Using Search Traffic Analysis to Understand Canadian Right-Wing Extremism Online

A promising technique for assessing the degree of support for right-wing extremism is the analysis of online search queries. Research shows that interest in extremist groups and influential leaders is male-dominated, with the highest interest levels in Ontario, Alberta and British Columbia. Right-wing extremist preoccupations are similar to those in Europe, with a strong anti-Semitic, anti-immigrant and racist focus. Results illustrate an increase in the popularity of extremist views consistent with the statistical increase in hate crimes.

Data Collection to Counter Right-Wing Extremism Online: The Challenge

The problem of right-wing extremism (RWE) is increasingly acknowledged as a significant threat by the Canadian security and law enforcement communities and highlights the vital role the Internet plays in nurturing the RWE’s culture of fear, hatred and mistrust^{Footnote 17} . Outside the security and law enforcement communities, Canadian researchers are working to build a clearer, data-driven picture of the RWE movement’s online activity, in order to inform government efforts and inspire new mechanisms for Canadians citizens to counter the culture of hate online. The number of Canadian researchers conducting large-scale data-driven analyses nevertheless remains relatively small. Barbara Perry and Ryan Scrivens’ seminal 2015 review of Canadian RWE included qualitative analysis of Canadian RWE web sites, and their more recent work on anti-authority movements included structured online data collection^{Footnote 18} . Similarly, recent work on the Canadian Soldiers of Odin by Yannick Veilleux-Lepage and Emil Archambault draws on structured data collection and network analysis^{Footnote 19} .  Otherwise, most work to date on the behaviour of Canadian RWE online has been conducted by other online interest groups who tend to focus on collecting evidence for advocacy rather than for large-scale data analysis^{Footnote 20} .

A number of factors help to explain the relatively small number of such studies. Simply maintaining access to extremism-related data is an ongoing technical challenge, which in many cases exceeds the capacity of social science researchers. In the wake of the Cambridge Analytica scandal and others, social media companies have dramatically limited programmatic access to data on their platforms^{Footnote 21} . These companies have also moved to more aggressively suspend and delete suspected extremist accounts, putting the data from those accounts further out of reach for researchers^{Footnote 22} . Similarly, many long-time RWE web sites, such as Stormfront and The Daily Stormer, have struggled to stay online (and therefore be available for analysis) in the aftermath of the 2017 Unite the Right rally in Charlottesville, Virginia^{Footnote 23} .

Even if these technical challenges can be overcome, collecting a verifiably Canadian RWE dataset is an added difficulty, since the Internet is an inherently trans-national environment. Establishing the geographic location of online activity using publicly available information is a perennially challenging task. Most often, researchers have to infer location based on factors like language, context and user-reported location, all of which are difficult to establish and verify at scale.

The final and perhaps most consequential challenge is establishing and implementing rigorous, ethically sound methods for collection and analysis. RWE activity in Canada and elsewhere often pollutes wider right-wing political discourse online. Otherwise ordinary political conversations can be hijacked by extremist users. Moreover, drawing clean analytical distinctions between what is acceptable, problematic, criminal and violently extreme online is much easier done in theory than it is in practice^{Footnote 24} . The RWE movement online further exacerbates this ambiguity by employing an arsenal of coded language and imagery, and by playing on the inherent obscurity of Internet culture. This creates a significant risk of over-collection and stigmatisation of legitimate political activity online. Even if they create a technological solution for collecting data, researchers must confront and address the very real privacy and ethical challenges of responsibly collecting and analysing this data. These and other challenges have created headwinds for Canadian RWE online research.

Measuring Appetite for RWE Content in Canada

Search traffic analysis offers a new opportunity for analysis of Canadian RWE activity that is both anonymous and geographically-specific. Rather than looking at what users post online, search traffic analysis examines what people look for through online search engines. Search data can also provide rich detail on demographic factors like age and gender.

Social media and forum posts are inherently performative, and as such provide insight into how users want to be perceived online. This can be valuable for the study of rhetoric and group dynamics, but the fact that users are posting in order to be seen makes it difficult to accurately establish users’ intentions and level of commitment. Some users may exaggerate their level of commitment by posting a high volume of high-risk content, while other more committed users may post very little or not at all, for a variety of reasons. Recent research in other extremist fields has shown, for instance, that female participation in online extremist activity is lower when the activity is thought to be public^{Footnote 25} . By contrast, search engines are widely used by all users, who do so without the expectation that they will be seen or judged by others. People tell Google things that they might not tell their closest friend or spouse. As such, search traffic provides both an intimate and broad-based picture of user interests and views. Analysis of search traffic requires the development and deployment of large databases of search phrases that indicate an interest or affiliation with extremist groups or ideologies. Much of the knowledge required to build this database is available from experts and researchers studying RWE, supplemented by additional research and prior experience.

In the present case, a database was developed containing more than 16,000 potential search phrases in English and French, including both Canada-specific phrases and phrases used widely elsewhere. Phrases covered a range of subjects, from Québec ultranationalist hardcore bands, like “Coup de Masse”, to acronyms and phrases like “LOTIE” (Lady of the Invisible Empire, to describe a female member of the KKK) and “Meine Ehre heißt Treue,” the motto of the Nazi SS, which is now used by neo-Nazi groups globally^{Footnote 26} . Each search phrase was coded according to a range of variables, including the type of content being sought, whether it included a specific group or ideology, and a risk rating based on the intent of the search and the level of prior knowledge required to perform the search. Searches that required substantial prior knowledge and/or convey an intent to act (for instance, join a group or commit an act of violence), received higher risk ratings. Searches that convey little more than curiosity received lower risk ratings.

