Radical Engagement

Hypothesising a radical multi-platform journey

Team Members

[DMI] Jeroen de Vos, Nick Forrester, Daria Asmolova, Paola Verhaert, Lorenzo Piazzoli

[Police] Nico, Arthur & John

Contents

Introduction

Finding a focus

In order to understand the highest priority areas to look at, for the three members of the police who were leading this project, we mapped out the journey our experts believe that the radical Islamist message takes. This looked at the key actors, down the left hand side of the whiteboard and the key touch points along the top. It quickly became clear from our discussions that the most important area was when those sympathetic to radical Islamist content start to actively engage with the content online by posting themselves. Either side of these interactions the police experts described how messages are disseminated by the mainstream media on the one hand and active propagandists, often loosely connected to Isis, on the other.

Another important factor in settling on our research question was the objective of the police to better understand how radical Islamist messages are communicated online. The objective, to take a broad look at the forms of radical content online, would help us to resolve certain ethical issues further down the line.

Settling on a research question

When mapping out the journey, the platforms we were most interested in were Facebook, YouTube and Twitter, because our experts felt these were the platforms where those sympathetic to the radical Islamist message might encounter this kind of content.

We settled on the following question: The relation of language and distribution in the dynamics of platform-specific engagement with and exposure to radical content. More broadly our objective was to understand the hypothetical journey of radicalization through Youtube, Twitter and Facebook Identifying the distribution of local and global radical networks and main actors.

Ethical considerations

Discussed the police’s targeted approach in the light of the increasing scrutiny on the intelligence services gathering of data following the Snowden leaks.

The police brought with them data that we could use that would allow us to build on the intelligence they’ve built up over many years of professional service. This brought with it some troubling ethical problems. The police described their diligent and targeted approach to forming a watchlist, which is in start contrast to the bulk data collection of national secret services. However the police have the ability and often the desire to connect a digital identity to an offline identity (like a passport number or a home address), which meant that individuals identified during the research project might end up on a police list. Members of the group were uncomfortable with the thought that our work could result in a breach of individual privacy rights and might have unforeseen consequences further down the line.

Furthermore, the university's approach to gathering twitter data presents a wealth of personal data which the police could utilise, as one member of the police commented, “the university policy is much more invasive than that of the police”.

At this point, the police were keen to point out that it was not their desire to identify individuals from the work of the summer school. For this reason and for the ethical concerns of the group we decided upon two measures.

Firstly we decided not to look to find personal or profile data on individuals who post online content. We were more concerned with the content that’s currently out there and how the message is communicated and what are the characteristics of the network where these messages thrive. Secondly, we decided not to use the data the police brought that identified actors of interest to them. We instead used their expertise to identify significant content producers and media outlets and this was considered an equally effective way to understand the network.

Another overarching principle was to use open source data sources. This meant that all data used should be publically available and all targeting methods should not use information isn’t available online.

Research Questions

What is the relation of language and distribution in the dynamics of platform-specific engagement regarding exposure to radical content?

Aims:

• Understanding the hypothetical journey of radicalization through Youtube, Twitter and Facebook.

• Identifying the distribution of local and global radical networks and main actors.

Tools used:

• Netvizz

• TCAT

• YouTube data tools

Methodology

The methods employed in this research are operationalized through three sub-researches. Since specific content users have not been traced through either of the platforms, this research is considered a multi-platform research rather than a cross-platform analysis. Together, the three analyses could give an insight into the potential journey of a radicalising person by means of various types of media, messages, languages and localities. In broad strokes, the following section will explicate the multi-platform analysis in relation to a potential journey of radicalization in the Netherlands.

1. The Youtube network analysis was drawn based on a list of Dutch Youtube videos provided by experts, and a list of international videos, which were found during the research phase through online research on Youtube and Google. The analysis provides a video ecosystem that offers an exploratory insight into potential exposure to radicalised content.

2. A Facebook network of 5 radical groups, based on an expert list, shows radical content spreading in open Dutch Facebook groups. Since the Facebook specificities have a stronger invite to engage with the content, this network could be considered to portray a more engaging mode with radical content.

