Critial reflection: Memes & Trolls 4 Trump

Critical reflection of the DMI Winter School 2017

Els, Daan, Tanja and I (Lauren) were placed in group 3: Memes & Trolls 4 Trump. The main aim of our group was to map the community that supports Trump by creating and sharing memes on the internet. After some central brainstorming with the whole group (about 30 people), the group was divided into subgroups. Our subgroup (11 members) was responsible for image-based research. Our corpus was the entirety of memes from the “GodEmperorTrump’-Facebook page. We started our research in three-layers: The image content (what is shown on the meme?), a textual analysis (the codification of emotions through words) and the circulation of the images (what kind of memes appear on what kind of websites?).

Working with a group of 11 members, all with different backgrounds and varying levels of skills, turned out to be as complicated as it sounds. Luckily, Marie took the lead and split the main group into smaller groups; each with their own subtopic. Unfortunately, nobody in our group was really experienced with analyzing data. As such we were, in the first three days, mainly busy with cleaning the dataset, replacing links with images and manually sorting the images or the dataset itself.

The communication was rather poor, because of the group division and varying abilities. The group members who -kind of- knew what they were doing, preferred to work in silence and just work out the tools for themselves. From my perspective: I felt a bit useless. Imagine that two people have to solve a difficult math problem and only one of them knows a little bit about the subject, while the other knows absolutely nothing. The person who knows a bit wants to try out all kinds of methods to discover which one is the right one. If the person that knows nothing asks for a step-by-step explanation, the process of figuring out the best method will only take longer. I kind of felt like that person that asks for explanation all the time.

A way to prevent such feelings of incompetence in the future, would be to divide people into groups by expertise, not only by interest. Or provide lessons to bridge the skill differences in advance. We could choose our favourite subject now, but that was not good for the balance in our group. It would be more fruitful to work in small groups that consist of designers, journalists, and people who really know how datasets work: What you can do with them and more importantly, what you cannot do with them. Our main hurdle was that we had all kinds of great ideas, but the execution of these ideas turned out to be insufficient, time and time again.

Unfortunately, this resulted in few usable findings. Our group was mainly occupied by thinking about new research methods, after the previous ones had failed. Below, we will explain what we did further.


On the first day, our project group mainly focused on becoming familiar with the data. Our group looked at several memes to discover how many of them there were and which of them were most liked. A new column was added to our excel sheet containing our data, to show the percentage of interaction per meme. This percentage was calculated with a formula. This formula took into account the number of reactions, likes and comments on each meme.

The aim of our research remained unclear for quite a long period of time. As a result, a lot of time was spent simply working with the tools without a clearly defined research question. This resulted in a lack of structure in our work process.

Bernard helped us transform the IMG URLs into visible images in our Google spreadsheet. The images had to be downloaded with the tool “Tab Save”, as the images took up a large part of the memory on the hard drive. This task was divided between all members of the project group. After some time, it became evident that one of the group members was able to perform this action much quicker with a different tool. Unfortunately, this was only clear after a while, after a lot of work had already been put into this task. This caused us to lose a lot of time.

The downloaded images were processed through Image Plot. This tool appeared to be reasonably complex, making the purpose of the tool unclear to most group members. Image Sorter was utilized to categorize the images based on their visual characteristics as well. The idea behind this was that, after sorting, the resulting green area could be used as a means to recognize the Pepe Memes. Unfortunately, this approach was also unsuccessful. The predominantly green area hardly contained any Pepe Memes. The output was visually engaging but did not contain any new information.

At a later stage Google Reverse Image Scraper was employed. This tool did not perform according to our expectations. After importing a number of images, the tool crashed continuously. All members of the group attempted to locate the ‘God Emperor Trump’ memes with the use of the Reverse Scraper. This would have resulted in valuable research as the alt-right could have been located with this approach. However, since the tool crashed repeatedly, only 1% of the data could be used. The chosen subset was supposed to contain the most popular memes. Despite our attempts, data was inaccurately characterized and thus the 1% of the memes were not those with the highest participation rate.

