Cecilia Vicini Ronchetti
In 2018 and 2019, a great number of stories appeared in the popular press about a generation of mostly young white men having been radicalised on YouTube. The New York Times, for instance, reported on Caleb Cain, a young boy who, in 2013, was suffering from depression and looking for ways to better his life (Roose, 2019). Given that he already spent most of his time on YouTube, he also turned to the online video platform with the question on how to better his life. Cain describes his experience at the time as falling down a rabbit hole. Following some videos about neuroplasticity, he stumbled on a series of content creators who presented themselves as self-help gurus and father figures, teaching young men how to navigate the, in their view, confusing and harsh modern world. One particular individual that had a profound impact on Cain was the self-proclaimed Canadian philosopher Stefan Molyneux. It was only after having consumed much of Molyneux’s content and having been recommended more far-right videos, that Cain found out that the self-help advice from this fatherly figure came with extreme radical ideology.
Molyneux gained a reputation for operating like a cult leader, telling children that they needed to cut ties and abandon their families if their parents did not take care of them. While starting mostly as a libertarian in his early YouTube days in 2006, Molyneux, around 2014, increasingly engaged with white supremacist talking points. For instance, Molyneux engaged in extreme debates around the so-called Great Replacement and scientific racist ideas about the biological connection between race and IQ. Furthemore, the Southern Poverty Law Center has labelled Molyneux “a libertarian internet commentator and alleged cult leader who amplifies "scientific racism," eugenics and white supremacism to a massive new audience.” Following the 2019 terrorist attack in Christchurch, New Zealand, the New Zealand Royal Commission reported that the radicalised far-right individual that entered a Mosque to kill 51 people was claimed to have consumed and donated to far-right YouTube channels such as that of Molyneux. It was only in 2020, when Molyneux had amassed a staggering one million subscribers on YouTube, the online video platform updated its terms of service and removed his content from their platform (Alexander, 2020).
Following the above-mentioned events, research into YouTube radicalization has taken off. One line of research has focused particularly on the role played by algorithmic recommendations, which on YouTube account for around 70 percent of time spent on the platform (Yesilada & Lewandowsky, 2021). Another line of investigation, critical of the technological determinism behind this line of research, has argued that YouTube ‘simply’ offers an alternative market for political content not provided by legacy media and the popularity of far-right content is part of supply and demand dynamics (Munger & Phillips, 2020). In addition, scholars highlight that in YouTube’s attention economy, ideological entrepreneurs use influence tactics to establish and maintain a dedicated fanbase. In other words, they engage in influencer drama, participate in debates and interviews with other channels, and perform qualities such as authenticity, relatability, and accountability in their videos (Lewis, 2018, 2019, 2020). Through such efforts, ideological entrepreneurs, like other YouTubers, aim to establish a parasocial relationship with their audience which can in turn increase radicalisation among followers of alt-right content creators.
One important aspect that has received less attention in the literature is that on YouTube, audiences do not merely watch content; they can edit and re-upload favourite clips in a more participatory culture style (Jenkins, 2006), or, more often, audiences perform small acts of engagement (Picone et al., 2019) via (dis)likes, shares, and comments to let content creators know what they think. On YouTube, as Lange (2009) has argued, ‘reciprocity’ is one of the defining cultural dynamics. Maddox (2021), in the context of ASMR videos, has stated that reciprocity is best understood “as transactional tingles: relaxation [or another experience] in exchange for likes, clicks, and views within the attention economy.” While this reciprocity is more easily understood regarding likes, clicks, and views, the YouTube comment section allows for a more elaborate feedback mechanism. On the one hand, comments reveal audience reception, where audiences, following Hall’s (1973) classic observation, perform dominant, negotiated, or oppositional readings of video content. On the other hand, however, as Ha et al. (2022) point out, “YouTube comments have all the characteristic traits of participatory culture, meaning that it is a site of intense collective sense-making and knowledge production.” The ideological entrepreneur on YouTube thus engages audiences through the camera's lens while simultaneously creating a space for viewers to engage with other audience members in collective sense-making. Both these dynamics on YouTube strengthen parasocial ties, where pseudonymous individuals develop, adjust, share, and construct their worldviews.