3. The Twitter analysis focuses on the discourse surrounding five international ISIS-related media outlets. An extensive co-hashtag analysis delves into the multilingual discourses surrounding those outlets with special attention to the recent Orlando shooting, which occurred on June 12th, 2016. Integrating the English and Arabic language, this network shows the content actively spreading and integrating in a large geographic scale.

Dutch Youtube video networks

Youtube is arguably the most mainstream video sharing platform and attracts users globally to share video content online. These users include radical jihadists, who exploit the mainstream video sharing platform as a portal for propaganda. The content shared by sympathisers of radical jihadism ranges from speeches to videos portraying extreme violence. While the latter are most often removed quickly by Youtube, radical jihadists continue to use the platform as a means of attracting new people to share their radical ideology. While groups like ISIS and Al Qaeda do not act as central actors on the platform, nor have stable accounts or channels, many of their followers use the platform to upload personal videos. Within this reality of highly accessible radical content, we aimed to research how Youtube users might become exposed to this content through the system of recommendation algorithm of Youtube. This exploratory analysis is not exhaustive, as we have had no access to the Youtube algorithm. We used the Youtube API to extract related content to Dutch videos which were flagged as radically jihadist. Using the Youtube data tool, we exported the data into Gephi, which we used to visualise the data, and derive findings on the network of Dutch radical jihadists on Youtube.

Questions:

• How are the radical islamic videos in the Netherlands connected in Youtube’s system of recommendations?

• Which topics are related to the radical islamic video cluster of the Netherlands?

Methodology:

• We started from a list of Youtube videos, which were provided by the experts, and were flagged as containing radical content.

• The videos in the provided list contained pro-ISIS messages, or expressed sympathy with Salafist groups in the Netherlands.

• The videos related to this list were derived using the Youtube data tool, and imported into Gephi.

• In Gephi, the videos from the list were removed as seeds in the data laboratory, as to be able to explore an unbiased network.

Configuration of the image at depth 1:

• Node colour: connectedness of video in network.

• Node size: general degree measure of video in network.

Configuration of the image at depth 2:

• Node colour: modularity measure, meaning measuring relatedness of content of the videos inside one cluster.

• Node size: view count.

International Youtube video networks

Popular international YouTube videos: the context

Our subject matter experts did not provide us with the list of internationally-produced videos, which were considered extreme and radical, so as a first step we decided to identify the most viewed videos with the interrogatory purpose of exploring the context around these videos.

• We took two keywords, jihad and mujahideen, from the keyword list provided by subject matter experts to search for videos with titles containing these words.

• The terms were chosen based on its broad use in the context of extreme and aggressive messages, but also, most importantly, these words are known to general public and are often associated with radicalism. Thus, we assumed that for someone new to the topic, the journey of looking for extreme content could start with these words.

• We chose to avoid the term ‘isis’ deliberately as there are too many news around this word, which would make our dataset too big to analyse.

• We picked top 5 videos for each keyword based on video view count. Mocking videos were skipped, because we weren’t going to explore the anti-radical sentiments.

• Please note that the top videos are not necessarily extremist, but are rather related to the subject of war / jihad in general. For example, we kept a video with nasheed to see whether it had any connection with the videos around conflicts.

• We used Netvizz’s Youtube Data Tools → Video Network → Crawl depth 1 (depth 2 would return too many results for videos with a total count of millions of views).

International Youtube videos and Dutch videos: is there a connection?

The next question we asked ourselves was about the connection between international videos and Dutch videos that were considered extreme: was it possible for someone starting the search on jihad and mujahideen to find propaganda videos in Dutch?

• We took the same ten videos from the previous query and added ten Dutch videos provided by the subject matter experts.

• The second query took ten ISIS-related videos recently removed from Youtube and ten Dutch videos from the list of subject matter experts.

• We also set crawl depth to 1 as depth 2 would generate too much data.

Radical ISIS / ISIL extremist videos: What is the context?

