online hate speech

Online hate speech is a type of speech that takes place online with the purpose of attacking a person or a group based on their race, religion, ethnic origin, sexual orientation, disability, and/or gender. Online hate speech is not easily defined, but can be recognized by the degrading or dehumanizing function it serves.

Multilateral treaties such as the International Covenant on Civil and Political Rights (ICCPR) have sought to define its contours. Multi-stakeholders processes (e.g. the Rabat Plan of Action) have tried to bring greater clarity and suggested mechanisms to identify hateful messages. Yet, hate speech is still a generic term in everyday discourse, mixing concrete threats to individuals and/or groups with cases in which people may be simply venting their anger against authority. Internet intermediaries—organizations and social networks that mediate online communication such as Facebook, Twitter, and Google—have advanced their own definitions of hate speech that bind users to a set of rules and allow companies to limit certain forms of expression. National and regional bodies have sought to promote understandings of the term that are more rooted in local traditions.

The Internet's speed and reach makes it difficult for governments to enforce national legislation in the virtual world. Social media is a private space for public expression, which makes it difficult for regulators. Some of the companies owning these spaces have become more responsive towards tackling the problem of online hate speech.

Politicians, activists, and academics discuss the character of online hate speech and its relation to offline speech and action, but the debates tend to be removed from systematic empirical evidence. The character of perceived hate speech and its possible consequences has led to placing much emphasis on the solutions to the problem and on how they should be grounded in international human rights law. Yet this very focus has also limited deeper attempts to understand the causes underlying the phenomenon and the dynamics through which certain types of content emerge, diffuse and lead—or not—to actual discrimination, hostility, or violence.

Definitions

Hate speech

The concept of hate speech touches on the clash of freedom of expression and individual, collective, and minority rights, as well as concepts of dignity, liberty, and equality. It is not easily defined but can be recognized by its function.

In national and international legislation, hate speech refers to expressions that advocate incitement to harm, including acts of discrimination, hostility, radicalization, verbal and/or physical violence, based upon the targets' social and/or demographic identity. Hate speech may include, but is not limited to, speech that advocates, threatens, or encourages violent acts. The concept may extend also to expressions that foster a climate of prejudice and intolerance on the assumption that this may fuel targeted discrimination, hostility, and violent attacks. At critical times, such as during political elections, the concept of hate speech may be prone to manipulation; accusations of instigating hate speech may be traded among political opponents or used by those in power to curb dissent and criticism. Hate speech (be it conveyed through text, images, and/or sound) can be identified by approximation through the degrading or dehumanizing functions that it serves.

Legal scholar and political theorist Jeremy Waldron argues that hate speech always contains two messages: first, to let members of the out-group feel unwelcome or afraid; and second, to let members of the in-group feel that their hateful beliefs are legitimate.

Characteristics of online hate speech

The proliferation of hate speech online, observed by the UN Human Rights Council Special Rapporteur on Minority Issues poses a new set of challenges. Both social networking platforms and organizations created to combat hate speech have recognized that hateful messages disseminated online are increasingly common and have elicited unprecedented attention to develop adequate responses. According to HateBase, a web-based application that collects instances of hate speech online worldwide, the majority of cases of hate speech target individuals based on ethnicity and nationality, but incitements to hatred focusing on religion and social class have also been on the rise.

While hate speech online is not intrinsically different from similar expressions found offline, there are peculiar challenges unique to online content and its regulation. Those challenges related to its permanence, itinerancy, anonymity and complex cross-jurisdictional character.

Permanence

Hate speech can stay online for a long time in different formats across multiple platforms, which can be linked repeatedly. As Andre Oboler, the CEO of the Online Hate Prevention Institute, has noted, "The longer the content stays available, the more damage it can inflict on the victims and empower the perpetrators. If you remove the content at an early stage you can limit the exposure. This is just like cleaning litter, it doesn't stop people from littering but if you do not take care of the problem it just piles up and further exacerbates." Twitter's conversations organized around trending topics may facilitate the quick and wide spreading of hateful messages, but they also offer the opportunity for influential speakers to shun messages and possibly end popular threads inciting violence. Facebook, on the contrary, may allow multiple threads to continue in parallel and go unnoticed; creating longer lasting spaces that offend, discriminate, and ridicule certain individuals and groups.

