Using Agent-Based Modeling to Gain Insight Into Interest Development

• In Event: 66.069 - Motivation and Complexity: The Merits of a Dynamic Systems Approach to Theory and Research on Motivation in Education

Abstract

Objectives

The purpose of this paper is to describe a project to build and test an agent-based model of interest development in a single community. Agent-based models are computational tools that can provide insights into dynamics of complex social systems (Macy & Willer, 2002).

Theoretical Framework

Psychologists note that interest is often generated, supported, and sustained by others (Hulleman, Durik, Schweigart, & Harackiewicz, 2008; Renninger & Hidi, 2002). But psychological theories tend to miss how individuals’ access to social supports for developing particular kinds of interest varies from place to place in ways that are consequential for individuals’ developing interests.

We conceptualize interest as developing within dynamic intersections of people, tools, and activities that can be characterized in terms of flows, networks, and trajectories (Latour, 2005; Leander, Phillips, & Taylor, 2010; Packer, 2010). Tools for studying network dynamics like agent-based models are particularly appropriate for gaining insight into how interest related to STEM develops over time in a place, given constraints on the number of opportunities to pursue particular activities, the expertise and passion of mentors, and the co-participation of friends in free-choice activities.

Methods and Data Sources

In our project, we are building an agent-based model to (1) gain insight into dynamics of STEM interest development and (2) support community planning to strengthen and diversify the ecology of supports for interest development in a single urban community.

There are multiple sources of data that are informing our model building activities. The first is a longitudinal study following a cohort of all children in the district (n = 170) in a single grade level, as they move from elementary through middle school. The study is surveying students each year as to their extracurricular participation and interest in science. These data serve as a basis for calibrating our model. The second source are interviews with 15 children about their free-choice activities. Third, we draw from community members’ own theories about how interest develops as a source for rules in the model, a practice consistent with theory-based community planning models (Connell & Kubisch, 1999).

Findings

In our project, two model building activities have generated insights the team has found useful: rule specification, an activity in which social scientists and community participants help define the rules governing individual agents, interactions, and environments, and model adjustment, an activity in which social scientists work with modelers to explore initial model results and make adjustments to rules and means for exploration. Our model predicts that increasing co-participation in activities with friends and the number of STEM-related free-choice activities are potential leverage points for supporting interest development. Our survey and case study data suggest place-based constraints on both, however, in terms of the size and composition of friendship networks and access to data.

Scholarly Significance

Our agent-based models present a novel approach to exploring the dynamics of interest development that take into account dynamics that psychological models do not. In addition, they are useful guides to community planning, in that they can illuminate leverage points for change.

Authors

• William R. Penuel, University of Colorado - Boulder

• John Falk, Oregon State University

• Lynn Diane Dierking, Oregon State University

• Matthew Hendrey

• Joey Harrison

• Nancy Staus, Oregon State University