This project aims to study parasocial ties in Molyneux’s community in relation to his content. More specifically, through a study of audiences’ small acts of engagement, the project aims to understand how the logic of parasocial relationships, mediated by YouTube’s infrastructure, mixes with far-right ideological discourses on white supremacy and misogyny.
This project uses a unique historical dataset to study audience engagement with Stefan Molyneux. As mentioned above, Molyneux’s channel was removed in 2020 by YouTube, resulting likewise in the removal of his entire video archive. In 2019, Dutch journalists captured a wealth of political content to study radicalization on the platform. Since then, their efforts have stopped, but their data collection is now used by a.o. Daniel Jurg for further historical excavation. This collection includes a dataset of video transcripts and 1.9 million comments from 2006 to 2018 under Stefan Molyneux’s videos, including likes, dislikes, usernames, timestamps, and video ID. Using this data set allowed us to reconstruct audience engagement with Stefan Molyneux to understand the dynamics within its participatory culture.
All transcripts and comments from Molyneux’s channel from 2006 to 2018 were downloaded as a CSV file from the SQL historical YouTube database and uploaded to 4CAT, where we used various analytical modules to perform a computational grounded theory (Nelson, 2020).
The overarching research question of this project is:
How did Molyneux’s YouTube community evolve from its origin to his channel being deplatformed?
What are the defining political moments in this evolution?
What are the key changes within the comment section during this evolution?
Molyneux brands himself as a philosopher and man of logic, reason, and science. We expect to find that audiences value Molyneux’s rationality and calm demeanour. This might make his white supremacy discourse appear less based on racism and more on rationality. On the level of misogyny, we know from Caleb Cain that Molyneux mixes personal stories and an understanding of the struggles of being a young white man with a larger ideological agenda that espouses hypermasculinity and sexism. We expect to see this also captured in the comments. Ha (2021) has argued that reactionary ideological entrepreneurs on YouTube evoke discourses on so-called ‘white vulnerabilities.’ We expect that audiences might share such ‘white vulnerabilities’ in the comments section and draw connections to their (parasocial) relationship with Stefan Molyneux and hyper masculine and white supremacist ideas.
We aimed to inductively map audience engagement with Molyneux using 4CAT and internet archives.
We first uploaded the dataset with the 1.9 million comments to 4CAT. To create the first graph visualising the amount of comments per month, we ran the monthly histogram feature to take note of commentary peaks in Molyneux’s channel. Secondly, we tokenized the body column per year, filtered stop words (for many languages), and lemmatised the content. Then, we visualised a rank flow diagram for the most frequently used words in the comment section by running a term frequency-inverse document frequency (tf-idf). Thirdly, using the filter option, we filtered the original dataset by value, creating a new dataset with comments that included the word ‘white’ in the ‘body’ column. We then tokenized this new dataset, after which we used the function ‘extract co-words’ to find the words most used in combination with the term ‘white’ sorted by year, only keeping unique co-word pairs per post. To make sense of the changes in these co-word frequencies related to the term ‘white’, we traced back popular thumbnails and video titles using the wayback machine, which correlated with these most popular terms.
Fourthly, we used the ‘count values’ function to determine the top 15 commenters per year, and visualised this in a rankflow diagram. From this diagram, we selected six super-participants by their activity, favouring commenters with a longer history as top commenter over commenters with a more short-lived status as top commenter. For these six selected super-participants we filtered the dataset by their username in 4CAT, creating six separate csv files for every user with their comments to closely analyse their commenting trajectory. Moreover, we utilised the same method to analyse the commenting trajectory of Stefan Molyneux himself, since he became a super-participant himself from 2014 onwards. We then combined the seven csv files into one file, visualised this data in a timeline, and plotted every comment in different sizes depending on the amount of likes it received (the bigger the bubble, the higher the amount of likes). Lastly, we selected two quotes per participant to illustrate the evolution of their ideology.
