Introduction to the report ‘Liking violence: A study of hate speech on Facebook in Sri Lanka’
By Sanjana Hattotuwa
“Hate speech” on the Internet is a global concern and with no kill-switch solution. Depending on the location online, language and media used, context and sometimes even the nature of the actors concerned, dealing with hate speech is a vexed challenge from parent to policymaker. This hasn’t stopped politicians, with little to no understanding of underlying technical challenges or repressive governments, who often seek a monopoly around the dissemination of defamatory propaganda seeking to control hate speech. Parochialism and expediency drive most efforts around hate speech related policy responses and legislation. In Sri Lanka, online social media and web based platforms, accessed increasingly over smartphones and tablets, provide an important, necessary vent for critical dissent, in a context where mainstream media does not and cannot afford the space for questioning or content that holds the government accountable for heinous crimes and outrageous corruption. The growth of content creation and consumption online, wider and deeper than any other media in the country and at an accelerated pace, has also resulted in low risk, low cost and high impact online spaces to spread hate, harm and hurt against specific communities, individuals or ideas. Conspiracy theorists, fringe lunatics and trolls have since the first days of the Internet inhabited online spaces and engaged with devoted followers, or sought to deny and decry those who question them. The growth of hate speech can be seen as a natural progression outward from these pockets of relative isolation, and is also pegged to the economics of broadband internet access and the double digit growth of smartphones – an underlying, coast to coast network infrastructure capable of rich media content production and interactive, real time engagement. This infrastructure has erased traditional geographies – hate and harm against a particular religion, identity group or community in one part of the world or country, can for example within seconds, translate into violent emulation or strident opposition in another part, communicated via online social media and mediated through platforms like Twitter, Facebook and also through instant messaging apps for mobiles like iMessage and WhatsApp, in addition to the older SMS technology.
A central challenge around addressing hate speech is that it is technically impossible – given the volume, variety and velocity of content production on the Internet today1 – to robustly assess and curtail, in as close to real time as possible, inflammatory, dangerous or hateful content just in English, leave aside other languages like Sinhala or Tamil. Once content is produced for the web and originally for a single platform, given user interactions and responses, it often replicates and mutates into other content over dozens of other websites and platforms, making it impossible to complete erase a record of its existence even if the original was taken down, deleted or redacted. This makes it extremely hard to address the harm arising out of hate speech, since there is so much of it around in digital form over so many media.
Another challenge is in defining hate speech. Overbroad legislation risks the law being used to curtail and stifle dissent. Loosely defined laws allow perpetrators of hate speech to get away by referencing the freedom of expression. Policymakers who have to respond to angry communities and individuals who are the targets of hate speech, if they are important constituencies, often respond with promises to address a problem they in fact cannot. Internet Service Providers and large corporations like Google, Facebook and Twitter have developed robust guidelines around the content they will allow on their platforms, but these seem to only work best around output that is in English. For example, this brief study is testimony to the sheer volume of hate freely disseminated in Sinhalese on Facebook, even though the company has clear guidelines around such content which includes the banning and blocking of users.
Reflecting the lack of any universal definition of hate speech – content acceptable to or posted lawfully in one country or region can be deemed hateful and unlawful in others, even on the same platform or site – the term is, unsurprisingly, variously defined across leading web companies. Google’s YouTube defines it as2,
…content that promotes hatred against members of a protected group. For instance, racist or sexist content may be considered hate speech. Sometimes there is a fine line between what is and what is not considered hate speech. For instance, it is generally okay to criticise a nation, but not okay to make insulting generalisations about people of a particular nationality.
Facebook defines hate speech as3
Content that attacks people based on their actual or perceived race, ethnicity, national origin, religion, sex, gender, sexual orientation, disability or disease is not allowed. We do, however, allow clear attempts at humour or satire that might otherwise be considered a possible threat or attack. This includes content that many people may find to be in bad taste (ex: jokes, stand-up comedy, popular song lyrics, etc.).
Addressing that it hadn’t done enough in the past to address hate speech4, Twitter’s current rules and policies note that5,
Users may not make direct, specific threats of violence against others, including threats against a person or group on the basis of race, ethnicity, national origin, religion, sexual orientation, gender, gender identity, age, or disability.
Add to these varying definitions is that the challenge of defining hate speech – in contradistinction to for example content that is just mildly offensive, distasteful, satirical or acerbic – is deeply rooted in context and expression. What could be a generally accepted turn of phrase used in colloquial speech, when translated into English and out of context, can be seen as hate speech in line with the guidelines noted above. Similarly, hate speech can be easily disguised by resorting to non -English and non-textual expression – or a combination of both. This study has a number of examples where inflammatory and defamatory content against a specific community finds expression and openly resides on Facebook because
it is in Sinhala, a language that clearly lies outside existing language competencies of Facebook’s automated and human curated monitoring frameworks. This brings us to a key challenge around hate speech – it always requires context to understand and address, and increasingly, the intermediaries in both supporting and curtailing the spread of it are corporate entities, not governments. Machine level and algorithmic frameworks to identify and block hateful and harmful content often fail, simply because they flag too many false positives (content erroneously flagged as hate speech) or allow so much of hate speech to pass through (in, as noted earlier, languages other than English) that their core purpose rendered irrelevant. This puts the burden of addressing this content on users themselves, who through reporting mechanisms baked into all the major only social media platforms, can choose to report hate speech with relevant context. Only as effective as the numbers who report hate speech, these reporting mechanisms also take some time to kick-in from the time of submission to the actual deletion or blocking of the original content, page, account or user. At a time of heightened violence, this time lag is unhelpful. There is also no guarantee the (corporate) owner of an app, service, platform or website agrees with the reporting of hateful content. Studies show, for example, significant variance in dealing with hate speech even within Facebook 6.
