Relational learning and fairness

This thesis will focus on relational learning in the modeling of text and user roles in networks, and the relative treatment of individuals as related to algorithmic fairness. With the exponential growth in social network data, the need for models of user interaction data is growing. This work presents a model which agglomerates users into archetypes based on topical modeling of the contents of their interactions. It further proposes models and a fairness metric for the creation of classifiers for individuals which control for the relative treatment of individuals