Data was collected over a 30-day period from 18 December 2018 to 16 January 2019, resulting in 13,357 relevant searches from across Canada. Interpretation of these results requires caution: collection over a longer period will provide greater fidelity and enable longitudinal analysis. It is nevertheless possible to highlight some early findings on the themes, demographics, and regional distribution of RWE searches.

Initial Findings: Search Appetite for RWE Content in Canada

Canadians searching for RWE-oriented content are mostly male (67 per cent), and the most prolific searchers are between 25 and 34 years of age (24 per cent). The most common searches at every age are looking for RWE groups and influential personalities. Searches for groups are more prominent among younger users, while older users tend to search more for influential personalities. The trend away from groups and towards personalities increases steadily with age: while searches for influential personalities make up 30 per cent of searches performed by those aged 18 to 24, they make up 49 per cent of searches performed by those over 65. One possible interpretation is that younger users show more interest in joining RWE organisations, while older users show more interest in consuming online content produced by prominent RWE personalities.

Ideologically, Canadian users are mainly seeking neo-Nazi and white supremacist content, with a consistent habit of seeking out RWE bands and record labels. Neo-Nazi related search terms account for just over half (52 per cent) of all searches. While the most prominent search phrase related to neo-Nazism was “Heil Hitler” the next most popular neo-Nazi search term is “Vaginal Jesus”, which is a now-defunct white power grindcore band from Connecticut known for songs like “Happy Hanukaust” and “The Gestapo Stomp”. Nazi military paraphernalia also featured prominently, including the iron cross pin/badge/patch and SS Totenkopf, which could either refer to the SS unit responsible for administering the Nazi death camps, or to a Belarussian neo-Nazi band by the same name.

White supremacist-related searches, which made up just over a quarter of searches nationwide (28 per cent), also focused heavily on music. The most widely searched term in this category is “Graveland”, which is a Polish pagan metal band active since the early 1990s that has been banned by German authorities for promoting white supremacist rhetoric. A variety of searches related to Tightrope Records (an online white power retailer based in the US) indicate Canadians are seeking out their music and merchandise. Finally, a group of related searches for The Daily Stormer indicate that Canadians are trying to find this white supremacist web site, which has struggled to stay online after the 2017 Unite the Right rally^{Footnote 27} .

While nearly all of the most searched terms in Canada refer to non-Canadian content, the most searched for online personality is in fact Canadian: Evalion, a YouTube star made Internet-famous in 2016 for videos praising Hitler and instructing viewers on how to identify Jews^{Footnote 28} .

Nationally, three provinces account for roughly 70 per cent of all medium- to high-risk searches in Canada, and have the highest per capita search activity: Alberta (8.003 searches per 100,000 population), British Columbia (6.871 per 100,000 population), and Ontario (6.130 per 100,000 population). While Alberta has the highest per capita search appetite, searches from Ontario account for the largest proportion of searches in the highest risk category: searches that show an intent to commit an act of violence or join a violent group. Alberta accounts for nearly all searches (78 per cent) specifically seeking out membership in a violent group. The two most widely sought-out groups in Alberta are Atomwaffen and Soldiers of Odin. While the number of searches is relatively small, this trend merits continued attention.

While Alberta had the largest share of searches related to joining a group, Ontario’s high-risk searches are more focused on intent to commit violence, comprising 43 per cent of all such searches in Canada. The most targeted group in Ontario are Jews (23 per cent of searches), followed by blacks (15 per cent). Nationally, the most targeted group are homosexuals (25 per cent of searches), followed by Jews (22 per cent) and blacks (11 per cent). This is broadly consistent with data on police-reported hate crimes in Canada compiled by Statistics Canada, in which Jews, blacks and people targeted for their sexual orientation are three of the four most victimised groups^{Footnote 29} . Canadian Muslims are the fourth victim group targeted most often in Statistics Canada reporting, but searches targeting Muslims for violence are all but absent in search data, composing only 4 per cent of searches with violent intent. Given the prominence of Muslims as a target of hate crime in Canada, it is likely that searches targeting them will increase over a longer duration of data collection.

Conclusion: Building a Holistic Evidence Base to Counter RWE

Statistics Canada’s hate crime data showed a 47 per cent increase in hate incidents from 2016 to 2017, which is a sharp increase over previous years^{Footnote 30} . This statistic, coupled with the search traffic analysis above, demonstrates that the culture of fear, hatred and mistrust nurtured by the RWE has both online and real-world manifestations. Addressing the RWE problem in Canada will require concerted effort and a holistic evidence-base from both domains, based on a recognition that the online and offline worlds are mutually reinforcing.

While RWE activity in public online venues will continue to be an important component of this evidence base, search traffic provides a less performative, more intimate source of evidence for what users are seeking from the RWE online environment. This opens up a much wider range of possible responses, including many that can be implemented earlier and more constructively than traditional security responses.