Date of Award
Fall 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Sociology
First Advisor
Erikson, Emily
Abstract
The dissertation consists of three relational-based approaches to analyzing bias in socialnetworks and social data. The first study introduces two group fairness metrics that incorporate network structure, revealing its substantial impact on fairness outcomes, which is often overlooked by traditional metrics. The second study investigates information diffusion processes, highlighting the potential for inequality generated by conventional seeding strategies. It focuses on designing equitable network interventions, considering each group’s relative influence. I calibrate the diffusion models to actual diffusion processes to assess various network influence measures and their likelihood of producing equitable outcomes under a set of fairness constraints. The analysis additionally brings to light inherent trade-offs in attaining fairness in information dissemination. The third project introduces a novel approach to sampling social media data that helps mitigate bias when using the sample to estimate population level characteristics. At its core, the approach employs an embedding space to identify a large set of sampling locations orthogonal to a particular topic of interest. Multiple evaluations of the approach show that it effectively mitigates selection bias, resulting in representative samples.
Recommended Citation
Sachs, Jeffrey K., "Beyond Bias: Three Papers on Bias in Social Networks and Mitigating Bias in Social Science Research" (2023). Yale Graduate School of Arts and Sciences Dissertations. 1105.
https://elischolar.library.yale.edu/gsas_dissertations/1105
