Advancing the Art of Simulation in the Social Sciences

Robert Axelrod. 2003. Advancing the Art of Simulation in the Social Sciences. Japanese Journal for Management Information System 12 (2):16-22.

This paper provides a theoretical background on agent-based modeling (with a focus on social sciences). It describes simulation as a third way of doing science, in contrast to both induction and deduction. It finally offers advices for doing simulation research, focusing on programming, analyzing and sharing the results.

Simulation means driving a model of a system with suitable inputs and observing the corresponding outputs. One purpose of simulation is to be used as a scientific methodology (prediction, proof and discovery). Using simulation for prediction can help validate or improve the model upon which the simulation is based. However, even highly complicated simulation models can rarely prove completely accurate (it does not aim to provide an accurate representation of a particular empirical application). While induction can be used to find patterns in data, and deduction can be used to find consequences of assumptions, simulation modeling can be used as an aid intuition. Its goal is to enrich our understanding of funcademtal processes that may appear in a variety of application.

Emergent properties
Large-scale effects of locally interacting agents

Adpative rather than rational strategies
When agents use adaptive rather than optimizing strategies (rational), deducing the consequences is often impossible. Thus, simulation is often the only viable way to study populations of agents who are adaptive rather than fully rational. While people my try to be rational, they can rarely meet the requirements of information, or foresight that rational modle impose (Simon, 1955; March 1978)

Complexity
The complexity of agent-based modeling should be in the simulated results, not in the assumptions of the model.

Three apsects of the research process need to be taken care of once the conceptual model is developed:

Programming
The programming of a simulation model should achive three goals: validity, usability, and extendibility

Analyzing the results
Despite the purity and clarity of simulation data, the analysis poses real challenges. Understanding the results often means understanding the details of the history of a given run (results are path-dependent). In order to determine whetheer the conclusion from a given run are typical, it is necessary to do seveal dozen simutation runs using identical parameters (using different randome number seeds) to determine which results are typical and which are unusual. The statistical method for studying the effects of the changes will be regression if the changes are quantitative, and analysis of variance if the changes are qualitative. As always in statistical analysis, two questions need to be distinguished and addressed separately: are the difference statistically significant (meaning not likely to have been caused by chance), and are the difference substantively significant (meaning large enough in magnitude to be important)

Sharing the results
The basic problem is that it is hard to present a social science simulation briefly. It may be necessary to explain very carefully both the power and the limitations of the methodology each time a simulation report is published