Year of Publication

2021

Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation

College

Engineering

Department/School/Program

Computer Science

First Advisor

Dr. Judy Goldsmith

Abstract

In an increasingly digitally interconnected world, the study of social networks and their dynamics is burgeoning. Anthropologically, the ubiquity of online social networks has had striking implications for the condition of large portions of humanity. This technology has facilitated content creation of virtually all sorts, information sharing on an unprecedented scale, and connections and communities among people with similar interests and skills. The first part of my research is a social network evolution and visualization engine. Built on top of existing technologies, my software is designed to provide abstractions from the underlying libraries, drive real-time network evolution based on user-defined parameters, and optionally visualize that evolution at each step of the process. My software provides a low maintenance interface for the creation of networks and update schemes for a wide array of experimental contexts, an engine to drive network evolution, and a visualization platform to provide real-time feedback about different aspects of the network to the researcher, as well as fine-grained debugging tools. We conducted investigations into the opinion dynamics of networks when multiple agent “archetypes” interact together with this platform. We modeled agents’ archetypes with respect to two attributes: their preference over their friends’ opinion profiles, and their tendency to change their opinion over time. We extended the current state of agent modeling in opinion diffusion by providing a unified 2D trajectory/preference space for agents that incorporates most common models in the literature. We investigated six agent archetypes from this space, and examined the behavior of the network as a whole and the individual agents in a variety of contexts. In another branch of work using our software, we developed a network of agents who must carry out both economic and social activities during a pandemic. Agents’ decisions about what actions to take (self-protective measures like masking, social distancing, or waiting to run errands) are based on several factors, including perception of risk (obtained from news reports, social connections, etc.) and economic need. We show with preliminary testing that this platform is able to execute standard pandemic models successfully with the incorporation of the economic and social dimensions, and that this paradigm may provide useful insight into effective agent-level response policies that can be used in concert with other top-down approaches that comprise most of the recent pandemic response research. We have investigated the implications of varying behavior profiles within a network of agents, and how those behavioral compositions affect the overall climate of the network in return, and this software will continue to facilitate similar research into the future.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2021.421

Share

COinS