu: Agent-based computational economics using netlogo

Abstract

Agent-Based Models (ABMs) reproduce human decision-makers in a virtual reality, or artificial world. In an artificial world, decision-makers interact with one another, generating possible histories that social scientists can observe, run several times or control (Lane 1993). This description may be obscure to those who do not deal with ABMs in their professional lives, but it acquires a clear meaning if one observes that all of us are acquainted with a specific sort of ABM, namely, videogames. As a matter of fact, videogames embed players in a virtual reality in which certain characters— the agents—operate: monsters, astronauts, soldiers, settlers, and so on. Each of these agents follows certain rules of behavior—a behavior algorithm—which concerns behavior in isolation and, most importantly, behavior when an agent meets other agents. However, possibilities are so many that random meetings can easily push the game into any of a number of very different outcomes. Videogames generate possible histories of interactions between their characters, some of which may be quite surprising. Now suppose that agent-based modeling is made available to social scientists. Social scientists may be able to characterize the typical behavior of certain social actors, e.g., consumers, producers, governments, trade unions and so forth, and the typical behavior of these actors may be expressed as a behavior algorithm. Social scientists may want to know which possible histories could be realized, given these premises, and whether these possible histories were very different from one another or tended to approach a sort of equilibrium. In order to answer these questions, social scientists may build an ABM the agents in which are the social actors relevant to a