Using an Agent-Based Model to Understand Complex Patterns of Teacher Professional Development

• In Event: The Problem of Linear Thinking in Research on Teacher Learning: Can Complexity Theories Help?

Abstract

Problem and Objectives
In this paper I developed and ran an Agent-Based Model of teacher professional development as a potential method for implementing complexity theory in education. Agent-based Models (ABMs, sometimes also referred to as individual-based models) have emerged in the aggregate strain of complexity theory (Railsbeck & Grimm, 2012). They are extremely useful in situations where patterns can emerge in space as well as time (for example, where learning can happen in a classroom and also over the school year). The purpose of an ABM is not to find a single outcome or effect but rather to identify the multiple patterns of interaction between agents and conditions that can result in multiple effects.

ABMs have been used by complexity researchers outside of education because the models help to understand how agents adjust their behavior to the current states of themselves, of other agents, and of their environment. They also help researchers understand what happens to the system because of what individuals do and also what happens to individuals because of what the system does. ABMs are considered ‘naïve’ models because they do not aggregate agents’ behavior or response into abstract variables. Rather, the models directly represent the agents’ behavior in a virtual environment that ‘runs’ so that researchers can see what happens over time.

Procedures and Contribution
To build an ABM for teacher professional development, I first conducted an extensive literature review to develop a set of hypotheses for the initial conditions, processes, structures, behaviors and interactions of teacher professional learning including how these elements vary by scale, location, etc. I then used NetLogo, an ABM virtual environment, to develop the model. The model focused on a hypothetical professional development scenario where 30 teachers attend a workshop and then return to their individual schools to implement what they have learned. Within the scenario, I started with the most simplistic version and then added elements in an iterative process until the model became more and more complex. Iteratively including more and more elements and processes helped to identify what is important in changing teacher behavior and system (in this case their schools’) behavior. I was able to vary elements for both individual teacher participants and by their school conditions. These elements included: length of workshop, the degree of collaboration in their schools, teachers’ motivation for attending, etc. I then ‘ran’ the scenario to see whether and how teachers’ behavior changed as a result of varying conditions. This iterative process led to a working scenario of how patterns of conditions, processes, and teacher behavior might predict learning from professional development.

The use of ABMs in education may hold promise for improving predictive ability. If we can establish that the patterns of relationships identified through modeling are isomorphic with the real worlds of teacher learning, then we can predict what will occur in future situations given a set of circumstances. Once we can predict, we can intervene to improve teacher professional development.

Author

• V. Darleen Opfer, RAND Education