Written evidence submitted by Demos (OHC0031)

This document is a response to a call for evidence by the Home Affairs Committee in connection with the Committee’s inquiry into hate crime and its violent consequences. It focuses on the current state of the far-right in the UK and the threat it poses, the far-right’s use of social media, and the question of how far social media companies need to take action. In this submission, we have drawn primarily from Demos’ own research into this subject, rather than summarising external sources.

1.       The current state of the far-right in the UK and the threat it poses

The contemporary far-right in the UK is rapidly evolving: groups form and dissolve quickly but draw on a similar base of supporters. Recently, the UK witnessed the emergence of far-right street-based groups alongside traditional parties, whose modus operandi is volatile street demonstrations and marches as opposed to formal electoral politics.

In 2011, Demos undertook an online survey of Facebook members of the (now dissolved) English Defence League (EDL). We estimated the total active membership to consist of least 25,000 people. 81% of surveyed supporters were male, 72% were between 16-30 years old and a significant percentage (28%) was unemployed. Supporters were more educated than their reputation might suggest: 30% were educated to university or college level, and 15% held a professional qualification.[1]

The biggest concerns of EDL supporters were immigration and Islamic extremism. We also found that EDL supporters’ lack of confidence in the legal and justice system as well as their pessimistic outlook about the future were important factors in explaining their support for the group. We encountered similar results in a cross-country study of online followers of far-right parties and street-movements across Europe.[2]

Importantly, it appeared from our survey that a clear majority of EDL supporters believed that voting did matter (approximately consistent with the national average). This suggests that EDL supporters had some faith in the power of parliamentary democracy. Supporters also had broadly similar levels of trust in parliament and political parties as the national average. Despite the outbreaks of violence at many of their demonstrations which indicate that the organisation included violent elements, supporters cited ‘rule of law’, ‘individual freedom’ and ‘respect for human rights’ among their top values.

In odds perhaps with the image they projected, the EDL and other large far-right street-based groups were far from one-dimensional, and members’ views were varied. The EDL can probably best be described as a populist movement that contained some extreme right-wing and sometimes Islamophobic elements. Although some supporters voiced illiberal and intolerant sentiments in the survey (and at demonstrations), many members were in an important sense democrats.[3]

2.       The far-right’s use of social media

Far-right groups are often active users of social media. As part of a recent study entitled Plugged In: Social Media for Social Action Demos interviewed individuals and organisations in the UK using social media to promote their cause. Our interviewees included the Democratic Football Lads Alliance (DFLA), who are often labelled as a far-right group. Like many other groups campaigning for social action, the DFLA use social media to fundraise, raise awareness about their cause and recruit new members. Social media offers individuals and organisations – including far-right groups — a valuable platform to reach large audiences, network with like-minded people, organise events and fundraise, without spending prohibitive amounts of time or resources doing so.[4] 

The DFLA and other far-right groups are particularly keen users of social media platforms, as they feel that their message is often boycotted or misrepresented by mainstream media. These alternative online channels allow them to bypass the mainstream press and communicate directly to their audience. Far-right groups also use social media to share mainstream media news stories that fit within their narrative and increase the credibility of their message. The EDL, for instance, used social media to share “real” news stories portraying Muslims in a negative light that fit with their view that Islam is incompatible with the values of Western societies.[5]

While social media platforms are a useful tool for far-right groups to broadcast their views, one should be careful not to equate online followers with active members. In our 2011 Facebook survey of the EDL Facebook members we found that only 44% of the respondents had participated in a local demonstration in the past six month, 24% travelled 100 km or more for a national demonstration and 52% took part in online activism.[6] Nevertheless, our subsequent interviews of UK activists using social media to promote their cause indicate that passive or low-level engagement online can lower the barriers of entry into social activism offline. Several interviewees felt that social media allowed them to build credibility for their organisation, and that engaging people online was a vital first step to activism offline.[7]

It is worth noting that Islamophobic and xenophobic messages on social media are not the sole preserve of far-right groups, or users that are sympathetic to their views. In a recent study, we identified 143,920 Tweets over a year (the equivalent of 393 tweets a day) sent from the UK considered to be derogatory and anti-Islamic, sent from over 47,000 different users. We found that Tweets containing hateful anti-Islamic messages were sent by from several different groups or online tribes:

●       users engaged in anti-Islam and far right conversations

●       casual users getting involved in the immediate aftermath of terrorist attacks or Brexit

●       football fans

●       users supporting Indian nationalism

●       users involved in the gaming community [8]

We also found that this discussion driven as much by events as by ideology, with key events such as terror attacks and the Brexit referendum causing increases in the volume of Islamophobic and xenophobic tweets.[9]

3.       Should social media companies take action?

There are a number of important questions to consider when assessing social media companies response to hateful content sent or shared on their platforms. The first concerns content which is reported to a platform as clear violation of UK law or a platform’s own policies. All major social media companies have existing measures in place to tackle hateful content online, including bans on hate speech, threats and extremist content. For instance, Facebook prohibits content that contains any “direct and serious attack on any protected category of people based on their race ethnicity, national origin, religion, sex, gender, sexual orientation, disability or disease”. [10]