As we had researched the context around radical Dutch videos, we also wanted to explore the context of the extreme ISIS / ISIL videos.

• From the website, which tracks Islamists propaganda daily, posting links to their videos and other communication pieces, we chose ten recent Youtube videos, which were deleted as they violated Youtube use rules. We assumed they contained extremist propaganda.

• As there wouldn’t be many video views (such videos are deleted within hours), we chose depth 2 to analyse this video network.

Mainstream / alternative media and extremist videos: Is there a connection?

The last search of this section was about investigating the potential connections between news and documentaries about radical Islamists and videos that contain violent propaganda. In brief, is it possible for someone who starts with viewing professional coverage of ISIS-related issues to stumble across banned videos?

• For this search, we narrowed down our list of recently deleted videos to five and added two most viewed videos about ISIS published by Al Jazeera and Vice this year.

• We chose crawl depth 1 to process the data, because depth 2 would return too many results due to the media videos on the list.

Dutch Facebook Groups / Communities

Facebook is known for its global social sharing functionality where users can interact with each other and share content they want their friends or the public to know about. When it comes to online radical jihadism, Facebook has been in the spotlight a number of times. Almost all major radical jihadi media outlets have Facebook pages, communities and groups. Historically, several known radical jihadists had Facebook user accounts. In the French and Belgian terrorist attacks journalists found that several of the attackers had active Facebook accounts from which they interacted with other (potential) radical jihadists and participated in Facebook communities, pages and groups. With this knowledge in mind, subject experts provided specific (inter)national Facebook Groups & pages/communities, which they derived from general (public) media sources. These groups were chosen since they were known to be active in relation to radical jihadism, but not under the active police investigation. By scraping these sources with Netvizz we wanted to see how the groups were connected with each other. After scraping these groups and pages, they were visualized with Gephi. Within Gephi the data was separated by user activity and post activity from which insights might be seen on how these groups are connected with each other.

Questions:

• How are the (inter)national Facebook groups & pages/communities connected?

• What is the difference in connectivity between groups versus pages/community connectivity?

Methodology:

• Started from a list of Facebook Groups provided by experts. In our case the research was done on the base of five data sets.

• The sets contained: one community page and four group based pages.

• The community group was built around a specific subject connected to radical jihadistic fighters. While the other four groups were more salafistic orientated.

Configuration of the image:

• Node size: in-degree

• Node colour: modularity

International Twitter network: bilingual content analysis of ISIS Media Outlets

Summary of the queries:

isis (AjnadMediaProduction, AlBayanRadio, AlFurqanMedia, AlHayatMediaCenter, AmaqAgency, Dabiq, IslamicState, أخبار_الخلافة, إذاعة_البيان, الدولة_الإسلامية, النشرة_الإخبارية, تقارير_الولايات, دولة_الخلافة, دولة_الخلافة_الإسلام, قناة_ناشر, مؤسسة_الفرقان, مركز_الحياة_للإعلام)

Date range chosen:

15 - 21 May 2016 - peak of activity observed in this week

7 - 14 June 2016- fewer videos posted on youtube

We originally wanted to map Twitter to YouTube to identify trends in the network.

Our approach to the data and what we hoped to get out of it

The data has been analyzed in using a co-hashtag network. This method has been used to better understand the relation between different topics especially when arabic and english hashtags were used simultaneously.

Findings

Video content: YouTube issue findings

Videos have been long used as one of the main means to deliver a message, whether for individuals, brands or organizations. The popularity of this format continues to grow and, according to Cisco, “Internet video streaming and downloads are beginning to take a larger share of bandwidth and will grow to more than 80 percent of all consumer Internet traffic by 2020”.

It’s no surprise that platforms, which were originally designed to connect people by more traditional formats of communication like texts, image and link sharing, have started offering video services as well. Facebook and Twitter now have a function of direct video uploads, and an encrypted message-sharing apps like WhatsApp and Telegram facilitate video sharing as well.

