Author Date

2021-03-15

Degree Name

Department

Computer Science

College

Physical and Mathematical Sciences

Defense Date

2020-11-10

Publication Date

2021-03-17

First Faculty Advisor

Dr. Nancy Fulda

First Faculty Reader

Dr. Christopher Archibald

Honors Coordinator

Dr. Seth Holladay

Keywords

anonymity

Abstract

An ever-increasing number of Americans have an active social media

presence online. As of March 2020, an estimated 79% of Americans were active

monthly users of some sort. Many of these online platforms allow users to

operate anonymously which could potentially lead to shifts in communicative

behavior. I first discuss my compilation process of the Twitter Anonymity

Dataset (TAD), a human-classified dataset of 100,000 Twitter accounts that are

categorized by their level of identifiability to their real-world agent. Next, I

investigate some of the structural differences between the classification levels

and employ a variety of Natural Language Processing models and techniques to

shed some light on the behavioral shifts that were observed between the levels

of identifiability.