Representation & Quality of Digital Data for Health Research

Digital data are increasingly attractive sources for information regarding the health and well being of individuals. However, one significant challenge of using these data is that they are demographically biased and do not contain demographic information about the individuals posting.  This project seeks to assess the representativeness of health-related data gathered from Twitter by automatically predicting the demographic traits of users tweeting. Using a machine learning approach, we seek to develop and implement simple, scalable tools for detecting Twitter users' age, race and gender and apply these to tweets regarding obesity and foodborne illness at the national level.  To learn more about this project, visit the website of Dr. Elaine Nsoesie, project PI.  A paper associated with this project, which applies our constructed gender classifier, can be found here. You can also access a preliminary draft of a paper reviewing existing approaches toward the automatic detection of social media users' demographics here.

Social Media and Public Mourning on Twitter

Emerging work in the field of social media and mourning suggests that social media sites encourage users to maintain continuing bonds with the deceased, and help mourners establish community with one another.  These studies have primarily examined Facebook and Myspace, two highly personalized social media spaces.  However, it is possible that conversations and behaviors regarding death and mourning vary according to the features and norms of usage within particular sites. In this paper, we analyze the tweet-feeds of deceased Twitter users and find that on Twitter these processes are more diverse than early studies of social media suggest. We find that Twitter users engage in a unique blend of public and private behavior pertaining to death and mourning, and that comments left for the deceased within this space range from highly personal, intimate communications to abstracted and impersonal.  Overall, our findings suggest that the structure and function of Twitter has broadened the scope of conversations about death and remembrance in contemporary Western culture by fostering behaviors that incorporate a blend of personal and public communication.  

This paper is under review but you can learn more about this project via UW Today.  It was also covered by Mashable, Bustle, LiveScience and Seattle's local NPR station, KOUW.

Using Twitter for Demographic and Social Science Research

Despite recent and growing interest in using Twitter to examine human behavior and attitudes, there is still significant room for growth regarding the ability to leverage Twitter data for social science research. In particular, gleaning key demographic information about Twitter users - a key component of much social science research - remains a challenge. This central component of this project is a paper that develops an accurate and reliable process for extracting demographic information from social media sites based on data encoded as images rather than text using crowdsourced human intelligence.  The methods described in this paper have been applied to a variety of other projects, including an assessment of same-race connectedness within associative networks on Twitter, an examination of how users within different racial categories discuss the issue of gun control on Twitter, and analysis of variation in Twitter profile content by race and gender.  I also helped facilitate a workshop hosted by IUSSP designed to make digital data more accessible to demographers at the 2016 Population Association of America annual conference.

Examining Community Policing on Twitter: Police action and community response

A number of high profile incidents involving black citizens have brought issues of police-community trust to the forefront of the public's attention.  Community policing tool is often cited as a potential means of mediating this tension, and social media platforms are sometimes cited as contexts in which widespread and convenient community policing may occur.  However, as of yet there is little work that seeks to understand how the police use Twitter as a platform for community engagement.  Drawing upon the framework of community policing, this project examines how law enforcement agencies leverage Twitter as a platform for connecting and engaging with citizens, as well as how citizens respond to this behavior.  Existing work related to this project examines the content of conversation between citizens and police and explores how police use Twitter to engage with the public/how the public responds to this engagement.

United We Tweet? A Quantitative Analysis of Racial Differences in Twitter Use (dissertation - available on ProQuest)

This study is grounded in the perspective that individuals who use Twitter exist within a racialized social structure, and that if handed a flexible platform for communication they may establish different patterns of use. It acknowledges Twitter as a novel social context in which users co-create meaning and structure, and is informed by theory addressing the role of race and racial identity within both online and offline spaces. Chapters analyze black-white racial variation in self-presentation, site use, and network formation using digital traces from two datasets of Twitter of users in the United States. Results indicate that while Twitter is in many ways a race-neutral context, black users are less likely to disclose personal identity indicators, tend to tweet at others less frequently and with a smaller volume of personal ties, and often have higher levels of racial homophily within their networks than white users.  White users are more outwardly vocal, more likely to disclose personal identity indicators, and more likely to engage with Twitter as an information space. Overall, Twitter appears not to be immune to the influence of offline biases and identities, and there are some black users for whom the narrative of Black Twitter – or Twitter as a community building space – may hold true.