Besides that, Google Image Sorter recognized memes based on existing images solely as the original image instead of as the meme. As such we could only recognize the most interacted with memes manually and failed to localize where these memes originated from or were commonly used. However, we did manage to do some in depth (manual) research on memes themselves:


The God-Emperor or how memes helped electing Donald Trump

The main goal of our research was to map the meme-posting community of Trump-supporters. Our data were collected from the Facebook page named ‘GodEmperorTrump’. In this section, I will provide a brief overview of the role played by online Trump supporters and their memes in the 2016 election.

Interestingly, the corpus for our research project, the GodEmperorTrump Facebook page, was deleted just a couple of days before we started. Luckily, Bernard already collected the data before that moment. Nevertheless some questions arise: what kind of page was ‘GodEmperorTrump’? And why was it deleted?

To understand the GodEmperorTrump page and its users, we have to sketch a bigger picture of the so-called alt-right, which can be seen as both a political ideology and a somewhat underground online community, which resides largely in anonymous image boards and social media like Reddit/pol/, 4chan and 8chan. Many users of this website show disdain for the more mainstream social media like Facebook and Twitter, as well as traditional news media, which they find to be ‘politically correct’, or in their own terms, bastions of ‘SJW’s’ (Social Justice Warrior).

Whereas Facebook and Twitter actively intervene and ban users who (regularly) post e.g. racist and sexist content, the websites mentioned above tend to minimize their interventions, providing users with a lot of freedom of expression. When frontrunners of the alt-right movement like Milo Yiannopoulos are banned from Twitter, the alt-right community sees this as confirmation of their believe that freedom of expression isn’t guaranteed in ‘mainstream’ social media. For them, the GodEmperorTrump page being deleted is such a confirmation.

But what made Facebook decide to delete the page? What kind of content was distributed at GodEmperorTrump, and is still distributed at comparable pages such as Reddit/pol/? Mostly, self-edited images, commonly known as memes. Memes, in the original sense intended by Richard Dawkins who coined the term, are cultural signifiers that spread simply because they are good at spreading. The title of the Facebook page itself refers to often used memes of Donald Trump as ‘God Emperor’ (a term derived from the game Warhammer). Lots of variations are used, such as Trump as Napoleon or Caesar. What they have in common is a depiction of Trump as mighty strongman or dictator.

Another example of a commonly used meme in the community is Pepe the Frog. Pepe evolved from a rather ‘neutral’ meme that could mean anything, to a symbol of the alt-right. This happened largely by coincidence, but the consequences seem irreversible: the Anti-Defamation League, an American organization opposed to anti-Semitism, listed Pepe on their database of hate symbols.

This is not to say that all Pepe and GodEmperor memes are explicitly and intentionally racist. There is a lot of irony in the meme-culture, that isn’t always simple to spot. It takes a fair bit of ‘Reddit/4chan-litteracy’ to really understand the symbols and slang that are used in this community. But racist or not, it is clear that memes like Pepe were used almost as a campaigning tool. Therefore, it makes sense to research these memes to try to further understand what role they played in the election.


Although we spent a lot of time trying to answer our research questions, the only information we gained was that which we found outside of our use of the winter school tools. So, what would we like to see differently next winter school? As mentioned above, the way the program was set up was not beneficial to our research. In the future it would help if all participants were properly prepared to handle the data they will have to work on. The difference in skill should have been beneficial, but the difference in knowledge on the subject matter, instead, worked against us.

Aside from more guidance on the topic itself, more time was of the essence. Every problem that was solved turned into another problem. Perhaps, our research could have been more successful with more time. One week to master a skill that we had previously only touched upon was too short for us. Not to mention everyone in our group was too busy attempting to fix errors or trying out new research methods, when the ones we used failed, to answer our questions.

Although we cannot speak for the other disciplines that were present in our research group, I think that for us, journalism students, more knowledge on the use of tools and data analysis could have provided us with the necessary skills to at least keep up with the participants that did have some idea as to what they were doing and how to accomplish their goals. For us, the gap in knowledge and the lack of time for the other participants to explain the tools and ways of analysing data to us was, among other things, fatal.

Of course, although many things could have gone better, we were not completely without any results. Despite being unable to answer any of our research questions or finish any of the research, some discoveries were made that needed more time to be fleshed out. As such, these discoveries were not sound enough to be used as actual research results, but were nonetheless interesting.

