Blogs of War

Hizballah Cavalcade

Internet Haganah



Kremlin Trolls

Making Sense of Jihad

Selected Wisdom

Views from the Occident


American Terrorists

Anwar Awlaki

Al Qaeda


American Al Qaeda Members

Inspire Magazine

Revolution Muslim


News, documents and analysis on violent extremism

Saturday, March 30, 2013

On Comparing White Nationalists to Anarchists

On Thursday, the International Centre for the Study of Radicalisation published Who Matters Online: Measuring Influence, Evaluating Content and Countering Violent Extremism in Online Social Networks, a research paper by myself and Bill Strathearn, which examined a substantial amount of data on the nature of Twitter networks used by extremists.

The paper was originally intended to focus primarily on metrics, and it originally set out to examine followers of prominent white nationalist Twitter accounts. However things got complicated when we saw what the data said about hashtags and links used by members of the dataset. The short version of this, as I explained for Foreign Policy:
The number 1 hashtag used by people following white nationalists online was #tcot -- "top conservatives on Twitter." Number 3 was #teaparty, and number 5 was #gop. The content of links tweeted by the users also skewed sharply toward mainstream conservative and Republican content. ... [A]s of last week, the most influential users from the group we studied were still heavily engaged with partisan Republican politics, based on an examination of the 50 most recent tweets from each of the 100 most influential users still active on Twitter. The top hashtag remained #tcot. Coming in at number 3 was #cpac2013, and #gop was number 4. The tag #nra vaulted from 102nd place in last year's data to number 6, no doubt due to recent discussions on gun control, while #teaparty dropped to number 10.
One of the joys of research is discovering things you didn't know before, and this project was full of such moments. It's not that I was particularly surprised that there would be Republican-related content -- there are legitimate reasons why we call some kinds of extremism right-wing and others left-wing -- but I hadn't expected that content to be so prominent. Like many others who study extremism, I was focused on the ways extremists are disenfranchised from the mainstream, rather than how they might seek to engage it.

When we saw the findings from the first round of data collection, we could have just left the statistics in there without further comment or context, but that felt like a cop-out and it probably wouldn't have lessened what we expected would be some controversy. So the question then became how to offer more context without straying too far from the point of the paper, which was to develop and evaluate metrics for processing large amounts of social media data.

To that end, we performed a secondary analysis that looked at the followers of anarchist Twitter accounts. This has prompted some objections that a comparison between white nationalists and anarchists is fundamentally flawed or unfair, and that a comparison to black nationalists would have been more appropriate both for a one-to-one comparison as well as for the purpose of offending everyone's politics equally.

These objections are not unreasonable, especially from the perspective of the people making the complaints on both the right and the left who are primarily concerned with the politics of the paper rather than the metrics. But I think we made the right decision given the paper's actual focus, and I want to explain why we did what we did in a little more detail. The most important reason is one that should be obvious if you read the paper but may be less obvious if you read the executive summary then skipped to the part that offended you, or just read news coverage of the paper and not the paper itself.

The paper is not primarily about politics. 

The sections concerning hashtag and link content comprise about one quarter of the total paper. The goal of the paper was always to evaluate metrics and develop techniques for sifting large datasets. Of the 25 percent that concerns politics, probably a third of that consists of disclaimers. The political nature of the findings was unexpected; if the content data had been uncontroversial, the hashtag and link content would have been even less than 25 percent.

We had the resources to analyze one additional group of accounts. So when we thought about a comparison, it had to primarily serve the paper's primary goal -- testing the metrics. In this respect, anarchism is one of the toughest challenges we could devise, for the reasons outlined in the paper. We felt that if the metrics performed well on anarchism, they would perform well on almost any kind of ideology.

While white nationalism has many different facets, it has a strong center of gravity around the relatively simple concept of excluding non-whites. Anarchism, in contrast, is by definition diffuse and unorganized, encompassing a wide variety of views and attitudes and clustering much more loosely in terms of charismatic leaders who attract "followers" in the non-Twitter sense. The metrics had performed extraordinarily well at surfacing people who were ideologically engaged with white nationalism. If they performed well on anarchists, it was a much stronger proof of concept.

To be clear, this wasn't just about the metrics. We did also want to provide some balance to the political content from the left (and I honestly expected to see more engagement with mainstream Democratic politics despite the definitional bias of anarchism against political parties).  But the metrics had proved to be unexpectedly powerful, which also has a lot of complicated implications.

If we had chosen black nationalism as the comparison group, I am virtually certain we would have succeeded at a tit-for-tat outcome of offending all mainstream political groups equally, but I am also reasonably certain it would skewed the metrics in a way that exposed us to criticism of inflating our success rate, because white nationalism and black nationalism both have the same strong center of gravity -- race.

Anarchism is an ideology that was both politically and structurally different from white nationalism. And the results were illuminating; they showed that the metrics performed less effectively on anarchism but were still effective enough to suggest that the metrics were doing exactly what we thought they were -- surfacing people engaged with the seeded ideology from a large dataset.

If our goal was only to provide a political tit-for-tat, mentioning Democrats as often as it mentioned Republicans, then black nationalism would have been the right choice. But the goal was to provide a more robust and informative overall view of how our techniques worked, and anarchism was the right choice for that purpose.

