The nature of the mobile-enabled swarmcast means it may appear to be degraded, but it has really only reconfigured.
This observation made in 2014 was based on earlier studies of Jabhat al-Nusra, ISIS and the online activity of the wider Jihadist movement – produced at a time when some commentators already claimed ISIS capacity on social media had been ‘degraded’.
The ‘bluckling’ of the ISIS media system, much talked up by decliners at the end of 2017, and presented as “not just a media decline—it is a full-fledged collapse” will, in 2018, turn out to be to have been a lull as the swarmcast reconfigured, rather than signs of collapse.
This post shows:
- How 6th grader math class could explain what is wrong with the current decliner logic and their approaches to quantifying output.
- Why decliners ignore 2016 – and why 2016 destroys the claim of a pseudo-correlation between content production and area controlled.
- What about 2014? – how 2014 further disrupts decliner logic.
- A thought experiment – what would decliner logic predict about the direction in volume of social media Global Coalition?
War by the numbers
Wars often turn on numbers, data, and the real-life experiences of the individuals represented by those numbers. The habitus of those interpretting information about the war also influences how it is understood. This includes the lived experience of target audiences in the physical world and virtual domains. As Douglas Rushkoff predicted, the battle for reality continues online.[i]
In this vein the Global Coalition information operations have long sought to present ISIS or Islamic State as being in ‘decline’ or weak. This has included tweeting images which question the ability of ISIS to produce media content and presenting this as a measure of their decline.
While the tactic is not new – it is currently recognised by ISIS and earlier versions were discussed in depth by Anwar al-Awlaki – there is risk in confusing wishful thinking, the Information War, and academic study.
For example, ISIS has launched a new magazine al-Anfal, now on the 7th issue and the newspaper al-Naba is now on issue 116. It is noteworthy that al-Anfal is vastly better known by those in an Arabic habitus than by English and faux-Arabic language commentators – for whom it has barely warranted a mention.[ii]
6th grade math class – A holiday anecdote
The holiday season is often a time when people encounter others with different perspectives and experience. A maths teacher I spoke with highlighted why, the even in an information war, accurate data work is important to avoid becoming transparent propaganda.
In response to hearing that claims of decline are often calculated by counting the pieces of content irrespective of type, content, or length, the maths teacher originally responded; “that’s just stupid”. He was ready to dismiss the premise that anyone would count an hour-long video, a newspaper, a speech from the leader, an infographic and a picture as if they are all the same.
His reasoning was persuasive, he said;
If I set five assignments, then ask two students to hand in their homework
- Student A hands in two of the assignments each on a separate sheet of paper
- Student B has done all five assignments in his book & hands in the book
Student A has not done twice as much work as Student B because they handed in two sheets while Student B only handed in one book. The maths teacher concluded, I’ve taught a number of 6th graders that would have a problem with that logic.
Anecdote aside, there is a problem trying to build credibility with an audience using an approach which infantilises the intended audience with an argument which 6th graders can see through. This is the 6th grader problem.
Despite the clear limitations in the data, (for example, previous research has proven ISIS weekly production exceeded the CTC estimate of monthly production. (New Netwar p. 38) ) some have claimed that “the destabilization of the Islamic State’s territorial strongholds is correlated to a decrease in the volume of media production”.[iii] The problems with this way of thinking are numerous, of which the two most prominent are;
- There is no calculation of correlation in the published research, so it is an assumed pseudo-correlation which is not based on any demonstrable relationship in the data.
- The apparent relationship between territory and production can only be maintained by cherry picking certain points in time – entirely ignoring 2016. In fact, the entire decliner ‘narrative’ relies on ignoring the fluctuations in content during 2016.
Research has already shown why the summer of 2015 is convenient to cherry pick as a start point if your goal is to claim decline. As discussed in depth elsewhere, picking a time which logic dictates is most likely to be an outlier can make for a nice soundbite for coalition propaganda, and may even sound smart in 20 seconds of air time, but does not work for an authentic understanding of the movement.
