USER PREJUDICES: SOCIAL MEDIA CAMPAIGNS IN BATTLES FOR THE U.S. PRESIDENCY
Final Project, DGT HUM 201: Intro to Digital Humanities
Given the time constraints of this project, only a portion of the dataset will be used. The full dataset will be further analyzed for my PhD dissertation.
About The Data
The data set was found through the “Data Is Plural” newsletter. It was pulled from Fusion.net (now SplinterNews.com). This dataset was used by Fusion.net to create a database of images posted by presidential candidates during the 2016 election. The images are pulled from Facebook, Twitter, and Instagram. The compilers of this data used Clarifai to automatically tag the images. There were a few challenges when working with this data set.
Firstly, the way the data was compiled was not the most accurate in that an automatic tagging program uses a learning algorithm based off previously tagged images and uses the correlations between the images and the tags to do the tagging. The compilers themselves give an example of a family portrait being tagged with “zombie” because of the color palette. The accuracy of the tagging depends on the previous database, the clarity of the image, and any image filters or alterations. The idea of going back and then editing the tags manually was not a conceivable goal during the length of a 10-week course. The next large challenge was the sheer amount of data I had to work with; my computer did not have the processing power for it. In order to work with the data within the time frame I had to narrow down the data to just a few candidates.
I know, I know. Sexism isn’t unexpected; or at least, it shouldn’t be. When narrowing down the data to specific candidates to make it more manageable I was curious about the major differences among the democratic candidates. Many political commentators blamed a split democratic party for Trump’s presidential win during the 2016 election. Many noted that Bernie Sanders took the younger vote from Hillary Clinton. With the precedent of a tech-literate Obama presidency and a generation of young voters heavily engaged in internet debates, a social media strategy has become a necessary element to a successful political campaign.
I began by looking at the Clinton campaign and was curious to see what posts of her social media campaign elicited the most online interactions. When I pulled her top 10 posts a concerning pattern emerged…
THE HILLARY CLINTON STATS
5 out of 10 pictures are with Bill Clinton
3 out of 10 are “political” but 2 of those 3 them are in support of the Obama presidency
Looking at the layout of Clinton’s top 10 posts it became pretty obvious that when it came to social media engagement. The first thing I noticed is the internet public was much more engaged with Hillary when Bill Clinton was present in the picture, but that the pictures of the two them that were most popular were pictures that referenced a family moment (Christmas, Anniversary, New Years Day, etc.). People cared that Hillary Clinton was a wife and a mother, they cared that she was a woman, or rather, that she showed herself in gendered roles that relegated women to private lives within the home. During a class presentation a colleague pointed out that an alternative way of looking at this pattern was that Hillary Clinton’s posts were most engaged with when her connections to powerful men were mentioned. If you included Obama’s posts, 7 out of 10 posts involved men. It was very easy to see that whether it was intended or not, online engagement with Hillary Clinton was gendered.
Looking at the history of the presidency and public relations, presidents that are relatable when it comes to their private lives do well in the public eye. It was quite possible that this focus on private family life for presidential candidates would be a normalized trait across the board and across genders. But it wasn’t…
THE BERNIE SANDERS STATS
2 out of 10 posts are family pictures
7 out of 10 are political and 6 of those 7 are his personal political stances
Well, my gut instincts were right. The first men-identifying candidate, Bernie Sanders, disproved any possible theory that this could be equal treatment across the board. This bothered me, I thought I shouldn’t be surprised, that we’ve been seeing how Hillary Clinton was judged by much harsher criteria than her man counterparts. But it still bothered me.
After looking through the images a dozen times over I finally realized why I was still surprised. Throughout this and previous political campaigns, Clinton supporters and activists noted again and again how unequal the treatment of this woman politician was. This discovery was different because the social media users that were interacting with her posts were not just other candidates or news outlets, they were members of the general public.
Hillary Clinton’s top posts were all on Facebook. An article titled “Social Media Use in 2018” published by the Pew Research center claims that “roughly two-thirds of U.S. adults (68%) now report that they are Facebook users.” The similar percentage of U.S. citizens voted in the 2016 election. Now, taking into account that there are more people in the U.S. than there are U.S. citizens there are more U.S. users on Facebook than there are voters. This engagement on the social media platform is much higher than engagement in U.S. politics which further supports the importance of this tool when understand public sentiment. As of Wednesday, March 20, 2019 Hillary Clinton’s Facebook page has a little bit more than 9.6 million followers. Followers of a Facebook page are more likely to be presented posts from that page regularly on their newsfeeds. Now, a follower does not imply either support nor dislike for a candidate but does show a level of commitment to engagement with that online presence.
These are all important demographics to consider when we think about the importance of this engagement trend with Hillary Clinton’s images.
Now let’s consider the demographics previously mentioned and the knowledge that those engagements are done by a large U.S. public. It is this Facebook public that has passed judgement on Hillary’s constructed online profile. It was not only the political commentators in the media or in government that have framed Hillary’s engagement with politics through her gender. Out of her top 10 posts, only one takes a personal political stance. The truth of the matter is that no one cares about Hillary Clinton’s politics, not the media, not politicians, not the public. Netflix hosts a per-decade mini-documentary series from CNN. When talking about the 2000s in the seventh episode titled “Yes We Can” it explores Obama’s presidential campaign and understandably goes into Hillary Clinton’s previous campaign. Clinton was losing the primary to Obama until a turning point in which she broke down during one of her roundtable discussions about women in politics. The media and other politicians heavily critiqued her for this demonstration of emotion and weakness, but here the public countered with their renewed support. This preference for a Hillary Clinton that plays into the stereotypes of her gender (whether genuine or not) is not a unique occurrence in her political career.
Moving Forward: Challenges and Next Steps
Unfortunately, there was only one other presidential candidate that identified as a woman during the 2016 election: Jill Stein. For an unknown reason, she was not included in the database and her data could not serve to shed more light on women candidates. As this dataset does not include any other seats in government there is limited information available.
Looking at the network graphs created from weighted tag relationships among her posts following her apology for Emailgate there is only one relationship that is shown when filtered for highest weighted connections (Musicians – Audit). There relationship between these two nodes is unclear although “audit” does support a social media campaign focused on damage control. As for centrality measurements, degree and betweenness centrality point towards “musician” as being a significant node, but the reasoning behind that remains unclear. It could be a fault in the tagging program or a fault with my calculations, this requires further study. After hours of filter, network analysis does not seem to bring any illuminating patters. That, or I have yet to find the right combination of filters.
This has easily become a much larger project than intended. The next major step is to set aside a few weeks to play with this data and see what are some other findings I find in the process.
“2000s.” Yes We Can, 7, CNN.
“Hillary Clinton Proved That Sexism Is Worse than Racism in America.” Daily Kos, Accessed 23 Mar. 2019.
Lavigne, Daniel McLaughlin, Ross Goodwin, and Sam. “Here’s How We Captured and Tagged 70,000 Photos to Analyze the Bizarre Imagery of the 2016 Campaign.” Splinter, Accessed 22 Mar. 2019.
Netflix. Accessed 23 Mar. 2019.
Smith, Aaron, and Monica Anderson. Social Media Use 2018: Demographics and Statistics | Pew Research Center. 1 Mar. 2018,
“Social Media Demographics to Drive Your Brand’s Online Presence.” Sprout Social, 5 Feb. 2019,