Pepe and /pol/

Pepe was first and foremost a comic strip character created in 2005 by Matt Furie, and is still widely used on the Internet for mainstream purposes. It’s surprising, however, to see the recurrence with which the extreme right uses it and the violence of the messages with which this image can be associated.

Back in 2021, right after watching the very good documentary “Feels Good Man” I was to convince a professor to work with a group of students on the use of Pepe by the Alt-right as part of a course on populism and the far right. I decided to do some analysis on a /pol/ archive from 4chan to try and better understand the far right’s link with this character.

4chan/pol/ and its audience

We’ll be focusing here on the use of the Pepe the frog image on the /pol/ sub-section of the 4chan site, as this is a platform particularly attached to this image and very much associated with the alt-right.

4chan is an imageboard created in 2003 by Christopher Poole, inspired by similar sites that already existed in Japanese. The particularity of these imageboards is that they function around images that start each conversation and are associated with the majority of the comments that follow them. Images are thus central, and a communication tool assumed by the platform and its users. Since 2015, Christopher Poole has sold the platform to Hiroyuki Nishimura, a Japanese recently residing in Paris who was previously behind a very popular anonymous Japanese-language messageboard associated with the Japanese far right.

In addition to this emphasis on images, 4chan is anonymous: users are not registered, and most have the default name of “Anonymous” associated with a number so they can follow the thread of their messages. Some may also choose to associate a name with their message, but there is no account system on the site. Only an insecure password protects this pseudonym from being used in parallel by others.

Dataset

Sadly, I’m not able to trace back the exact Terabytes size dataset I used at the time but this dataset covered all 9,882,027 images posted on /pol/ between September 2013 and July 2017. Each image was processed in its thumbnail form, and the time of its posting was deduced from a timestamp in the file name.

Analysis

At the time, I tried to gage the proportion of Pepe posted in this thread, I trained an image classification model trained on over 5,000 with a less than ideal estimated accuracy around 90%. I then tried to apply this model to all thumbnail in the dataset to get an idea of the number of occurrences of this meme in /pol/. Sadly my computer at the time wasn’t able to process this much data in a realistic time frame and I was left with a guess on only a part of this dataset.

By graphing the much less computer intensive timestamp, we see a gradual increase in image volume in this dataset since Donald Trump announced his candidacy in June 2015, with a peak in volume in November 2016, corresponding to the end of the American election campaign that saw Donald Trump win. Of the 487,891 images posted that month, 8,225 were said to contain Pepe. We have prioritized these 8,225 images for further analysis.

Last but not least, this interactive, day-by-day graph allows us to better understand the interests and values of the platform’s users by looking at what they react to. Here, we’ve highlighted local peaks (at an interval of 4,000) to which we’ve added the day’s top news.