The path from conspiracy theories to Trump support: a YouTube study

This project is one part of the larger Winter School 2017 study Beyond the bubble? Exploring the 2016 US election echo chambers

Team Members

Alex Gekker, Nikolaus Pöchhacker, Talía Castellanos, Manon van Hoek, Nienke Zoetbrood

Contents

Summary of Key Findings

While our research did find a connection between conspiracy videos and Trump, it was not a very strong one. Yet the research reveals a very specific formation of narratives around the new president of the united states as well as it shows that the topic is taken up by specific channels.The found videos present a distinct narrative, that presents Trump as a hope against the evil plans of the new world order. Although Trump often refers to the so called climate change hoax, surprisingly he is not a central actor within this issue space or connected narratives.

1. Introduction

Before and during the US presidential election 2016, Donald Trump created and supported several wild claims and conspiracy theories, such as the birther issue and Ted Cruz’s father's connection to the JFK assassination. Refuting the prevailing narrative of conspiracy theories appealing more strongly to conservatives, recent research indicates the appeal of conspiracy theories to both conservatives and liberals depends on political opportunity and power dynamics (Uscinski & Parent in Miller & Saunders 133). In further researching whether conspiracy theorists possibly have a pro-trump attitude, the aim of this project is to detect connections on YouTube from one information ecology of conspiracy theories videos to an assemblage of recommended Trump videos. While this approach is limited to the platform logic of YouTube and its recommendations, it shows the formation of an online discourse and the entanglement of narratives within the community of the medium. Within our research we tried to take the presumed position of a conspiracy theory believer and reconstruct the information ecology such a person would be exposed to, without starting from an obvious position, such as alt right or the official channels of political actors. By doing so, we tried to detect a possible filter bubble in a reverse way on youtube.

2. Initial Data Sets

We identified 20 original YouTube videos on conspiracy theories accessed the 10th of january 2017, which we used as a starting point for the network analysis. In the methodology we will explain why we have identified these videos as an entry point. The seed videos and the main conspiracy theories they refer to are:

3. Research Questions

In our research project, we are exploring this potential connection on YouTube by tracing recommendations of the platform, starting from often viewed conspiracy theory videos. Is it possible via the YouTube recommendation engine to end up with pro-Trump and/or anti-Clinton videos, discursively forming a political ideology and related issues, or are these discursive spaces separated from each other?

In this research youtube as a platform is conceptualized by us as a “discursive device” that shapes the formation of the conspiracy discourse by its recommendations. With this, we are cannot test the hypothesis that conspiracy theorists are potentially trump supporters (that would need further offline inquiries) in its totality, but reflect on the role of youtube in shaping the online discourse. By following the connections it is possible to explore, whether and if so how Trump and/or other central political actors are discussed in discursive conspiracy spaces.
Main question: To what extent are people who tend to watch conspiracy theories videos on YouTube directed towards trump-related videos (via the recommendation engine)?

Sub questions:

RQ1: Is the conspiracy space related to the trump space?
RQ2: Does the conspiracy space lead to other central political actors and to whom?
RQ3: What themes are connected to political actors?
RQ4: How are the political actors represented in there?

4. Methodology

We have performed a platform specific network analysis based on the recommendation engine of YouTube. The recommendation engine of YouTube seems to be working very much on an content-based filtering approach, i.e. it searches for similar content instead of content that has been watched by similar users (which would be collaborative filtering). At the same time the recommendations seem to be based strongly on the watched video and only loosely to the user profile. This could be observed in an comparative experiment, where every team member watched the same video, while being logged in on their personal accounts. The lists of recommended videos did, indeed, differ, but only to a minor degree. In image nr. 1 the similar recommendation for every team member are marked. Therefore, our methodological assumption that the recommendation engines has a strong impact on a shared information ecology formation was confirmed.

Image Nr. 1 - similar recommendations for different users

Phase 1

We started out with 20 different seed videos (see annex), each of the representing a distinct conspiracy, e.g. 9/11, New World Order, JFK assassination. These videos have been identified by natural language search on youtube and was selected based on view count (the highest possible) and “authenticity” of the channel, i.e. documentaries about a conspiracy was not seen as representative. The ids of these seed videos were used for crawling the recommendations (crawl depth: 1) with the youtube tool of the Digital Methods Initiative. The different recommendation networks subsequently were integrated into a common graph. In order to identify videos related to political actors, we first identified the most relevant, i.e. Trump, Clinton, and Obama, and coded them within the dataset and colored their respective nodes in the visualization, to make them identifiable in the graph.

It is important to note that with this method we only get a sample of the possible graph representing the discursive ecology of the conspiracy realm. This means that not every video related will be included, as well as there are nodes without outgoing connections. However, we kept those nodes, as they might represent relevant content of the sample, i.e. Trump or Clinton related videos, that we did not want to lose.

Phase 2

Inspecting the structure of the graph, we identified the conspiracy theories connected to Trump, Clinton, or Obama. Additionally, we analysed the content of the videos mentioning these actors in order to find out whether the overall message of the video was negative, neutral or positive towards the political actor and what general narrative was constructed in relation to them. We categorized the findings in an excel sheet.

Phase 3

Departing from identified bridge-videos, i.e. Trump videos that are connecting different clusters of conspiracy theories, we started a subsequent crawl to check, whether we end up in either a conventional news/issue space or if we are being redirected to official Trump sites.

