Lia C Bozarth

3rd Year PhD student at University of Michigan, School of Information (UMSI)

Projects on Fake News

Higher Ground? How Groundtruth Labeling Impacts Our Understanding of the Spread of Fake News During the 2016 Election

Sources differ in what they consider to be fake news sites. We observe that depending on the groundtruth, the prevalence of fake news varies significantly but temporal patterns do not. In addition, we show that the age, popularity, and ideological leaning of fake news sites are uncorrelated with whether it is more likely to be listed as fake news by a given source comparing to another. Finally, the agenda-setting difference between fake and real news sites remain relatively constant across different groundtruth lists. This project is currently under review. Download

Collaborators: Ceren Budak, Aparajita Saraf

Lay it Out: Detecting Fake News Publishers through Website Structure Data

We propose a novel website structure based domain-level fake news detection model that has performance results surprisingly comparable to that of existing content-based methods. Through feature analysis, we highlight that fake news sites have more clustered subpages and more ads links, whereas traditional news sites are more substantive and more likely to contain staff links. We also demonstrate that the performance of existing content-based models improve significantly by incorporating structural features, particularly when the definitions for fake and traditional news sites are lax. This project is currently between submissions. Download

Collaborators: Ceren Budak

Projects on Political Elites

From Greetings to Corruption: Politicians, Political Parties, and Tweeting in India

Classifying 1711 Indian political elites on Twitter and studying their feeds of about 4.6 million messages, we find that politicians are indeed more keen on establishing personal branding through low-substance, positive-focused messaging rather than broadcasting hard policies stands. Moreover, compared to the party in power, opposition politicians collectively post more detailed tweets and demonstrate higher negativity, especially regarding corruption, on social media. Finally, through contextual human examination of the most retweeted messages from two key leaders - the prime minister and the leader of the opposition, we find that there are qualitatively important distinctions between both the styles of key politicians. This project is currently between submissions. Download

Collaborators: Joyojeet Pal, Anmol Panda

Projects on Social Movement

Beyond the Eye-Catchers: a Large-Scale Study of Social Movement Organizations' Involvement in Online Protests

We build nested supervised learning models and identify more than 50 thousand social movement organizations (SMO) participating in 2 distinct movements on Twitter. We analyze SMOs' role from five different perspectives: commitment, knowledge sharing, community-building, network structural significance, and recruitment. This project is currently between submissions. Download

Collaborators: Ceren Budak

Slacktivist Contribution to Social Media Activism

We measure the value of slacktivists involved in more than 80 protest movement hashtags on Twitter, focusing on their value for content generation, communication, encouragement, and content diversity. Further, we move beyond the existing literature which primarily focuses on direct content contributions by investigating whether slacktivists induce further contributions from other activists by providing them simple cues of encouragement that enhance non-slacktivists' commitment to a movement. This project was published in ICWSM2017. Download

Collaborators: Ceren Budak