Agent-based Models, Social Network Analysis, and Machine Learning: Exploring Interaction in International Education

Group Submission Type: Formal Panel Session

Proposal

Conventional analyses used in large scale statistical studies are useful in describing and comparing populations on aggregate or “macro-level” measures, such as the average achievement at the school or district levels at one or multiple points in time. And regression analyses shed light on the relative influence of associated factors (e.g., SES, geographical location, parent education) on these aggregate measures. However, these approaches fail to address the influence that micro-level phenomena have on these macro-level patterns, such as the interactions and relational dynamics of individuals, inter-departmental communications, or stakeholder networks. Social institutions - whole communities, religious organizations, economies, households, or schools - are complex adaptive social systems (P Auspos & M Cabaj, 2018; M Jacobson, et al 2019) that are made up of individuals who interact and adapt or change because of these interactions. In complex adaptive social systems..
"Social agents interact with each other through connections and navigate their world by the adaptive processes they apply – making decisions, learning, changing – as they encounter and interact with other social agents."(J Miller & S Page, 2007, p.10).
The individual or inter-departmental interactions in education institutions and systems therefore influence all aspects of education, including student learning outcomes, uptake of a new curriculum or pedagogy, community involvement, alignment of pre- and in-service teacher training, violence tolerance in a school, positive youth development and successful entry into the workplace, to name a few. Thus, it is important that research and development efforts give space to investigating these lower-level interactions as they strive to understand higher level education outcomes and to understand the nature and processes that support education improvements.
There are a variety of tools that have been designed for investigating the interactions of individuals, stakeholders, or system components in complex adaptive social systems. These include, among others: social network analysis (M Jackson, 2008; M Newman, 2010), agent-based modeling (U Wilensky & W Rand, 2015), and machine learning or artificial intelligence - AI (Benaich & Hogarth 2021). These methods have been applied to a broad spectrum of social science research, including the diffusion of innovation (Valente, 1995), social influence on body mass index (R Hammond & J Ornstein, 2014), and transition to publics school choice (S Maroulis, et al, 2014). Given the history and extensive application of these tools in social research, it’s surprising that there has been limited application of these tools in international education.
The panelists will introduce three methods for studying interactional dynamics, as follows: Eric Johnson - social network analysis (SNA); Timothy Slade – machine learning (AI) and RTI’s Loquat application. Elizabeth Randolph - agent-based modeling (ABM). These introductions will be followed by a short, moderated conversation between the chair and the panelists about the practical applications of these tools before opening the conversation up to the audience in the final 30 minutes of the session.
Collectively, we seek to:
1. Increase audience understanding of SNA, AI, and ABM
2. Enhance the value education researchers place on studying local patterns of interactions
3. Achieve audience understanding of the practical applications of SNA, AI, and ABM

Sub Unit

• General Pool

Chair

• Jonathan Stern, RTI International

Individual Presentations

• Social Network Analysis Methods for International Development: Applications in Higher Education Development - Eric Johnson, RTI International

• Agent-Based Modeling: A Method for Understanding Individual,, Social, and Environmental Influences on Equality in the Classroom - Elizabeth Randolph, RTI International

• Right-fit EdTech: Leveraging Loquat and Other (machine) Learning to Support Foundational Learning in LMICs - Timothy Slade, RTI International

Discussant

• Elena Walls, USAID