The root of the problem, as Reuven Paz noted in 2007,
Jihadi militancy is … almost entirely directed in Arabic and its content is intimately tied to the socio-political context of the Arab world.[iv]
People who live in that socio-political context, or habitus, easily pick up on the factors likely to play into the spike in content over the summer 2015.[v] This is because;
the habitus is itself a generative dynamic structure that adapts and accommodates itself to another dynamic mesolevel structure composed primarily of other actors, situated practices and durable institutions (fields).[vi]
And because habitus allowed Bourdieu;
to analyze the social agent as a physical, embodied actor, subject to developmental, cognitive and emotive constraints and affected by the very real physical and institutional configurations of the field.[vii]
While it is tempting to focus on Western interpretations of Jihadist content, particularly if you can only draw on faux-Arabic, the habitus of the intended audience has to be foremost when analysing the meaning of the content.[viii]
Why decliners ignore 2016
Most of the recent claims of decline, including those by the Global Coalition, pick 2015 as the start point, then jump over the awkward hurdle of 2016 and head straight for 2017. This allows them to get past the fluctuations in 2016 and present “huge and steady decrease”.
To understand how this works, first it is important to understand the 2015 ‘highpoint’, from one of the often-referenced studies about 2015.
Journalists writing about the 2015 ‘highpoint’ have claimed ISIS “was producing more than 200 videos, radio programmes, magazines and photo reports each week” resulting in “just under a thousand unique data points, ranging from radio bulletins and electronic magazines to videos and photographic essays”[ix]
Perhaps not immediately clear to the reader;
- “Just under 1000 unique data points” is in reality 892. This makes the ‘highpoint’ sound 10% higher than the reality. 892 could be described as just under 900 … but it is just under 1,000 like I’m just under 8 feet tall.
- Composition of the content is around 80% pictures (just under 700 of 892 data points).
Even for the astute reader it can be hard to tell what is being counted. As best as can be ascertained from the terms ‘events’, ‘photo reports’, ‘photographs’ and ‘photos’ being used interchangeably, this is unlikely to be 700 photo-essays (one every hour for a month) as the description of “200 videos, radio programmes, magazines and photo reports each week” would require.
Alternatively, if one differentiates a photo report from a photo by assuming a minimum of two pictures, and the graph showing approximately 700 ‘photos’ accurately reflects the data, then 700 / 2 = 350 reports. ISIS photo reports regularly have more than two images, but this is just a theoretical minimum.
If the non-photos account for the remaining 200 pieces (approx.) of content, which recent articles suggest, and this is added to the maximum number of photo reports it would be around 550 pieces.[x]
This is still a lot, but would still eliminate the claim of “200 videos, radio programmes, magazines and photo reports each week”. Let alone the claim made in September 2017 of “hundreds and hundreds of unique media products, videos, magazines, radio bulletins, in lots of different languages coming out every single day.”
As noted in a previous post, just counting a photo, speech, and video as the same makes little sense;
“It should be needless to write, this audio-release [of ISIS leader’s speech] by al-Furqan is of much greater importance than a single image, or photo report – at least for IS sympathizers and operatives. Although currently we still find ourselves having to write it”.
How the 2015 data is compared to 2017 production is illustrative of the way the decliner narrative is constructed. Recent journalism around ISIS decline, (which has since been publicised by UK government on Twitter) has used some creative ways to display longitudinal data, which fall well short of what would be required to pass 6th grade math class.
Edward Tufte, has “set out a detailed analysis of how to display data for precise, effective, quick analysis” in his Visual Display of Quantitative Information. This includes demonstrating that inappropriate use of data can lead to an apparent, but entirely spurious, connection between the fortunes of the New York Stock Exchange and the level of Solar Radiation.
Contrary to the ideas set out by Edward Tufte, the chart featured in a recent wired article gives the impression of a steep, linear decline by evenly spacing the bars, even though the chronological space between data points varies significantly. Two of the time points are only a single month apart while the other gaps are 6 and 17 months respectively.[xi]
When the image is quickly resketched, so all months are represented equally – both months with and without data – the problems become evident.
- First, the impression of sharp decline is reduced.
- Second, what happened to 2016?
Why does recent decliner narrative rely on ignoring 2016 and compressing these 17 months in the graph? Was there nothing to report in 2016?
Data from February 2016 to March 2017 shows production actually fluctuated, despite research claiming a “huge and steady decrease” or 30% to 40 % reduction.[xiii] At times in 2016 content was rising, at others falling. This nuance is entirely missing from decliner logic (whether from Coalition Information Operations or recent journalism).
Decliner logic insists on there being a single direction and that direction being decline – this despite decliner commentary also claiming to have identified recently a 25% increase in content – (barely 20 pieces per week in November and 25 pieces for week in December).