It is imperative that these rules are enforced, requests from enforcement agencies for hateful content to be removed are swiftly acted upon, and efforts are taken to prevent offenders from rejoining platforms and resuming attacks. Although there is room for improvement, there has been some progress here over the last few years. A 2018 report by the European Commission found that social media companies are increasingly fulfilling their commitment to remove illegal hate speech within 24 hours—on average, Facebook, Twitter, YouTube and Microsoft removed 70% of all the illegal hate speech notified to them in January 2018 compared to 28% in May 2016.[11]

In calling for enforcement of these standards, however, there is a danger that important discourse which sits in the grey areas of our definitions of hate speech becomes prohibited. In our 2017 research on anti-Islamic content, we found that there was a prominent group of users sharing content which, though classified as likely to be hateful by our algorithms, were using slur terms in order to protest against their use by others, or as an act of appropriation.[12] A similar imperative exists in maintaining the freedom of political expression. If platforms are seen as banning individuals or groups from using social media platforms based on their political views, this would stifle free speech and legitimise the view of far-right groups that they are being boycotted by the mainstream media.

A second challenge is presented by content which is not proactively reported to platforms. Due in part to the speed and volume to which content is shared online - over half a million posts are shared on Facebook every minute - social media platforms cannot be expected to find and remove every instance of hate speech. The primary difficulty in doing so is not only a question of resourcing, but also because of the challenges in automating the identification of extremist and hateful content. It is a misconception to imagine that if social media companies can, for example, create algorithms that remove content infringing copyright or child abuse images, they must be able to create one to identify and remove hateful and extremist content.[13] Using algorithms to define and identify extremist content, however, is a very different endeavour to removing copyrighted material or child abuse images, and there is no technical silver-bullet for the former.

One difficulty here lies in the ambiguity of many acts of hate speech; an ambiguity which is less acute for copyrighted material, which can be identified through cross-referencing media with existing databases of material.[14] An important factor in the relative success with which images of child sexual exploitation have been detected online is the unified Child Abuse Image Database, developed by the Home Office, which brings together child abuse images collected by the police force and the National Crime Agency. In each of these cases, digital ‘fingerprints’ can be extracted from media and images, which can be cross-referenced to known prohibited content held in a database.[15] There is no such database for hate speech.

One reason for this is the inherent ambiguity and context-sensitivity of language. As we argue in a recent Demos Quarterly, "one person's hateful content is another's reasonable criticism," and algorithms are ill-suited to deal with these ambiguities.[16] For example, our research on Islamophobia on Twitter used algorithms to analyse tweets on the basis of keywords that they contain, but none of the algorithms trained were 100% accurate (our classification had a precision of approximately 72%).[17] While it may well be possible to raise this precision, or to identify re-posting of content previously removed or flagged as “harmful” or “extremist”, creating an algorithm to identify novel extremist or harmful content with perfect accuracy is impossible, as this type of natural language classification is inherently probabilistic.[18] This is not to say that algorithms will be useless in these cases; only that substantial human oversight will likely be needed in order to make decisions based on their output.

Social media companies are also under increasing pressure to proactively identify illegal hate speech content on their platforms . Although the Crown Prosecution Service has made it clear that what is illegal offline is illegal online, and the have published guidelines to clarify the circumstances under which prosecution should be brought, deciding when laws and legal guidelines apply is subject to interpretation.[19] This is especially true in the case of hate speech where the context of the communication and the use of language might not be familiar to social media companies and the line between separating rude comment from grossly indecent ones can be blurry.[20] In other words, the fact that social media companies have the ability to remove content posted on their platforms does not mean that they will be able to make the right decisions about which content constitutes or does not constitute hate speech.

We believe that while social media platforms have an important role to play in removing hate speech online, the drawing of the line between hate speech and legitimate opinion should not be left solely in the hands of these private companies. It is crucial that these decisions involve a plurality of voices, and there are key roles to be played by law enforcement, legislators, civil society and the users of these platforms themselves in deciding what constitutes acceptable discourse online.

In order to have this debate, we need to ensure there is independent oversight as to how existing rules are applied. Social media companies should be encouraged to be as transparent as possible to the public about the content they have removed, and the decision-making process used when removing content. It should be noted that platforms have made advances in this area - YouTube, to take one example, regularly publishes a detailed report on the ways in which their community guidelines have been enforced.[21] However, due in many cases to the impossibility of accessing content once it has been removed or deleted, it is often difficult to scrutinise exactly how these policies are being applied. This is doubly true for platforms such as Instagram, which has an extremely restrictive policy on researchers accessing content shared on the platform.

Alex Krasodomski-Jones, Head of the Centre for the Analysis of Social Media

Agnès Chauvet, Researcher at the Centre for the Analysis of Social Media

Josh Smith, Senior Researcher at the Centre for the Analysis of Social Media

September 2018