This study aims to focus these challenges around the significant growth of hate speech in post-war Sri Lanka, primarily directly against the Muslim community and Islam. The rise of Islamophobia in Sri Lanka is well documented7 and shows no signs of abating. Studies on this score are often anchored to the statements by extremist groups in public rallies, and physical acts of violence and intimidation. Equally remarkable though less studied is the growth of hate speech in online social media. As I noted in 2013 after the study of just around four of the most active extremist groups on Facebook8,
Even the most offensive anti-Muslim sentiments and statements have a growing audience and following in web based social media
That such content has a greater chance of going viral, and influencing real world action, when published in online fora as opposed to mainstream and traditional media Content is largely visual in nature, appealing to a demographic as young as 18 (who are still in school) Anti-Muslim hate speech is generally, qualitatively more vicious and venomous than anti-LTTE sentiments even at the height of war
Numbers of those joining these groups is on the rise, and the government is either unaware or unable to address this through counter-narratives and content in support of liberal values, tolerance and religious cohesion.
The focus of this study is to expand on these points. Sadly, the content for the research is growing in abundance. When juxtaposed with the increasing violence against sexual, ethnic and religious minorities and the open celebration of hate speech by groups like the BBS with total impunity, content online risks fanning even greater violence in the future. Even if to date there is little evidence of content online leading to actual physical violence, what is particularly disturbing – given the tens of thousands who are actively
producing and engaging with hate speech – is the radicalisation of youth, as young as 18, to an alarming degree. Though discussions and content are respectively conducted and produced in public fora over platforms like Facebook, the scale and degree of this radicalisation remains ironically hidden to politicians, policymakers and even most parents because of a digital media literacy gap. As I warned in 20139,
Given that the extremists are web savvy, and escape the usual checks on the spread of racist content by virtue of publishing material in Sinhala, it is to be expected that unless serious, meaningful and urgent measures are taken by government, hate will overcome more moderate voices online, and risk spilling over to real world violence on the lines of Black July 1983, against Muslims.
Obviously, the growth of hate speech online in Sri Lanka does not guarantee another pogrom. It does however pose a range of other challenges to government and governance around social, ethnic, cultural and religious co -existence, diversity and, ultimately, to the very core of debates around how we see and organise ourselves post-war. What this study lacks, by design, is a list of solutions to counter the growth of online hate speech. There is simply no panacea, no easy fix or solution in the short term that will effectively curtail the emergence of hate speech online in the future. Indeed, a government that protects instigators of hate is not one that can drive progressive policies around addressing a growing trend this same hate expressed online. Politicians who are digitally illiterate are equally ill- placed to bring about legislation that addresses hate speech even though it may appear to be expedient to do so in light of increasing violence. What this study aims to provide is evidence around what remains an under -appreciated driver of conflict and violence post-war. To acknowledge the scale and depth of the problem is a step beyond an ignorance that it even exists. Moving forward requires all levels of government, private corporations outside of Sri Lanka that host social media content, civil society within the country including the legal community, conscientious individuals and institutions in the diaspora and local ISPs, out of a duty of care for their customer base working in concert to address this explosion of hate speech online. Though it is unclear when and if a concerted, collective approach or a wider study around hate speech in Sri Lankan online fora will be undertaken, this report provides a starting point for informed discussions around how urgently this disturbing phenomenon needs to be studied and remedial measures, to the extent possible, taken.
My sincere thanks and appreciation to Shilpa Samaratunge, the lead author and researcher of this study. Despite being profoundly distressed by what she encountered , Shilpa’s sharp eye and intelligence was simply invaluable in matching existing research on online hate with content found on Sri Lankan websites and social media. Without her, this report would simply not be.
read as a PDFHate Speech – Executive Summary
September 2014
[email protected]
1 http://www.ibmbigdatahub.com/infographic/four-vs-big-data
2 https://www.youtube.com/t/community_guidelines
3 https://www.facebook.com/help/135402139904490/
4 https://blog.twitter.com/en-gb/2013/our-commitment
5 https://support.twitter.com/articles/20169997
6 http://ohpi.org.au/if-you-cant-recognize-hate-speech-the-sunlight-cant-penetrate/ 7 http://newint.org/blog/2013/04/15/islamaphobia-in-sri-lanka/
8 http://sanjanah.wordpress.com/2013/02/01/anti-muslim-hate-online-in-post-war-sri-lanka/
9 http://sanjanah.wordpress.com/2013/02/01/anti-muslim-hate-online-in-post-war-sri-lanka/