Using rolling mean to examine how production changed over 2016 uncovers some important results. For example, there was an increase of 132% in content production between low points in September / October 2016 and the decliner cherry-picked month of February 2017. This means, February 2017 occurred during a period of rising content production, rather than production heading in the single downward direction, as the Global Coalition imply and decliners explicity claim.
Equally, despite the counting of photos being used to show decline between 2015 and February 2017, video production increased. By adding all media types together, a decrease in easily produced photographs hides an increase in the higher resource requirement and greater impact of video production.
What about 2014?
Analysis which uses no-Arabic, faux-Arabic or google translate is no substitute for being able to read and listen fluently in Arabic. It is essential researchers are genuinely able to recognise the encoded references, historical precedents, and understand the habitus of the intended audience.
Recent commentary has suggested:
“there can be no questioning the fact that the Islamic State’s media capabilities largely relied on its territorial clout between 2014 and 2017”.
However, a different picture emerges once the flaw in decliner logic and fluctuations in content have been exposed by genuine data science. This different picture is one which has long been evident to anyone embedded in an Arabic context and “crystal clear to virtually anyone who has the linguistic capacity to grasp and the opportunity to witness what jihadists are actually saying, writing and doing, both online and offline”.[xiv]
Unfortunately, it is a picture which is too often obscured from view by commentators writing from the perspective of a Western, English language dominated, habitus. These commentators often exhibit masculinist understandings of power, success, and victory and rely on faux-Arabic or google translate.[xv]
Just as genuinely longitudinal data disrupts the narrative of a consistent downward direction in content production, analysis of the territory held alongside the media production also shatters the illusion of a pseudo-correlation between the territory held and content production.
The Global Coalition has reported on the reduction in ISIS territory as a percentage area held at the territorial highpoint of August 2014. Yet, decliners claim summer 2015 as the media highpoint, which occurs while ISIS was losing ground, including parts of Sinjar and Kobane (14% in total during 2015). Put simply, the count of content was increasing toward the claimed ‘highpoint’ while territory was decreasing.
Furthermore, the claims of a relationship between territory and media production after the 2015 media ‘highpoint’ are problematic because content production went up 132% between October 2016 and February 2017, yet territory held by ISIS went down. The Coalition territorial estimate for October 2016 shows that the territory ISIS had lost was 56% Iraq 27% Syria, by February 2017 this had become 63% and 35% respectively.
Contrary to what they expect, decliners asking ‘what about 2014?’ reminds us that the claimed media and territorial ‘highpoints’ occur a year apart and that the pseudo-correlation between the two evaporates when the change in production and territory are compared with genuinely longitudinal data rather than cherry-picking.
Conclusion: Appropriate use of Data Science emphatically destroys claims of pseudo-correlation between content production and area controlled. Saying ‘what about 2014’ just makes this clearer.
A supplemental thought experiment:
For the sake of a thought experiment, as the Global Coalition has been claiming success and territorial gains against ISIS – what would decliner logic predict about the volume of social media produced by the Global Coalition? So has the Coalition social media increased or decreased?
Using the last 3,000 tweets from each of the accounts run by the Global Coalition Against Daesh (Arabic, English and French respectively) all could be described as ‘in decline’ from an earlier highpoint. For Arabic and English, the ‘highpoint’ is in 2017, for French it is in 2016. If production is linked to battlefield success – Arabic and French speaking forces are in deep trouble – as production is down (over 90% for Arabic) from their social media height.
No doubt you are reaching for alternative explanations other than the collapse of coalition forces. If so you have understood the problem with decliner logic perfectly.
It is perhaps telling that the logic when applied to anything else is utterly transparent. The coalition propaganda even trips over its own logic, seeking to claim decline and how gaps in the production of magazines should be interpreted.
The above images show the Global Coalition has highlighted the time since the last issue of Rumiyya, but those with a longer memory will recall that before Rumiyya there was another now discontinued magazine, Dabiq. Both magazines are examples of communicating to the secondary – at best – target audience, namely English speakers. However, it is instructive as, in this case as Dabiq, there was a gap in production which began in September 2015 and lasted over 100 days (second longest gap between Dabiq issues). This, it should be noted, came the month after the claimed peak in ISIS content production, which decliners present as the zenith of the Caliphate.