While extremists have been using different platforms to spread their propaganda and recruit people for years, it’s particularly the ISIS group that has become successful in using digital channels, especially social media. Their tactics is professional from the content creation to distribution stage, with many experts trying to understand what makes ISIS successful in online propaganda.

As YouTube is still the biggest video sharing platform with the number of people watching YouTube each day increasing significantly over the past couple of years, we assume that the video sharing is not moving away from YouTube, but rather throwing more channels into the mix. Hence, trying to understand the connections and context of the extremist-related videos on YouTube, can help us find clues to answer the key questions of this research.

Dutch YouTube videos: issue findings

Dutch radical video network at depth 1:

The list of radical jihadist videos from the Netherlands, provided by the experts, was used as a base point to derive a larger Dutch network of related content. The most important findings are:

• The most connected videos in the main cluster is related to Dutch informative videos on ISIS, which do follow a divisive right-winged discourse.

• Left to the centre of the main cluster, the discourse changes to Dutch tales of martyrdom, and the benefits of dying in battle.

• On the right side of the main cluster, connected to this content, are videos on the extreme-right Dutch politician Geert Wilders, talking about Islam and jihadism.

• Salafist content related to the Hizb ut Tahrir grouping in the Netherlands is related to the network, but is not central to it.

• While the network contains content which both strongly opposes and sympathises with radical jihadism, it is interesting to mark that both extreme position are located in the same network.

Image 1: overview of the Dutch Youtube video network around radical jihadist content.

Image 2 (left): “How the propaganda machine of ISIS is conquering the Netherlands” as highlighted node / video.

Image 3 (right): “Martyr at the front line” as highlighted video / node.

Image 4 (left): “Geert Wilders on jihadism (debate)” as highlighted node / video.

Image 5 (right): “Street interviews in The Hague: Campaign It is time for Khalifah Hizb ut Tahrir the Netherlands” as highlighted node / video.

Dutch radical video network at depth 2:

This network was derived from the same video list provided by the experts. The depth from the seeds was set at 2 this time, resulting in a database of approximately 9400 videos. The bright red cluster represents the Dutch video network from the Netherlands. From this network, we were able to represent the following findings:

• We can see that the Dutch videos are very embedded into and intertwined with a network of surrounding clusters. The central content of the red cluster is Dutch videos on ISIS and jihadi fighters who have left Western Europe. The discourse in the centre of the red cluster is fairly neutral, and connects to mainstream news outlets of the neighbouring blue cluster.

• To the left of the centre of the bright red cluster, the videos relate to travel stories to Syria, some convert stories, Ramadan, and martyrdom. The titles of the videos are more radical than those in the centre of the bright red cluster.

• Towards the right of the centre, the red cluster starts intertwining with the green cluster, which contains videos about right-wing radical parties in the Netherlands. This part of the red cluster is anti-ISIS, and anti-Muslim.

• The green cluster is a political cluster, and contains videos on Geert Wilders, the PVDA, and PVV, related to jihadism and radical Islam.

• This part of the bright red cluster, as well as the green cluster, are intertwined with the purple cluster, which contains videos of Dutch convert stories. To the left of the centre, the purple cluster contains more videos of female convert stories, and stories of Ummahs.

• The mustard cluster to the bottom of this set of clusters contains videos that feed into the divisive discourse of west and east; here, we can find a lot of videos about discrimination of Islam in Holland, and “time for Khalifah” videos. The larger mustard cluster at the bottom contains a lot of videos on the Salafist Hizb ut-Tahrir in the Netherlands.

• The magenta cluster next to it, tells a similar story, but with “softer” titles. They are emotional video titles, urging people to choose the path of Islam.

• The large blue cluster contains videos in English, mostly from mainstream media outlets in the West, and mostly documentaries.

• The dark green cluster contains nasheeds.

• The dark red cluster on the left contains Qur’an recitations, some in Arabic and Urdu; this is the only cluster containing videos titled in Arabic.

Image: Youtube network around radical jihadist videos in the Netherlands, at depth 2.

Image (left): “My son left to Syria for jihad - RTL NEWS” as highlighted node / video in bright red cluster.

