Month: July 2017

Google search data Islamophobia

Can Google search data help solve Islamophobia?

For decades, social scientists have conducted research using some combination of surveys, census data, focus groups, interviews, and observation techniques. With the exception of covert observation, which brings its own ethical issues, these methods have something in common: the dishonesty of people. These methods are all subject to human lies and therefore unable to paint a reliable picture of society’s true beliefs and darkest fears. In fact, the most objective forms of data are given up willingly, in private, where people are free from the worry of being judged. Short of stealing people’s diaries or tapping their phone calls, what else can researchers do to gather the most objective data possible?

Better than surveys

In our digital era the most obvious answer is also the correct one. But until now, few people have thought to leverage this tool and publicise their findings in such an accessible way and at such a pertinent time. What is the technology we all use to ask questions, seek validation, and search for the most outrageous things? Why of course, it’s Google. Many people would be embarrassed to publicly display their Google search history. I know mine is full of very silly things. But at the same time, these queries are deeply revealing; which is precisely why they strike a nerve. They display some of our deepest secrets. For example, a few years ago I used to get occasional panic attacks. I remember waking up at 3 am in an unfamiliar country, caught in the midst of an attack, gasping for breath. To calm myself, I searched Google for reassurance that it was ‘just’ a panic attack.

Google as ‘truth serum’

People search Google for all manner of things. Seth Stephens-Davidowitz (see below for video of his recent RSA talk), the researcher who produced this study, found many searches for terms involving ‘abortions’, ‘closet gays’, ‘penis size’, and ‘breastfeeding of husbands’ (the latter being apparently popular in India). He also found other more sinister patterns, ones suggesting American racism was far more widespread than previously thought. In fact, search data shows the idea of America as a ‘post-racial’ society, much-touted after the 2008 election of Barack Obama, to be quite absurd. Google showed American racism and Islamophobia were thoroughly alive and kicking, even in places where people didn’t publicly admit to holding racist views. They espouse very different opinions in the privacy of their own homes, face-to-face only with Google. It’s Google as ‘truth serum’. Almost ten years later, with Trump at the helm, perhaps America is finally showing its true face.

Tracking Islamophobia in searches

Obama’s address to the nation after the 2015 San Bernardino attack, provides an interesting example of how search data reflects hidden social views. In the speech, he aimed to calm outraged people and restore order to the country. In particular, he wanted to counteract the backlash that Muslim-Americans would surely face. While he was speaking of Muslims as ‘our friends’, ‘neighbours’ and so on, Google search data was telling a different story. After each terrorist attack (and this happens in the UK too) the volume of negative and violent searches about Muslims skyrockets. Islamophobic searches like ‘kill all Muslims’, become alarmingly high.

During most of Obama’s speech, these searches didn’t reduce or even level off. Instead they became even more frequent. This makes sense, because challenging people’s world views acts as an attack on their fundamental identity. In response, most will cling tighter to whatever they believe. But later in his speech, Obama changed tack. He introduced new images: not just of Muslim-Americans as friends and neighbours, who should be respected, but also of ‘Muslim soldiers’, willing to die for America, and ‘Muslim athletes’, representing the country on the world stage.

From ‘terrorists’ to soldiers and athletes

And then, something changed in the data. Islamophobic searches slowed down, to be replaced with searches for ‘Muslim athletes’, and ‘Muslim soldiers’. Something had resonated with the people searching; instead of responding predictably to Obama’s perceived ‘attack’ on their entrenched world views, they had become curious. I believe this happened for two reasons, partly because the idea of Muslims as athletes and soldiers resonated with ‘patriotic’ American audiences. But also because these images perhaps helped to ‘de-otherise’ public perceptions of Muslims. By drawing on resonant all-American themes, Obama associated Muslims with a set of positive images rather than just trying to convince wider America to accept them as a group. In response, albeit temporarily, the volume of Islamophobic searches slowed and included more positive searches.

This is encouraging in some ways, because despite the fleeting nature of this positivity, its presence suggests two important things, 1) that Islamophobia is largely a problem of perceptions, and 2) that the tide can be turned back. Negative views of Muslims have become deeply entrenched over the last three decades. Islamophobia as a public perception is regularly reinforced by mainstream media, by certain think tanks and their ‘experts’, and by reactions to the terrible deeds of ISIS; a group that has hijacked the image of Islam worldwide.

How can this data help us?

Can Google search data offer us the chance to fix some of society’s ills? Its revealing nature shows our darkest fears in a way no survey can ever do. Having this information (anonymous of course) could be used to bring issues into the open and address their root causes. In the case of Islamophobia, analysing Google searches could reveal where the gaps and misperceptions lie in wider society’s understanding of Muslims. It could allow us to categorise the fears, misunderstandings, and false perceptions. This could inform the design of social initiatives targeting specific problems, helping people understand each other better and gain a stronger sense of reality over perception.

Reflections: Terrorism and Social Media Conference 2017

Last week, in a sleepy Welsh city by the sea, a group of social media and terrorism researchers came together to discuss the latest challenges in the field.

I learned a lot, met people doing admirable work and came away inspired with ideas to shape my own research in the future. This post is a short synopsis of topics from the conference that struck me as important, interesting and/or particularly thought-provoking.

The visual web

Maura Conway’s opening keynote was peppered with mentions of the visual web – and it’s importance in the study of terrorist and extremist activity. All extremist groups have a visual profile, and many use images as a central feature of their propaganda and recruiting efforts.