5. Findings

The connection between Trump and the conspiracy issue space

Recommendations on YouTube show that people who click on conspiracy theory videos are viable to be recommended to videos related to Trump. This is not a big connection statistically. Only 18 out of 1332 recommended videos are Trump related content, but it shows that the connection nevertheless exists. Also, in relation to the sample that we created, the Trump videos show an above average indegree, i.e. videos with Trump content are recommended in average more often (avg. indegree: 28) than the other videos (avg. indegree: 18). While this may not be a statistically significant finding, it indicates a probable centrality of Trump related narratives within the conspiracy space.

We researched recommended videos with crawl depth 1, which means that only the recommendations of the seed video are analysed, as well as recommendations of that video. The fact that Trump videos appeared in this network analysis shows that, although their number is limited, conspiracy videos are in close proximity of Trump videos.

Image Nr. 2 - graph with marked trump videos

Political actors related to conspiracy theories

The conspiracy videos are also related to other political actors, mostly Hillary Clinton and Barack Obama. Of the 1332 recommended videos, 5 recommended videos on Clinton and 36 on Obama. The content of these videos were overall negative towards Clinton and Obama.

Appearances of the political actors

The numbers indicate that Trump is not as strongly connected to the conspiracy realm than expected. Even though Obama had a longer timespan to “become an issue” in these discourses, a stronger connection of Trump would be expected, if conspiracy means trump support. However, as we will see later, the quality of the narratives is quite different.

Clinton is a reptile, really!

There are a few conspiracy clusters that are specifically connected to Trump videos. These mostly revolve around conspiracy theories about 9/11, Big Pharma, GMO’s, Reptilians, Pokemon Go and the climate change hoax. One cluster of conspiracy theories that seems particularly closely related to Trump is the one on the New World Order. This is also the cluster of conspiracy theories to which Clinton and Obama are related.

The only videos that are related to Clinton are on her role in the new world order and the illuminati, and the theories about her being an alien reptile.

Image Nr. 3 - Clinton as a reptile http://areazone51ufos.blogspot.nl/

Theories that seem not to be directly related to Trump or any other political actors focus for instance on the global financial crisis, the murder of JFK, vaccinations and Paul is dead. So we might say that some conspiracy spaces are more prone to be related to a Trump space than others. Additionally, Trump videos function as a bridge within the recommendation, connecting different conspiracy spaces. Especially, the the connection between the mono-thematic cluster of the New World Order is connected to the more diverse channel of Alex Jones, which is a major player in the conspiracy theory community (see also image nr. 4)

Image Nr. 4 - Trump as a bridge between NWO and Alex Jones

Trump vs. the establishment

In the conspiracy realms Clinton and Obama are seen rather negative, because they are part of the establishment and therefore central actors within these conspiracies. Most of the analysed videos are highly critical of the government - whether it is due to their role in mass surveillance, the 9/11 coverup or in the new world order. Trump is seen as an outsider who is not connected to the traditional government. Whereas Obama and Clinton are seen as part of the New World Order and other major conspiracy theories, Trump is not associated with these theories. Even in some cases, he was presented as the saviour who would defeat Obama’s new world order.

6. Discussion

During the analysis of the collected graph, we additionally aimed at analyzing the content of direct neighbours of Trump related videos. This way, we would have been able to see, which kind of content is linking to Trump videos within this discursive issue space. However, even with combined efforts we were not able to adapt and run the scripts necessary to extract the neighbours, as the scripts we tried to utilize were originally developed for twitter data. It might be a promising direction to go further and adapt these scripts for Youtube.

It is important to note that with the method applied in our research project we only get a sample of the possible graph representing the discursive ecology of the conspiracy realm. This means that not every video related will be included, as well as there are nodes without outgoing connections. However, we kept those nodes, as they might represent relevant content of the sample, i.e. Trump or Clinton related videos, that we did not want to lose.

7. Conclusions

Trying to detect a filter bubble on YouTube, we researched whether videos on conspiracy videos lead to the trump filter bubble. Following the recommendations of 20 different conspiracies, we mapped the network videos that youtube would recommend you. What we did found indeed is the possibility to end up with pro-trump videos within just two steps. One one hand, the amount of content is not overwhelming, but only a small share of the whole scraped content (ca. 1.35%). Yet, on the other hand, the Trump related videos show in average a higher in-degree than the other nodes.

While our research did not find a strong connection between conspiracy videos and Trump, it reveals a very specific formation of narratives around the new president of the United States as well as it shows that the topic is taken up by specific channels.The found videos present a distinct narrative, that presents Trump as a hope against the evil plans of the new world order. Although Trump often refers to the so called climate change hoax, surprisingly he is not a central actor within this issue space or connected narratives.

An additional main finding of the research project is, that there is a connection to Trump, but only as part of a very specific narrative. The position of Trump as an anti-establishment outsider was integrated into the realm of conspiracy theories and reinforced this image. While, there is this overall positive image of Trump, an interesting follow up question could be, if Trump in his official function as a President is able to stabilize this narrative over time. Additionally, recommendations departing from Trump videos just brought us back in the conspiracy realm, not uncovering any connection to official news sites or the mainstream discourse. This indicates that the interplay between users curating their content and the recommendation engine of youtube are creating a somehow solid filter bubble revolving around very specific topics, which cannot be escaped without explicit user intervention.

8. References

Joanne M. Miller & Kyle L. Saunders. “Conspiracy Theories in the United States: More Commonplace than Extraordinary.” Critical Review, 28:1, 127-136, 2016. < http://dx.doi.org/10.1080/08913811.20> Accessed 12 January 2017.