Image (right): “Wilders: IS declares war to the Netherlands” as highlighted node / video in light green cluster.

Image (left): “Dying As A Muslim (Greatest Gift of Allah) - HD” as highlighted node / video in magenta cluster.

Image (right): “Dutch women convert to Islam” as highlighted node / video in magenta cluster.

Image (left): highlighted node / video in light purple cluster.

Image (right): “Interview with Okay Pala Together against anti-Islam policy - Hizb ut Tahrir the Netherlands” as highlighted node / video in mustard cluster.

International YouTube videos: issue findings

Popular International Youtube videos at depth 1:

• As the videos were chosen by view count, those related to Taliban, a longer-standing issue in international conflicts compared to Syria, had larger network of connections.

• Taliban / Afghanistan cluster is relatively isolated from the Syria-related videos.

• Even though some videos about Chechen fighters were produced before the war in Syria broke out, overall, this content has many connections to videos about Syria. The reason could be a growing involvement of this region in supporting ISIS.

• There were also Indonesian and Albanian videos in the mix, which might support the argument of this conflict being used by ISIS to unite fighters from different regions.

• The link to Arabic videos is the weakest. As none of us could read in Arabic, we did not understand the titles of these videos, but based on the language alone, videos in Arabic seems to be isolated compared to other languages.

• Nasheed cluster also has weak connections with the war-related videos. However, there is one big ‘traveling’ hub, a video about what jihad is, that channels viewers to war-related videos, proving, once again, that generally jihad is associated with war and not known for its original concept of spiritual struggle with yourself.

Image: video network around popular international videos

International Youtube videos and Dutch videos: is there a connection?

Findings:

• In depth 1, Afghanistan / Taliban / Chechnya have no visible connections to Dutch videos.

• Extreme ISIS videos also have no connection to the Dutch videos.

• While Dutch videos related to Syria were scattered in different clusters, they were not connected to the similar content in English or other languages. This might be caused by the different methodologies chosen to create the lists for international and Dutch videos.

Image: popular international videos and extremist Dutch videos

Image: radical ISIS videos and extremist Dutch videos

Radical ISIS / ISIL extremist videos: What is the context?

Findings:

• The video network around deleted videos is split into many different groups, which predominantly contain videos in Arabic.

• French language appears in at least three clusters, while at least one cluster consists of videos with Turkish titles (mentioning the word ‘kurds’).

• From the video titles, we were able to derive mentions of countries like Algeria, Morocco, UAE and Lebanon, but the Netherlands haven’t appeared in this search.

• There were also a number of clusters related to popular culture and lifestyle, including mbc pro sports, extreme sports like parkour and skateboarding, iPhone, Coca-Cola, Egyxos (TV series) and more. We may assume that it shows that those who upload / watch radical videos also use Youtube to discover content for leisure.

Image: video network around radical ISIS / ISIL extremist videos

Mainstream / alternative media and extremist videos: Is there a connection?

Findings:

• Videos in Arabic are split into 3 clusters, which have very few connections to media-produced videos like documentaries and news stories.

• While both Al Jazeera and Vice videos were related to ISIS, a new cluster on Israeli-Palestinian conflict emerged.

• The were almost no links between media videos and radical content.

• The closest link between media-produced videos and extreme content in Arabic were a couple of links between a video about the Israeli strike on Gaza and videos in Arabic (at the top).

• It should be noted that in some media videos links to more radical content are shared by the users in the comments section.

Image: video network around media-produced content and ISIS / ISIL extremist videos

International Twitter network: bilingual content analysis of ISIS Media Outlets

Methodology

Our analysis of Twitter is based on a co-hashtag analysis over to distinct date ranges. The first date range, 15th-21st May 2016, was selected because there appeared to be a significant spike in activity during this time. The second date range, 7th-14th June, was selected because of the fierce fighting that was reported in the news in Raqqah, Manbij and other prominent towns in Syria.

The subsequent data visualisations take into account three variables from the co-hashtag data.