There were two other significant considerations:

  • Extremism is an extremely diverse topic, and there are limits to my ability to deeply familiarize myself with every type. I know somewhat more about anarchism than black nationalism, partly because it seems to me the former movement is more robust at the moment. So I could more realistically identify seed accounts and evaluate the accounts identified by the metrics. 
  • Anarchists are much more effective online than black nationalists, within the scope of what I have been able to examine, and the paper is about online activism. 
I was asked why we included the lists of accounts for white nationalists but not the accounts for anarchists. The main reason is that white nationalists were the primary case study, and anarchists did not receive the same degree of analysis. Secondarily, as noted in the paper, there is a much lower social stigma and much lower threshold for casual involvement with anarchism. If someone uses a swastika and "1488" in their Twitter profile, I don't feel like I'm "outing" them as a white nationalist and an extremist. The privacy factors are more complicated for anarchists and I'm not comfortable throwing out lists of accounts on that basis, even after considering the complaints, which do deserve consideration. 

The nature of the anarchist vs. white nationalist comparison is the most serious and thoughtful complaint I've seen about the paper, but there are a couple other comments and interpretations I've seen circulating, which I'll address only briefly. 
  • Anarchists aren't "violent" extremists.  The paper itself concerns only visible online engagement with the seed ideology. It does not address the risk of violence by social media users, and it does not characterize any user as prone to or involved with violence. Only a very small percentage of extremists of any ideology are violent, whether they're anarchists, white nationalists or anyone else. The paper is looking at extremist ideologies and not violent offenders. The V in CVE -- countering violent extremism -- is a misnomer that runs through the whole field of CVE, which I've addressed and criticized in the past. That said, you can certainly find violent anarchists, but in this country, anarchist violence is relatively rare and usually confined to property crimes. I could have made this clearer in the paper, and maybe I should have. I'm saying it now. 
  • From the right, The data is "made up" to "smear" the GOP and Tea Party. From the left, The data shows Republicans are racists. It's not, and it doesn't. I wasn't thrilled to discover that we had stepped on a political landmine, but once we did, we had little choice but to let it blow up. Anything else would have been dishonest. The paper clearly and repeatedly states that this data in no way characterizes Republicans or conservatives as racists, but it does characterize ideological racists as being very engaged with mainstream conservative politics. If you don't understand the distinction, I suspect there's a good chance you don't want to. 
  • This research is about Big Brother/government surveillance squashing free speech. The government had no role in sponsoring this research, nor did it have any input into the project. This was always done with private activism in mind, although I'm not an idiot and I understand how government might try to use this technique. If it's upsetting to you that Big Data can identify your political ideology, it should be, but private companies and likely the government are already doing this kind of analysis on you on a daily basis if you use social media. At least with this paper, you can read a transparent account of how it's done, and in what ways it succeeds and where it doesn't. And if you've ever signed up for Klout or Kred, you have no business being upset about this kind of analysis.  
There are a lot of other complicated issues arising from this research, and I'm not going to try to address them all here today. It's a lot to take in. I expect we'll be talking about these issues for some time to come, and on Twitter, and on this blog in the weeks and months ahead. I welcome the opportunity, but please be patient as I pace myself. 

Buy J.M. Berger's new book, Jihad Joe: Americans Who Go to War in the Name of Islam, on sale everywhere.


Labels: , , , , , ,

Views expressed on INTELWIRE are those of the author alone.



Tweets referencing this post:



", granular analysis..."

ISIS: The State of Terror
"Jessica Stern and J.M. Berger's new book, "ISIS," should be required reading for every politician and policymaker... Their smart, granular analysis is a bracing antidote to both facile dismissals and wild exaggerations... a nuanced and readable account of the ideological and organizational origins of the group." -- Washington Post

More on ISIS: The State of Terror

"...a timely warning..."

Jihad Joe: Americans Who Go to War in the Name of Islam:
"At a time when some politicians and pundits blur the line between Islam and terrorism, Berger, who knows this subject far better than the demagogues, sharply cautions against vilifying Muslim Americans. ... It is a timely warning from an expert who has not lost his perspective." -- New York Times

More on Jihad Joe


INTELWIRE is a web site edited by J.M. Berger. a researcher, analyst and consultant covering extremism, with a special focus on extremist activities in the U.S. and extremist use of social media. He is a non-resident fellow with the Brookings Institution, Project on U.S. Relations with the Islamic World, and author of the critically acclaimed Jihad Joe: Americans Who Go to War in the Name of Islam, the only definitive history of the U.S. jihadist movement, and co-author of ISIS: The State of Terror with Jessica Stern.


Newest posts!

Best-Liked on the Most-Wanted

A Closer Look at the SPLC's Hate List

Inspire 10: Still sucks

Bum Rap

Should we disrupt terrorists on the Internet?

Former friend rips U.S. jihadi Hammami in online p...

Omar Hammami and the Trolls of Jihad

Review: The Last Refuge

FBI Releases Redacted Informant Manuals In Respons...

Signs of Life (Maybe) from Omar Hammami


New York Pipe Bomb Suspect Linked to Revolution Muslim

The Utility of Lone Wolves

Interview with Online Jihadist Abu Suleiman Al Nasser

A Way Forward for CVE: The Five Ds

How Terrorists Use The Internet: Just Like You

PATCON: The FBI's Secret War on the Militia Movement

Interview About Jihad With Controversial Cleric Bilal Philips

Forgeries on the Jihadist Forums

U.S. Gave Millions To Charity Linked To Al Qaeda, Anwar Awlaki

State Department Secretly Met With Followers of Blind Sheikh

State Department Put 'Political Pressure' On FBI To Deport Brother-in-Law Of Osama Bin Laden In 1995

FBI Records Reveal Details Of Nixon-Era Racial Profiling Program Targeting Arabs

Gaza Flotilla Official Was Foreign Fighter in Bosnia War

U.S. Had 'High Confidence' Of UBL Attack In June 2001

Behind the Handshake: The Rumsfeld-Saddam Meeting