If gaps in production of a prominent magazine (available in English) indicates weakness and territorial losses, as the Coalition implies, how was it Dabiq exhibited this ‘weakness’ during the period the Coalition refers to as the ‘high point’ of ISIS media production?
Indeed, as the Coalition and decliner journalists focus on a masculinist, post-Westphalian measure of victory, ISIS engage with their intended audience on a different plane. As narrated in a recent video اهل الثبات (The people who are steadfast) from ولاية كركوك (Kirkuk):
Morale is not something you can buy with money, and victory is not measured in square kilometres rather it is measured by the overall outcome, including the outcome in the hereafter, and not short-term achievements.
It is true that we lost ground, but with every day that passes the reality of the battle is becoming apparent to the Muslims worldwide, that this is a global campaign against Islam and the Muslims, it is a campaign against the Sharia and the very basic fundamentals of Islam.
Praise be to Allah that the mere existence of the Khilafah said what no long lectures and books could ever do to the hearts and minds of the Muslims worldwide.
I guess it is clear from the overall situation that we have already won the battle on the field of morale and ideas, winning it on the ground is just a matter of time, by the grace of Allah.
Victory is a complex concept in Jihadist interpretation of Islam.[xvi] Just as there is no meaningful relationship between Solar Radiation and NYSE, so Coalition propaganda, decliner logic, and the claimed pseudo-correlation between media and territory fail to provide an authentic representation of the current fortunes of the jihadist movement, their strategy nor their tactics.
As noted in 2014, the media mujahedeen are constantly reconfiguring and finding new outlets. In 2018, content production will continue to fluctuate.
The much talked up bluckling of the ISIS media system, presented as “not just a media decline—it is a full-fledged collapse” is likely in retrospect to have been a lull as the swarmcast reconfigured, rather than signs of ongoing decline.
[i] See for example;
Rushkoff, Douglas. Cyberia: Life in the trenches of hyperspace. Clinamen PressLtd, 2002.
[ii] It is also worthy of note, that in highlighting the gap in Rumiyah production (above) and the ‘high’ point in 2015 (below) the Global Coalition get themselves in a logical tangle. This is because the second longest gap in the production of Dabiq began in September 2015 the month after the claimed peak in content production. Was it strong – as the ‘high point’ would suggest – or weak as the long gap in publication of a magazine would imply in their logic about Rumiyah.
[iii] Mehdi Semati & Piotr M. Szpunar, ISIS beyond the spectacle: communication media, networked publics, terrorism, Critical Studies in Media Communication Vol. 35 , Iss. 1,2018
[iv] Paz, Reuven. “Reading Their Lips: The Credibility of Jihadi Web Sites as ‘Soft Power’ in the War of the Minds.” (2007).
[v] Lizardo, Omar. “The cognitive origins of Bourdieu’s habitus.” Journal for the Theory of Social Behaviour 34.4 (2004): 375-401.
[vi] Lizardo, Omar. “The cognitive origins of Bourdieu’s habitus.” Journal for the Theory of Social Behaviour 34.4 (2004): 375-401.
[vii] Lizardo, Omar. “The cognitive origins of Bourdieu’s habitus.” Journal for the Theory of Social Behaviour 34.4 (2004): 375-401.
[viii] This is not to say the intended audience is sympathetic to the message, just that the message is intended for their consumption.
[ix] In the following section it is unclear whether the dubious data handling is due to the ignorance of the required skills / methods, or a sleight of hand in the presentation.
[x] The published research lists the titles of only 140 pieces of content
[xi] This trick is also used here: https://twitter.com/charliewinter/status/920651729393389568
[xii] There is a lot wrong with counting content – except where you are comparing total files detected for takedown vs. files released – this section shows what is wrong with decliner logic even in their own terms, rather than because the approach itself is insightful.
[xiii] This data comes from the torrent files released containing a collection of all the content for the ‘week’. These files were not released consistently in a seven-day cycle, but often varyied between a six and ten day cycle. As we have used the torrent file date to produce date of production, there are some periods of seven days without a total. This however, does not effect the total production and rolling mean is used to provide an authentic view of the ongoing levels of production.
[xv] From an IR perspective see:
Sjoberg, Laura. Gendering global conflict: toward a feminist theory of war. Columbia University Press, 2013.
For a jihadist perspective see:
Anwar al-Awlaki, State of the Ummah.
[xvi] See for example; Constants of Jihad and the discussion of the contents of the book by Anwar al-Awlaki.