1. Population diversity - which is a weighting to identify the number of people using this hashtag in relation to the number of uses of the hastag. The scale is from red to blue. Blue being high population diversity (lots of people using the hashtag) and red for low population diversity (a relatively small number of people using the hashtag in relation to the frequency with which the hashtag is used).

2. Co-hastagging - this refers to the number of times hashtags are used together in the same tweet. Each mention two hashtags have together in the same tweet increases the strength of the connection between nodes.

3. Hashtag frequency- each node represents a hashtag and the size of that node refers to the number of times that hashtag was used in our dataset.

Language divide is connected by key actors

Configuration of the image:

Co-hashtag networks for the two weeks.

The strength of the connections depends on how many times two or more hashtags are present together in the same tweet. The color of the node is represent the user diversification: red for a low user diversification and blue for high.

Image: twitter network mapped in Gephi comparing a date range in May with one in June

Findings:

Languages are clearly split apart, but significantly more so in the first week than the second. There are some key actors and events that bridge the divide between languages, such as Dabiq (a magazine that publishes ISIS-related content and the SDF and YPG which are connected to the Kurdish army fighting in Syria. We found it interesting that these issues related to these actors might be bridging the divide between languages.

A closer look in the news found that there was fighting in Syrian cities between the Kurdish forces and others and this seemed to be of particular interest to English language tweets, whilst also bridging the gap between the two languages. Another major event which bridged the gap were the Orlando shootings, which happened in our second date range. We chose to focus on the second week because this could be our window into the events of the second week.

issues are connected differently in both languages

Configuration of the image:

This are two zoom in the co-hashtag network of the second week (7-14 June 2016).

Strength of the connections and colours have the same meaning as in the preview image.

The hashtag related to “Orlando” are underlined both in the Arabic cluster (left) and in the English one (right).

Image: network detail highlighting hashtags connected to Orlando shootings in Arabic and in English

Findings:

In Arabic the hashtag “Orlando” gets adopted into the narrative around the key battlegrounds in Syria, such as Raccah, Fallujah, Manbij, Islamic State (are all quite closely connected to Orlando).

In English Orlando is drawn into this narrative, but we see a version of the program Vs anti-program. Orlando is connected to the English translation of the key battlegrounds we see in the Arabic, but also strongly connected to hashtags like “prayfororlando” and “islamistheproblem”.

Other international events might be hijacked by ISIS actors

Configuration of the image:

Co-hashtag network for the second week (7 - 14 June 2016) with underlined unexpected topics that appear in the English cluster. The strength of the connections and colours have the same meaning as in the first twitter image.

Image: twitter network highlighting apparently unrelated events that appear in the network

We see how seemingly unrelated topics enter into the network, but we’re not totally sure why and our experts have pointed to the possibility that events are hijacked. This diagram points to events like the opening ceremony of Euro 2016 “showeuro2016” or the Lenovo Tech World event “lenovotechworld”, which become part of the network. The fact that these have a low population diversity may support this suggestion that there is an attempt to hijack these events by actors related to and supporting ISIS.

Dutch Facebook groups/communities: issue findings

The collaboration between the police task force brought us to the question what data is shareable? Starting with the Facebook groups that were openly sourced made us run into a couple of problems. First of all, The design of Facebook makes it hard to extract data. The API’s have been changed a lot and there is a difference between the kind of pages. Secondly, there is a distinction between Facebook profiles, Pages and Groups and Facebook Community Pages. And while Nevizz allows scraping pages and groups, the 'communities' were only sometimes scrapable as a 'page'. The implications of this distinction need further investigation. Moreover, there seem to be differences in functionality the community pages are seem to look more like Wiki pages. It more a platform that is been used where on messages are been posted and advertised. The community page seems to give an “official” status to a page.

The interaction between users (pink) and posts (green) of one of the most active radical Facebook pages. It shows a selection of about 10 people affiliate with many pots, where the page does have a larger actively engaging audience.