One look at the ISIS propaganda magazine, Dabiq, proves this point. And it’s not only about images, but also video, which terrorist groups have used for decades, from the grainy, muffled bin Laden recordings all the way through to the glossy ISIS productions. Far-right groups use images too – from the notorious Pepe the Frog to a range of logos featuring swords, swastikas and national flags.

The ‘post-truth’, digital era has ushered in a trend for using images as part of disinformation efforts; driving so-called ‘fake news’. A recent example springs to mind from the March 2017 Westminster attack. In the swirling social media aftermath of Khalid Mahmood’s actions there emerged a photo of a Muslim woman wearing a hijab, walking past victims across Westminster bridge, engrossed in her phone as she walked.

The image was quickly hijacked, attached to numerous false claims attacking the unknown woman for her apparent ‘disdain’ for the injured victims. These claims spawned thousands of comments where people released their Islamophobic feelings to the full, feeding into the milieu of anti-Muslim sentiment that presently hangs over society.

Of course, the truth was very different. The woman had been messaging friends and family to let them know she was safe after the attack. Despite the truth being outed, the damage had already been done. Social perceptions of Muslims as ‘bad’ had been further reinforced.

Back to Prof Conway’s speech; in which she highlighted the ‘strong signalling function’ of images, making them critical subjects for further analysis. Yet most terrorism analysts still focus primarily on text, because the analysis of images is more challenging. Visual analytics tools and techniques do exist, both qual and quant, with big data research on images being especially popular in communication science at the moment.

In short: we need to pay more attention to the visual nature of the internet – and focus more on these ‘low-hanging fruit’ of visual analytics in the study of extremism.

The far-right

TASM didn’t focus only on the Islam-related side of extremism, but showcased a balanced view across the spectrum, with plenty of emphasis on research into the far-right. I attended several interesting panel talks on this subject, and came away with a number of key points.

One piece of research compared Britain First with Reclaim Australia, aiming to draw out the nuances within the umbrella term ‘far-right’. Methodology involved corpus assisted discourse analysis (CADS) on a static dataset, showing text that Britain First and Reclaim Australia supporters had posted on social media over a three-month period.

The researchers used a social media insights tool, Blurrt, to gather raw data, then used Python scripts to sort it into a workable format before finally analysing using CADS. In particular, they focused on collocations to reveal telling patterns in ideas and sentiments across the two groups.

Findings included a strong pattern of ‘othering’ – the core ‘us versus them’ narrative (which is a common theme not just among far-right discourse but also in some mainstream media and foreign policy: e.g. the Iraq war – ‘Axis of Evil’).

It was unsurprising therefore to find that Muslims and immigrants were particularly targeted. In what appears to be an extension of the ‘us versus them’ theme, ‘metaphors of invasion’ were often found in the discourse of both groups.

Other common themes included mentions of ‘our women’, ‘our religion’ and ‘our culture’ as being under threat from the ‘invaders’. All these themes feel very masculine. It could be interesting to reflect on the proportion of these sentiments that come from male authors; and could also be worth analysing what far-right discourse looks like from a female perspective.

In general, researchers concluded that far-right propaganda is less ‘overtly’ violent than that of ISIS, and is mainly rooted in nationalistic tendencies. This raises many questions. Is this how the far-right have managed to fly ‘under the radar’ for so long? Are they seen as being defensive rather than offensive? (And hence the ‘good guys’ on some level).

Could that be a factor in the much-discussed media under-reporting of far-right crimes, while focusing almost hysterically on those perpetrated by jihadists? Or, are ISIS and similar viewed as ‘worse’ simply because they are more ‘other’ (i.e. racism)?

Resonant narratives

Just as in commercial marketing, narratives work best when they intersect with individual agency and contexts. In his panel talk, Dr Akil Awan pointed out that CVE campaigns must not neglect the real-world issues that allow extremist narratives to resonate in the first place.

So how do ISIS narratives achieve success? They play on themes of belonging and identity; important for people experiencing ‘dual culture alterity’, i.e. feeling alienated from both their parents’ culture and the culture of their country of upbringing. In these cases, a return to fundamentalism becomes an ‘anchor’; a default setting of identity in a sea of alienation.

Awan highlighted the disparity between perceptions and reality around the true numbers of Muslims living in European countries. The media drives much of this misperception; making people feel ‘under siege’, creating fear, driving societies apart and destroying any sense of cohesion. In such a milieu, it is easy for ISIS to ‘eliminate the grey zone’ by means of terrorist acts. The media has already primed society for ISIS to succeed.

Understanding perceptions is as important as understanding reality; because how people perceive something will guide their course of action in response to it. Current CVE campaigns (based around tools such as counter-narrative videos) are cheap to implement and make it look like governments are taking action.

But recognising the ‘lived experience’ of minority groups is one of the keys to successful CVE efforts; neglecting to do so is hypocritical and unlikely to be effective.

Conclusion

In closing, we heard from the arbiter of all this – Facebook. Dr Erin Saltman explained the tools Facebook uses to tackle the online side of extremism and terrorism. These tools include a database of extremist propaganda images that relies on machine learning to match images as they surface, and automatically remove them.

But machine learning has its limitations, and humans are still required to take into account context and nuance. At present, the two work in tandem to surface the content (machine learning) and then interpret it as needed (humans).

Other tools include Facebook Insights, commonly used in commercial marketing, but can also be leveraged to guide counter-speech initiatives and enable precise reading of audiences.

The age of social media, although still in its infancy, has already had profound impact on politics and society – as well as the individual psychology of internet users. The long-term effects are unknown, with many changes no doubt still on the way.