CCDS Seed Funding Highlights
The Center for Computational and Data Science offers seed funding for proposals that align with the mission of the Center: working to advance theoretical or applied research in the social sciences using advanced computational approaches, including human language technologies and data science. The goal of seed funding is to support pilot research that will lead to future grant proposals or research publications, as well as to support dissertation research that advances CCDS goals.
We are excited to showcase the research of iSchool faculty and PhD students, and how CCDS seed funding has impacted their work. Our first seed funding follow up features Mook Sikana Tanupabrungsun, a 4th year PhD student at the iSchool, who received seed funding from CCDS to support her PhD work. Mook received a Bachelor’s and Master’s degree in Computer Engineering from King Mongkut’s University of Technology Thonburi in Thailand. She recently defended her dissertation and will graduate in the Spring of 2018, and will be working for Microsoft, Redmond following graduation.
Mook’s dissertation, titled “Microcelebrity Practices: A Cross-Platform Study Through a Richness Framework”, examined the uses of social media by celebrities for growing and maintaining an audience. Specifically, she was interested in exploring the ways in which celebrities engage in several core practices: a) developing and maintaining their online identity; b) interacting with fans; and c) gaining visibility beyond the existing fan base, to expand and maintain audience by examining their activities on Twitter and Instagram.
Mook designed a mixed-methods study, using both computational methods and qualitative data analysis through a three-step analysis. First, she constructed datasets of Twitter and Instagram posts from celebrities over a period of five months, and drew samples of 1,000 posts from each platform. Drawing on the theory of microcelebrity and related literature, she used crowdsourcing to annotate the samples in three categories: identity, interaction and visibility. Specifically, given a post, she asked a crowdsourcer to answer whether or not a) the post expressed identity of the writer; b) it showed an attempt to interact with others; and c) it attempted to promote visibility beyond the existing fan base. In other words, she asked crowdsourcers to annotate a post as either rich or low in identity, interaction and visibility, which are the practices of microcelebrity. Mook developed a platform for annotations and used Amazon Mechanical Turk to recruit crowdsourced workers.
Then, she trained machine learning classification models on the labeled samples to automate the annotations of tweets and Instagram posts. The models were used to predict richness labels of the larger sets of unlabeled data. With the predictions, she used statistical and regression analysis to answer the research questions e.g. to examine the changes of microcelebrity practices over time by audience responses using three richness constructs.
The last step was designed to explain and confirm the findings from the quantitative analyses through the interviews with audience members. Mook identified a list of candidates who recently interacted with the celebrities in her datasets. The interview protocol was designed based on the quantitative findings and included questions like “Do you follow the same celebrities on both platforms (Twitter and Instagram)? Why/why not?”. In total, Mook conducted 15 semi-structured interviews which ranged from 54 minutes to 1 hour and 15 minutes, with an average of 58.4 minutes. All interviews were audio-recorded and subsequently transcribed by Mook for further analysis.
Three findings are highlighted. First, Mook found that celebrities used the two platforms, Twitter and Instagram, differently, and that mainstream and Internet celebrities emphasized different core practices. This finding was well explained by the interviews suggesting that the audiences had different expectations from different groups of celebrities. Second, microcelebrity strategies played an important role in an audience’s engagement decisions. The finding was supported by the interviews indicating that audience preferences were based on some core practices. Lastly, while their strategies had no effect on follow and unfollow decisions, the consistency of the practices had significant effects on the decisions.
When asked if there was anything that surprised her about her research Mook explained that, although she examined the practices of microcelebrity of both types of celebrities (mainstream and Internet celebrities) the interviews with audience members suggested that they perceived no difference between the two types of celebrities. For the public, they are all celebrities who are available but not approachable. Mook’s analyses also suggested some interesting patterns about how the audiences felt toward celebrities’ social media uses. For example, the more celebrities attempted to engage with the public, the less the public wanted to engage with the celebrities.
Mook’s research directly relates to the goals of the Center as it adopted advanced computational methods to understand a social phenomenon. Examples of methods used include machine learning and linguistic analysis. Mook explained that her study would not be completed without the support of CCDS. Mook used the funding to support the collection of qualitative data through semi-structured interviews. Finding willing participants for the study was challenging, and the Center’s seed funding was used as an incentive for her participants. Specifically, she used the funding to offer a $15 Amazon gift card for an hour interview for each complete session. When asked if she plans to continue to pursue this research, Mook said that she plans to conduct further analysis on her interview data to draw deeper implications, but does not have a specific research direction yet.
Mook stated that: “Getting support from CCDS has been a delightful journey for me. My work benefited not only from the seed funding support, but also the support from the infrastructure of the Center e.g., the staff and CCDS researchers who always provided thoughtful feedback. I’d like to thank the Center for their support throughout my dissertation.”
The Center wishes Dr. Tanupabrungsun a heartfelt congratulations on her successful dissertation defense and best of luck in future endeavors!