Scraping the six most active Dutch Facebook pages, showed there is some cross-page activity. Where for some pages the content is clearly driven by a select group of people, others are more diversely connected. When looking into the Gephi data visualisation you can see all groups/communities are connected. Specifically, they are mostly connected by certain users who interact within all groups and seem to be the thriving factor that make/keep these groups connected.

Limitations: Analysis of other the entire expert list of public Facebook pages shows, the facebook groups/pages seem do not have very much user activity and could be labelled as a platform for radical jihadist lurkers.

Suggestion for further research: Scrape similar international Facebook groups & communities/pages who support radical jihadism. Compare/combine them with Preliminary findings.

Future difficulties: In June 2016 the European Union (EU) announced a new online speech code to be enforced by four major tech companies, including Facebook and YouTube. That means that it will no longer allow hardcore or graphic radical content in general. Algorithms will delete or suspend this content as soon as possible. This all is part of a global notice and takedown initiative by the major tech companies.

Discussion

In this research, we have made some deliberate choices in our operationalisation of the sub-researches. The collaboration with the police led to the productive discussions on issues of privacy for both research subjects and researchers, on the nature of open source intelligence, the shareability of sensitive data, and the lack of scrutiny of research methods to some extent. This section will shortly discuss the implications of the choices we made throughout the research, and will end with an evaluation of the results of our collaboration.

The expert lists provided by the police have been rendered based on general information which is openly available. In addition, any Facebook group or Youtube channel listed was not currently under active investigation by the Dutch police. This makes the expert lists compliant with the criteria for open source intelligence. The entries have been used as a starting point for exploration of content over individuals. This implies that the explicit aim of this research was not to frame new individual targets, but rather to establish a better understanding of the modus operandi of potential radicalisation within Dutch context specifically. The line between visualising an online discourse or activity would sometimes get close to individual people, as is the case with the Facebook network analysis. But since the Netvizz tool anonymizes the results, we chose to look at the activity and spread without including unnecessary details to the graphs published on this wiki. Finally, the collaboration between researchers and subject experts appeared of great value. While we had the time to explore the data, Nico, Arthur and John were quick to help interpret and explain the results (as well as their limitations).

This research evolved around the question 'What is the relation of language and distribution in the dynamics of platform-specific engagement regarding exposure to radical content?'. It has been operationalized through the hypothetical journey of radicalization. The Youtube network analysis shows how Dutch clips with radical content are closely intertwined with more local political videos. It shows the films from the expert list do not exist in a vacuum but are embedded into the local Dutch sociopolitical context. At the same time, these videos are not actively integrated into the extremely radical videos network. The preliminary Facebook findings indicate that some groups are disengaged and rely on more traditional broadcast models, whereas others have active engagement. Tapping into the larger international debate that comprises ISIS related issues and news hijacking shows the jihadist radical discourse to be deeply diverged over both Arabic and English, in which some topics will exist in both discourses in a constellation, some actively integrate the two and others are native to one language alone.

References

1. Cisco Visual Networking Index: Forecast and Methodology, 2015–2020 http://www.cisco.com/c/dam/en/us/solutions/collateral/service-provider/visual-networking-index-vni/complete-white-paper-c11-481360.pdf

2. Why ISIS Is Winning the Social Media War https://www.wired.com/2016/03/isis-winning-social-media-war-heres-beat/

3. YouTube Statistics https://www.youtube.com/yt/press/en-GB/statistics.html

4. Chechen volunteers fighting ISIS in Syria and Iraq: Report http://europe.newsweek.com/chechen-volunteers-fighting-isis-syria-and-iraq-report-424525?rm=eu

5. The Facebook jihadists They may not be violent but they are still dangerous http://www.al-bab.com/blog/2016/feb/facebook-jihadists.htm

6. Jihad Selfies: These British Extremists in Syria Love Social Media http://www.vice.com/read/syrian-jihadist-selfies-tell-us-a-lot-about-their-war

7. Jihadist Kills Cops While Broadcasting Live Over Facebook http://dailycaller.com/2016/06/14/jihadist-kills-cops-while-broadcasting-live-over-facebook/