Constructivists rarely formally model their insights because they reject many of the strictures of dominant rational choice formal modeling (given interests, strict methodological individualism, logic of consequences). ABM is a useful modeling environment for constructivists precisely because it is not restricted to rational choice techniques.
ABM is a computer simulation technique with a distinguished history in computer, cognitive, and physical science that finds its roots in early artificial intelligence efforts. More recently, ABM has begun to make inroads into the social sciences. A number of economists, sociologists, anthropologists, and political scientists have begun to apply ABM to specific and general puzzles.2 The essence of this type of modeling lies in the creation of artificial agents that can be envisioned as individuals, organizations, or even states. The modeler endows these agents with individual characteristics (attributes that change from simulation to simulation), the ability to perceive their environment, and decision-making apparatus. The modeler then places the artificial agents in an artificial environment (social and/or physical if modeling spatial/environmental interactions) and lets them interact. The goal is to simulate and understand processes through which macro patterns emerge from the actions and interactions of agents (and their context). As Epstein and Axtell explain:
We view [ABM] as laboratories, where we attempt to "grow" certain social structures in the computer—or in silico—the aim being to discover fundamental local or micro mechanisms that are sufficient to generate the macroscopic social structures and collective behaviors of interest.3
This is a very general view of ABM. Thus described, game theoretic modeling could be undertaken in an ABM environment, though simulation is less utilized in game theory given the attractiveness of closed form analytic solutions often available. ABM is a useful alternative to tradi tional formal analysis because it can be used to explore complex/complicated problems that are too messy for game theory and other types of mathematical analysis.
ABM efforts usually begin by noting how the interesting problems in social science are analytically intractable. To analyze much of social life with traditional quantitative/modeling tools it has been necessary to abstract away much of the messiness that makes politics and other social endeavors interesting—nonlinearity, heterogeneity, nonrationality (strictly speaking), incomplete information, changing values and preferences, and even limited computational ability. Traditional tools such as game theory, econometrics, and other forms of mathematical analysis have a difficult time dealing with these noted characteristics of human beings and social life. They simplify—assuming linearity, homogeneity, rationality, stable preferences, and unlimited computational ability—in order to reach tractable solutions. To be clear, I am not condemning simplification—all modeling and indeed all analysis entails simplification. However, it is difficult to justify abstracting away the interesting parts of social life in the name of analytical tractability when other, similarly rigorous, methods exist that do not have to make similar simplifications.
The flexibility entailed by the ABM methodology—it is possible to endow the agents with almost any kind of attributes and decision-making rules imaginable—allows scholars to move beyond restrictive rationality and still retain rigor. This is an especially welcome development for con-structivists, plagued (unfairly) with criticisms of being "unscientific" in a social context that values science.4 With ABM, it is possible to model con-structivist relationships between agents and structures; that is, agents that are in a mutually constitutive relationship with their social context.
The agents in many ABM applications are heterogeneous and adaptive, rather than homogeneous and rational. These adaptive agents have limited computational ability and "live" in a world of less than complete information. Instead of calculating the optimal course of action based on (often) full information of all alternatives, adaptive agents "rely on heuristics or rules of thumb," that are learned over time, through experience.5 This is entirely compatible with the constructivist notions that agents follow norms or rules that are learned through experience within a social context.
In addition, adaptive agents "inhabit a world that they must cogni-tively interpret—one that is complicated by the presence and actions of other agents and that is ever changing."6 They "generally do not optimize in the standard sense . . . because the very concept of an optimal course of action often cannot be defined."7 Adaptive agents feel their way around and learn from experience in a highly uncertain world. Further, the context or world that agents inhabit is a product of their own actions and interactions. The study of complex adaptive systems and constructivism similarly hold that agents and their intersubjective context continually recreate one another.
ABM is a technique that facilitates computational explorations of relationships and populations that are inherently path dependent and not in equilibrium—the way that constructivists envision political systems. It provides a social laboratory for thought experiments—useful for social scientists and those natural scientists whose subject matter is inherently historical and not easily explored through experimentation (for ethical, moral, or practical reasons). In addition, ABM retains the rigor of formal theory and (for those that desire it) the stamp of science. Robert Axelrod calls it a "third way" to do science; combining aspects of both inductive and deductive approaches:
Like deduction, it starts with a set of explicit assumptions. But unlike deduction, it does not prove theorems. Instead, a simulation generates data that can be analyzed inductively. Unlike typical induction, however, the simulated data comes from a rigorously specified set of rules rather than direct measurement of the real world.8
However, as mentioned above, formal modeling has not been the tool of choice for social constructivist analysis. A number of the theoretical strictures of constructivism directly conflict with the foundations of most traditional formal modeling. Traditional formal modeling tends to be methodologically individualist, committed to objectivity, and wholly positivist epistemologically, in contrast with constructivism's denial of on-tologically primitive agents or structures, ontological commitment to in-tersubjectivity, and debated epistemological foundations (interpretivist or pseudo positivist). Is ABM any different? Can mutual constitution and the norm life cycle be captured in an agent- based model?
Both Jeffrey Checkel and Stefano Guzzini warn constructivists against implementing their insights through individualist behavioral mod-els.9 Guzzini cautions constructivists against "mixing an intersubjective theory of knowledge with an individualist theory of action,"10 while Checkel laments that "all too many constructivists rely" on behavioral models that "are decidedly individualist in nature."11
ABM may appear to exhibit exactly these problems. In these models social structure is often a very simple aggregation of agent actions, something Guzzini blames for "individualist reductionism."12 Additionally, the focus of ABM is the decision making of individual, autonomous agents. However, the ABM approach is not as individualist as it may appear, and the abstraction of some components of constructivist thinking may not be so damaging.
First, while the method of social aggregation is usually explicitly modeled in a simple fashion, this is, in some ways, immaterial to the question at hand of whether entrepreneurs can catalyze change. The more important point is that in ABM simulations the social context emerges directly from agent behaviors—agency constitutes social structure.
Second, as discussed in chapter 3, a focus on internal decision rules does not necessarily equal an individualist ontology. Constructivists have struggled with how to characterize individual agency—how to describe the process whereby agents make decisions and take actions.13 This is a relatively easy task for rationalist thinkers because of their reliance on methodological individualism.14 However, when agents are socially constituted, exist in intersubjective reality, and are assumed to follow a logic of appropriateness, the task becomes more difficult. At the same time, Checkel concedes "where to draw the line between individual and social ontologies is no easy task."
Individual agents, whether they are states or individuals, take actions. Constructivists rightly point to the need to represent this action as embedded in social relations—social structures define the goals, identities, and interests that lead to individual action. Fortunately, the notion of internal rules fits very well with Guzzini's suggestion that constructivists turn to schemata,15 and the composition of internal models can be represented in ABM applications as socially determined. Because of its flexibility in defining agents and their behavior, ABM is a useful tool for exploring con-structivist insights in general, and the norm life cycle specifically.
ABM is ideal for exploring the norm life cycle and the influence of norm entrepreneurs on norm emergence and evolution in preparation for examining the influence of an actual norm entrepreneur in the ozone depletion negotiations. In the modeling exercises that follow I put the norm life cycle on the computer and demonstrate how stylized norm entrepreneurs can catalyze both the emergence of norms and change in established norms over time. While this type of modeling study does not, by itself, generate empirical support for my arguments about universal participation, it does establish that norm entrepreneurs can in principle influence the emergence and evolution of norms, as constructivists argue.16 In addition, the modeling results outline the boundary conditions for the influence of norm entrepreneurs, and highlight additional empirical expectations that structure the case studies in chapters 5, 6, and 7.
Was this article helpful?