Date of Award

8-2018

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

School of Computing

Committee Member

Dr. Larry F. Hodges, Committee Chair

Committee Member

Dr. Jerome McClendon

Committee Member

Dr. Bart Knijnenberg

Committee Member

Dr. Shaundra B. Daily

Committee Member

Dr. Sabarish Babu

Abstract

Digital dating abuse is a form of interpersonal violence carried out using text messages, emails, and social media sites. It has become a significant mental health crisis among the college-going population, nearly half (43%) of college women who are dating report experiencing violent and abusive dating behaviors. Existing technology and non-technology based intervention programs do not provide assistance at the onset of abuse. The overall goal of this dissertation is to create a mobile phone application that consists of a detection tool that classifies abusive digital content exchanged between partners, an educational component that provides information about healthy relationships, and a list of nearby resources for users to locate help. For the user-interface design of this application, we conducted a focus group study and incorporated the themes generated from the study to create our Android prototype. We used this prototype to conduct a usability study to evaluate the overall user-interface design and the effectiveness of the features we incorporated into the app. Due to the lack of a publicly available dataset that could be used to create training and testing sets for the classifiers to detect abusive vs non-abusive text messages in the context of digital dating abuse, we first created and validated a dataset of abusive text messages. This dissertation describes the dataset creation, validation process and the results of an evaluation of different classification and feature extraction techniques. The combination of linear support vector machine, unigram input and tf-idf feature extractor with an accuracy of 91.6% was the most balanced classifier, classifying abusive and non-abusive text messages equally well. Finally, we conducted a user study to investigate different visualization paradigms that will assist users to trust the feedback regarding the possible abusive nature of their online communication. Three different visualization techniques were evaluated using survey questionnaires to understand which one is the most effective in invoking user trust and encourages them to access resources for help.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.