Agent-Based Models as a Tool for Social Work Research

Social work has traditionally emphasized the interaction of person and environment, with each influencing the other. From this point of view, the interaction of individuals creates the environment, which in turn influences the individuals. Traditional research methods, whether they involve experimental designs or statistical models, cannot capture this dynamic multilevel feedback process. In recent years agent-based models (ABMs) have begun to fill the gap. ABMs are computer simulations, in which the researcher defines the characteristics of individual agents which are then allowed to interact in silico. The resulting output can then be compared to empirical data. ABMs offer a powerful methodology for considering interpersonal processes in groups ranging from half a dozen to hundreds of thousands of individuals. They have found numerous applications in areas relevant to social work, including studies of residential and social segregation, cooperative behavior, formation of social networks and interpersonal aggression. They offer the potential for social work researchers to design and test models of interpersonal processes in such varied situations as therapy groups, social service agencies and neighborhoods.

This workshop will introduce ABMs to social work researchers. The workshop will discuss the following topics:

1) Philosophy of ABMS—why are they of interest to social workers? We will begin by discussing the interaction of person and environment as feedback between system levels, in which the group level emerges from the interactions of individuals and subsequently alters those individuals. A simple model of residential segregation will be used for illustration.

2) Basic implementation of ABMS. We will discuss the basic concepts of agent-based modeling, including agents, agent interactions, and the definition of such interactions within a computer program. We will briefly illustrate these concepts using an agent-based model of self-help groups.

3) Platforms for implementation of ABMS. We will discuss the two most common platforms for ABMs used in social science, NetLogo and RePast.

4) Empirical validation of ABMS. We will illustrate the empirical validation of ABMs using a model of interactions in therapeutic communities.

The emphasis throughout the workshop will be on discussion of the strengths and weaknesses of ABMs as a research tool, along with identification of specific areas of social work research where these models are likely to have an impact in coming years.