For example, Rhubini and Marie managed to collect just enough data to discover that a popular use of “Trump-memes” on the Alt-Right Facebook page “GodEmperorTrump” was future news prediction. Although their findings still required a lot more research, a preliminary video could be made:

Furthermore, some of the visualisations could have proven promising for our research, had they been based on the correct data:

All in all, it would have been great if any of the tools we tried to use could have answered our questions about the data. Perhaps with further instruction and more time, this research project too, could have been a success. Sadly, the tools could not be used to their full extent in our research group. So most of our data had to be analysed manually. Though, because of our manual analysis we can at least present to you the:

Top 10 Trump Memes On “GodEmperorTrump” was one of the main Trump-meme databases.

Here are the top 10 most interacted with posts on the Alt-Right Facebook-page:

  1. Hillary Clinton’s e-mail scandal.

What it’s about: Who hasn’t heard about Hillary Clinton’s e-mails being deleted? This meme deletes Hillary’s femininity by using a picture of her often associated with memes linking het to an alligator and dropping the F in female. Emale (a homonym of E-mail) is the only thing left. The meme also recalls the discussion on Clinton and feminism.

Why it went viral: Use of a renown person, a pun and repetitive use of imagery concerning recent events is a recipe for virality.

  1. Spongebob in a panic.

What it’s about: This meme portrays the reaction of Hillary voters after Donald Trump won the presidential election, despite Hillary Clinton’s concession speech.

Why it went viral: Spongebob is repetitively used as a character in memes. Furthermore, so called “chaos memes” are also found in different forms.

  1. “Pepe” map of States that voted Trump.

What it’s about: Remember this Map from Election Night?

Pepe the Frog was associated by the Alt-Right with Donald Trump during the presidential election campaign. The “Pepe States” are supposed to be the ones where Trump was elected.

Why it went viral: Accurate and recognisable. That’s all you need.

  1. EU reaction to new Putin-Trump relations.

What it’s about: Russia’s Vladimir Putin and new presidential elect Donald Trump are steering towards better relations between the USA and Russia. Meanwhile, in light of recent events between Russia and the EU, the European Union is wary of this friendship.

Why it went viral: Stick figures simplifying current affairs counts towards the shareability of memes.

  1. Starbucks employee fired for anti-Trump Twitter comment.

What it’s about: On Twitter Sam Montgomery alleged that Starbucks fired her after she tweeted she would spit in Trump-supporters cups. This turned out to be “Fake News”.

Why it went viral: Twitter’s concise yet informative messages are used for information sharing on all platforms. Had the story itself been true, this would have been a scandal.

  1. Fidel Castro after his death.

What it’s about: Cuba’s communist leader Fidel Castro passed away last November. The Alt-Right used him as a symbol for Communism.

Why it went viral: The high shock value of the text and currentness of this meme made it go up in the charts.

  1. Crying Democrat after Trump won the election.

What it’s about: Trumps win in the presidential election was met with emotional reactions from Hillary-supporters all throughout the USA. This particular woman, an American expat watching from Sydney, was photographed after her pantsuit and lament drew the attention of the Australian media.

Why it went viral: Janna DeVylder was not the only Hillary-supporter in tears, but the repetitive use of her picture on mainstream media turned her into an overnight meme.

  1. Bernie Sanders wiping off sweat after calling for Democratic vote on Hillary Clinton.

What it’s about: Memes portraying getting out of a sticky situation have used forehead wiping since the “Sweating Towel Guy meme”. After Bernie Sanders called for his voters to vote on Hillary Clinton instead during his speech this photograph was taken.

Why it went viral: The echo of the original Sweating Towel Guy meme helps this meme call on posters’ recognition. The commentary on current affairs helped this post reach new heights.

  1. Pepe’s pride after Trump won the White House.

What it’s about: As mentioned above, Pepe the Frog is used by the Alt-Right as a symbol for Trump. Trump winning the White House was cause for celebration amongst his followers.

Why it went viral: Pepe the Frog is already a meme, and thus has a high virality by default. The timing of this post made it shareable.

  1. Blaire White on illegal immigrants.

What it’s about: Transgender YouTuber and Political commentator Blaire White tweeted about the reaction to “The Trump Wall” of “Social Justice Warriors”, known for their outspokenness on subjects such as gender identity. Blaire White is known to say outrageous phrases (at times for the sake of saying them) and has been attacked for it.
Why it went viral: A YouTuber that has already built a network commented on current affairs.