Phase 1: November 2017
Given the sizeable dataset to work with and the richness of its contents, both the social network and textual analyses yielded numerous insights, from the expected to the surprising. The first was the prevalence of retweets (i.e. the term "RT"), by far the most frequently used term across all five days. It suggests that the #MAGA hashtag circulates within a populist echo chamber of sorts, rather than being a site of new knowledge creation that one might associate with a research topic hashtag (e.g. #blockchain). At the centre of the text and context of #MAGA is, of course, the polarizing figure of Donald Trump. Considering the pervasive collocation of #MAGA and @realDonaldTrump, the hashtag may be imagined as opening a direct line of communication to the president - for praise and lambastes alike. Indeed, if #MAGA can be said to mean anything, it is a reflection of Trump's public persona: big on self-referentiality, hazy on specifics. It would, for example, be nearly impossible to determine Trump's political values or even party affiliation from the bubble plots.
More surprising, as illustrated by the force-directed placement graph in which colour hue encodes the out-degree attribute, is how little Trump himself participates in the ongoing discursive construction of #MAGA - even on the anniversary of his election. In fact, the @realDonaldTrump handle does not appear once in the "screen_name" field of the dataset (that is, as a sender of a #MAGA tweet) during the entire five days of data collection. It would seem that #MAGA behaves less like a political slogan than as an extension of Trump's brand; its visibility - like a designer's name on clothing - due mostly to the free promotion by others.
With respect to the more direct question of examining #MAGA's uptake in the discourses of white nationalists and white supremacists, as investigated through the visual analysis of the "white" keyword subset, the results were inconclusive but worth further study as predicted. Hashtags symptomatic of white backlash to anti-racist critiques like #itsokaytobewhite, #whitelivesmatter, #whitepride and #whitegenocide have a small but not insignificant presence within the corpus. An unintended consequence of the "white" keyword analysis was the surfacing of conspiracy theory-oriented hashtags like #followthewhiterabbit, #quanon and #thestormiscoming, a facet of #MAGA's use that - by definition, as a fringe opinion - would not be discernible at the overview level of analysis. I was impressed by the capacity of visual analysis to unveil the semantic landscape of #MAGA's use, from its centre down to its margins; it would be an invaluable means for researchers to discover what might not be evident from tracking the #MAGA hashtag natively in Twitter.
For more information, please see the final report for the project.
Phase 2: January 2018 (ongoing)
Phase 2 was undertaken to address the data collection errors in phase 1; that is, the truncation of tweets at 140 characters that occurred due to changes in the Twitter API mid-project. The data cleaning script was also improved to restore more information that was lost from the escaped diacritics and emoticons. The data analysis is still in progress, but initial observations are reported below.
The phase 2 termsberry visualization showed that the top truncated (November 2017 data collection) and non-truncated (January 2018 data collection) tweets remained fairly constant. "RT" continues to be by far the most heavily used term, followed by @realDonaldTrump and the US flag emoticon. Beyond about 10 or so terms, however, keywords start to diverge, changing in response to contemporaneous events (e.g. "releasethememo") and calling into question the reliability of the truncated data from phase 1.
A word cloud (above) displaying the 145 most frequently used words overall in phase 2 of the project reinforces what we might expect to find. What is both surprising and concerning, however, is the considerable increase in visibility since phase 1 of conservative conspiracy theory-esque hashtags like #rednationrising, #qanon, #followthewhiterabbit and #thestorm, which only surfaced in the small "white" keyword dataset in the previous phase.