Through the histogram in Fig.1, we were able to take note of the months with the most engagement, relative to comment frequency. The first major peak corresponds with both the Zimmerman trials and Molyneux’s commentary in July 2013. Indeed, the most viewed and liked video of this month is “The Truth About George Zimmerman and Treyvon Martin” published the same day of the verdicts. Interestingly, this video, along with Molyneux’s other content, was endorsed and shared by David Duke, a former Grand Wizard of the Knights of the Ku Klux Klan, on his personal blog site (seen above and captured by the Internet Archive’s Wayback Machine). At the time of the data scrape, it had 1,135,366 views, 14,492 likes, and 2,516 dislikes. The second major peak is identified as April 2016 in which his interview “Race, Genetics and Intelligence | Helmuth Nyborg and Stefan Molyneux'' is one of the three better performing videos of that month, totaling 2016,666 views, 4,935 likes, and 358 dislikes at the time of the data scrape. The third and final identified peak takes place in August of 2018. This month’s most viewed video is Molyneux’s “Do You Have Any Questions Which Aren’t Race-Baiting?” which amounted to 463,386 views, 21,399 likes, and 338 dislikes when the dataset was captured. Though we did not have access to comment count in the cvs file containing all video titles, likes, dislikes, and timestamps, we estimate that the commentary peaks largely correlate to the most viewed or best performing videos of the months identified, where commenters would likely encounter a more active and engaging audience.
Using a top-word count of the nearly 2 million comments left by Molyneux’s audience, we were able to highlight dominant trends within his community and visualise it in Fig. 2. The years before 2014 demonstrate a largely libertarian discourse where popular terms include ‘government’, ‘money’, ‘market’, and ‘power’. Then, in 2014 a major shift can be seen whereby the libertarian rhetoric is suddenly replaced by terms which become much more frequent such as ‘woman’, ‘black’, ‘trump’, and ‘white’. This shift from libertarian values towards increasingly tokenized alt-right discussion points in Molyneux’s comment section attests to the changes in his content, including thumbnail style, topics, commenting habits etc. Further, we noticed a rise of the influencer Stefan, a difference from the colloquial ’Stef’.
In Fig. 3, we highlight the most significant topics per year around “white” as they were perceived by Molyneux’s community. After an in-depth analysis of the content creator’s Youtube channel through the decade for which we have a large amount of quantitative data, we traced these topics back to highly engaged-with videos to capture the critical political events that both Molyneux and mainstream media outlets were reporting on, showing the events’ influence over the discourse itself. To reiterate our earlier argument, 2013 was dominated by the Zimmerman trials, followed by the shooting of Micheal Brown and the Ferguson riots in 2014. 2015 saw a rise in dominance by the European migrant crisis and 2016 saw an increase in discussion of Black Lives Matter and, of course, Donald Trump’s election. 2017 highlights the misogynistic discourse on his channel, as made evident by the coverage of the Women’s March. Finally, in 2018 we see a surge in the conversation of white racism in South Africa. It is perhaps interesting to note that Molyneux’s style in both thumbnails and titles largely follows Youtube norms when they were published. For example, in the beginning of his Youtube residency in the late 2000s many of his videos are titled as if they were lectures (i.e., “Intro to Philosophy”) and the thumbnails are simply frames chosen from the platform from the videos itself. Then, from 2011 onwards, Molyneux begins to create more colourful thumbnails featuring text and stock images. Finally, from 2016, towards the end of both our dataset and his Youtube platform his thumbnails are increasingly saturated, black and white, and feature a lot of red in both text and images. Moreover, his titles become accusatory and utilise many capitalised characters.
Using Fig. 4, we were able to identify the top commenters under Molyneux’s videos from 2008 to 2018. This rankflow diagram provided us with information on which top commenters remained within the top 20 for multiple years, which in turn enabled us to identify six super-participants which remained within the top 20 over a considerable amount of time, enough for us to be able to track changes in their commenting behaviour. As visible in the diagram, from 2014 onwards, Stefan Molyneux himself appears in the chart as one of the top commenters and dominates the charts from 2015 to 2018.
To summarise, the contributions of this research are threefold.
The results of this research should be interpreted with caution due to the following limitations:
Transcripts of the videos were unusable. The datasheet of this project included not only 1.9 million comments from Molyneux’s videos, but their transcripts as well. The transcripts, auto-generated by Youtube, were incoherent and difficult to understand. Thus, we were not able to analyse it.
