Abstract
This thesis presents the results of investigation into methods of reducing bias in machine-learning-based content moderation. Using methods including regression models, decision trees, and transformer models, I plan to investigate how out-of-distribution data results in a decrease in model accuracy, which models are most resilient to these kinds of errors, and what other methods can be employed in combination with these models to otherwise improve out-of-distribution performance.
Content warning:
This project involves the analysis of data which contains instances of hate speech, discrimination, and other forms of offensive language directed toward minority groups and others. Some terminology is included in Chapter 4.1, Decision Tree Models as part of an analysis of model performance.
Advisor
Smith, Joseph
Department
Statistical and Data Sciences
Recommended Citation
Dieterich, Nathan J., "Reducing Out-Of-Distribution Bias in Machine Learning Models Used for Content Moderation" (2026). Senior Independent Study Theses. Paper 13292.
https://openworks.wooster.edu/independentstudy/13292
Disciplines
Data Science
Keywords
content moderation, machine learning, OOD
Publication Date
2026
Degree Granted
Bachelor of Arts
Document Type
Senior Independent Study Thesis
© Copyright 2026 Nathan J. Dieterich