Itinerance

Hate speech online can be itinerant. Even when content is removed, it may find expression elsewhere, possibly on the same platform under a different name or on different online spaces. If a website is shut down, it can quickly reopen using a web-hosting service with less stringent regulations or via the reallocation to a country with laws imposing higher threshold for hate speech. The itinerant nature of hate speech also means that poorly formulated thoughts, or under-the-influence behavior, that would have not found public expression and support in the past may now land on spaces where they can be visible to large audiences.

Anonymity

Anonymity can also present a challenge to dealing with online hate speech. Internet discussions may be anonymous or pseudonymous, which can make people feel more safe expressing their opinions but can just as easily accelerate destructive behavior. As Drew Boyd, Director of Operations at The Sentinel Project, has stated, "the Internet grants individuals the ability to say horrific things because they think they will not be discovered. This is what makes online hate speech so unique, because people feel much more comfortable speaking hate as opposed to real life when they have to deal with the consequences of what they say." China and South Korea enforce real-name policies for social media. Facebook, LinkedIn, and Quora have sought to activate a real-name system to have more control on online hate speech. Such measures have been deeply contested as they are seen to violate the right to privacy and its intersection with free expression.

Many instances of online hate speech are posted by Internet "trolls", which are typically pseudonymous users who post shocking, vulgar, and overall untrue content that is explicitly intended to trigger a negative reaction, though may also be indended to influence or recruit the reader to their beliefs, if they share the same opinion. Social media has also provided a platform for radical and extremist political or religious groups to form, network, and collaborate to spread their messages of anti-establishment and anti-political correctness, and promote beliefs and ideologies that are racist, anti-feminist, homophobic, transphobic, etc. Fully-anonymous online communication is rare, as it requires the user to employ highly technical measures to ensure that they cannot be easily identified.

Cross-jurisdictional spread

A further complication is the transnational reach of the Internet, raising issues of cross jurisdictional co-operation in regard to legal mechanisms for combating hate speech. While there are Mutual Legal Assistance treaties in place across Europe, Asia, and North America, these are characteristically slow to work. The transnational reach of many private-sector Internet intermediaries may provide a more effective channel for resolving issues in some cases, although these bodies are also often impacted upon by cross-jurisdictional appeals for data (such as revealing the identity of the author(s) of a particular content). Different jurisdictions also have unique definitions for hate speech, making it difficult to prosecute perpetrators who may seek haven in less stringent jurisdictions.

Unlike the dissemination of hate speech through conventional channels, victims of online hate speech may face difficulties knowing to whom they should turn to help, as the platform, their local law enforcement, and the local law enforcement of the person or people using hate speech, may all feel as though the issue does not fall into their jurisdiction, even when hate speech policies and laws are in place. Nongovernmental organizations and lobby groups have been raising awareness and encourage different stakeholders to take action.

Artificial intelligence

Some tech companies, such as Facebook, use Artificial Intelligence (AI) systems to monitor hate speech. However, AI may not always be an effective way of monitoring hate speech, since the systems lack the judgment skills that humans have. For example, a user might post or comment something that classifies as hate speech, or violates community guidelines, but if the target word is misspelled, or some letters are replaced with symbols, the AI systems will not recognize it. This weakness has led to the proliferation of attempts to circumvent censorship algorithms using deliberate misspellings, such as the use of "vachscenes" instead of "vaccines" by anti-vaxxers during COVID-19. Therefore, humans still have to monitor the AI systems that monitor hate speech; a common problem in AI technology which is referred to as "Automation's Last Mile.", meaning the last 10% or 1% of the job is the hardest to complete.

Frameworks

''Stormfront'' Precedent

In the aftermath of 2014's dramatic incidents, calls for more restrictive or intrusive measures to contain the Internet's potential to spread hate and violence are common, as if the links between online and offline violence were well known. On the contrary, as the following example indicates, appearances may often be deceiving. Stormfront is considered the first "hate website." Launched in March 1995 by a former Ku Klux Klan leader, it quickly became a popular space for discussing ideas related to Neo-Nazism, White nationalism and White separatism, first in the United States of America and then globally. The forum hosts calls for a racial holy war and incitement to use violence to resist immigration. and is considered a space for recruiting activists and possibly coordinating violent acts. The few studies that have explored the identities of Stormfront actually depict a more complex picture. Rather than seeing it as a space for coordinating actions. Well-known extreme right activists have accused the forum to be just a gathering for "keyboard warrior