Datasets lack information about dislikes and replies to the comments. Meaning that we were not able to see the dislikes of the comments, which made analysing the opinions of his audience less clear. Likes and dislikes in comments can be used to gauge the general sentiment of a community's discussion. Likes and dislikes can also be used as a way to encourage participation in a discussion. If someone receives a lot of likes on their comment or post, then that might encourage them to continue participating in the discussion. Yet, working with only likes can make the discourse distorted.
From this research it can be concluded that radicalisation (or at the very least a drastic change in discourse) within an online community or digital culture does not happen in a vacuum. Political events strongly correlated with comment and content changes in Molyneux’s community. Thus, we cannot overestimate the role of algorithmic recommendations, falling into the trap of technological determinism, in which the algorithm is over credited for its role in online radicalisation (Yesilada & Lewandowsky, 2021). However, we also cannot underestimate the role the affordances of a digital platform like YouTube play in enhancing community building in this case study, since Stefan Molyneux does use influence tactics to create parasocial relationships with his followers. Thus, the popularity of far-right content is not simply part of supply and demand dynamics (Munger & Phillips, 2020). Taking a more in-depth look at how Molyneux’s content has changed in relation to Youtube’s evolution of its affordances can help us further understand the role these affordances play in the radicalisation of online communities.
Despite some limitations, one of the implications that arose in conducting this research is tied to the importance of internet archives as content moderation and de-platformization become norms for platform conduct. In-depth research and analysis of content creators who have been banned and whose accounts or channels are completely deleted from a platform cannot be completed without some use of internet archives, such as the Wayback Machine. Though some digital communities may be found harmful to the overall environment of a certain platform, their radicalisation and their internal dynamics may still include vital information for further academic analyses dealing with alt-right communities.
Alexander, Julia. “YouTube Claims Its Crackdown on Borderline Content Is Actually Working.” The Verge, December 3, 2019. https://www.theverge.com/2019/12/3/20992018/youtube-borderline-content-recommendation-algorithm-news-authoritative-sources.
Alexander, Julia. “YouTube Bans Stefan Molyneux, David Duke, Richard Spencer, and More for Hate Speech.” The Verge, June 29, 2020. https://www.theverge.com/2020/6/29/21307303/youtube-bans-molyneux-duke-richard-spencer-conduct-hate-speech.
Anderson, Eric. Inclusive Masculinity. 2010. Milton Park, UK. Taylor & Francis Books
Connell, R. W., and James W. Messerschmidt. “Hegemonic Masculinity.” Gender & Society 19, no. 6 (2005): 829–59. https://doi.org/10.1177/0891243205278639.
Donovan, Joan. Meme Wars: The Untold Story of the Online Battles Upending Democracy in America. S.l.: Bloomsbury, 2023.
Donovan, Joan, Becca Lewis, and Brian Friedberg. “Parallel Ports. Sociotechnical Change from the Alt-Right to Alt-Tech.” Edition Politik, 2018, 49–66. https://doi.org/10.14361/9783839446706-004.
Ha, L., Graham, T., & Gray, J. (2022, October 5). Where conspiracy theories flourish: A study of YouTube comments and Bill Gates conspiracy theories. Misinformation Review. Retrieved January 19, 2023, from https://misinforeview.hks.harvard.edu/article/where-conspiracy-theories-flourish-a-study-of-youtube-comments-and-bill-gates-conspiracy-theories/
Hall, Stuart. Encoding and Decoding in the Television Discourse. Birmingham: Univ. of Birmingham, 1973.
Hawley, George. Making Sense of the Alt-Right. New York: Columbia University Press, 2019.
Jenkins, Henry. Fans, Bloggers, and Gamers: Exploring Participatory Culture. New York, NY: New York University Press, 2006.
Lange, Patricia G. Thanks for Watching: An Anthropological Study of Video Sharing on YouTube. Louisville: University Press of Colorado, 2009.
Lewis, Becca. “Alternative Influence.” Data & Society Research Institute, September 18, 2018. https://datasociety.net/library/alternative-influence/.
Lewis et. al . “Parallel Ports: Sociotechnical Change from the Alt-Right to Alt-Tech.” Accessed January 19, 2023. https://mediarep.org/bitstream/handle/doc/13283/Post_Digital_Cultures_49-65_Donovan_ea_Parallel_Ports_.pdf?sequence=5.
Lewis, Rebecca. “‘This Is What the News Won’t Show You’: YouTube Creators and the Reactionary Politics of Micro-Celebrity.” Television & New Media 21, no. 2 (2019): 201–17. https://doi.org/10.1177/1527476419879919.
Maddox, Jessica. “On the Limits of Platform-Centric Research: YouTube, ASMR, and Affordance Bilingualism.” International Journal of Communication 15 (2021): 1120-1140.
Munger, Kevin, and Joseph Phillips. “Right-Wing YouTube: A Supply and Demand Perspective.” The International Journal of Press/Politics 27, no. 1 (2020): 186–219. https://doi.org/10.1177/1940161220964767.
Nagle, Angela. Kill All Normies: Online Culture Wars from 4chan and Tumblr to Trump and the Alt-Right. Winchester, UK: Zero Books, 2017.
Nelson, Laura K. “Sociological Methods & Research.” Essay. In An Introduction to Computational Grounded Theory. SAGE Publications Ltd, 2019.
Picone, Ike, Jelena Kleut, Tereza Pavlíčková, Bojana Romic, Jannie Møller Hartley, and Sander De Ridder. “Small Acts of Engagement: Reconnecting Productive Audience Practices with Everyday Agency.” New Media & Society 21, no. 9 (2019): 2010–28. https://doi.org/10.1177/1461444819837569.
Roose, Kevin. “The Making of a YouTube Radical.” The New York Times, June 8, 2019. https://www.nytimes.com/interactive/2019/06/08/technology/youtube-radical.html.
“Stefan Molyneux.” Southern Poverty Law Center. Accessed January 19, 2023. https://www.splcenter.org/fighting-hate/extremist-files/individual/stefan-molyneux.Yesilada, Muhsin, and Stephan Lewandowsky. “A Systematic Review: The YouTube Recommender System and Pathways to Problematic Content.,” Internet Policy Review 11. 2022. https://doi.org/10.31234/osf.io/6pv5c.
|jpg||Areagraph_03_Tavola disegno 1.jpg||manage||302 K||21 Oct 2019 - 13:36||EmilieDeKeulenaar|
|jpg||Atlantis_WikiTimeline_Tavola disegno 1.jpg||manage||86 K||21 Oct 2019 - 13:28||EmilieDeKeulenaar|
|jpg||Crusade_WikiTimeline-02.jpg||manage||70 K||21 Oct 2019 - 13:27||EmilieDeKeulenaar|
|png||Screenshot 2019-07-22 at 15.22.51.png||manage||429 K||21 Oct 2019 - 13:20||EmilieDeKeulenaar|
|png||Screenshot 2019-07-22 at 16.42.17.png||manage||527 K||21 Oct 2019 - 13:37||EmilieDeKeulenaar|
|png||Screenshot 2019-07-23 at 12.25.46.png||manage||60 K||21 Oct 2019 - 13:24||EmilieDeKeulenaar|
|png||Screenshot 2019-07-23 at 16.10.01.png||manage||327 K||21 Oct 2019 - 13:31||EmilieDeKeulenaar|
|jpg||WW2_WikiTimeline-03.jpg||manage||66 K||21 Oct 2019 - 13:28||EmilieDeKeulenaar|
|png||cluster 2.png||manage||1 MB||21 Oct 2019 - 13:44||EmilieDeKeulenaar|
|png||image-wall-e3b55f6d8e296e95f13bd18fc943dd55.png||manage||934 K||21 Oct 2019 - 13:33||EmilieDeKeulenaar|
|png||pasted image 0.png||manage||1 MB||21 Oct 2019 - 13:23||EmilieDeKeulenaar|
|png||pasted image 2.png||manage||1 MB||21 Oct 2019 - 13:32||EmilieDeKeulenaar|
|png||unnamed-2.png||manage||12 K||21 Oct 2019 - 13:34||EmilieDeKeulenaar|
|png||unnamed-3.png||manage||11 K||21 Oct 2019 - 13:34||EmilieDeKeulenaar|
|png||unnamed-4.png||manage||54 K||21 Oct 2019 - 13:37||EmilieDeKeulenaar|