The Obligation of Social Platforms to Remove Hate Speech and Illegal Speech

The obligation of social platforms to remove hate speech and illegal speech enshrined in German law and in Europe’s 2016 Code of conduct of countering illegal hate speech online. Hate existed before technology, but the emergence of the social platforms from the development of the internet created a new dimension to the complex hate speech topic.

Social platforms facilitate borderless communication, allowing for artistic, cultural, ideological, and political expression. More so, it allows the inflow of daily news while raising awareness on human rights violations. The platforms also allow the spread and normalization of hateful orotundity and strategic targeting of the minority groups (Alkiviadou, 2018). Alongside technological developments, hateful expression from internet use is a significant human rights challenge. A large number of social media users nationally makes it necessary to pay attention to the digital system. More so, the social media websites contain organized and semi-organized groups which focus on promoting hateful statement while at the same time targeting a particular group of individuals. Also, social media platforms contain content regulation frameworks that are assessed to determine the degree ti which the regulation facilitate appropriate dealing with online hate. The legislation designed to deal with aspects of online hate is challenging to use on the current offenses. The current offenses based on online hate speech and illegal material fail to describe the real attribute oh the internet since the crimes developed before the use of the internet or not to protect hate speech victims (Bakalis, 2018). Hence, estimate dictates that the portion of online hate that is investigated amounts to about 9 %.

 

History of the Obligation of Social Platforms to Remove Hate Speech and Illegal Speech:

Social media platforms are associated with significant, high-profile instances where racist material and fake news spread through German arms of powerful social platforms. Germany created the Network Enforcement Act or Netzwerkdurchsetzungsgesetz (NetzDG) law to deal with online hate expression coming into effect on 1 January 2018 (BBC,2018; Echikson & Knodt, 2018). The law contains fines of up to 50 million for online platforms in case of failure to delete illegal content (Echikson & Knodt, 2018). The strategic decision to deal with Xenophobia and Racism aims to criminalize public incitement to hatred and violence against individuals defined by reference to ethic, descent, or narional origin, religion, color or race (Press Corner, 2020).

In this regard, the Code of Conduct on countering illegal hate speech online was presented on 31 May 2016 by the European Commission and four huge information technology companies namely YouTube, Twitter, Microsoft and Facebook (Press Corner, 2020). The code of conduct in attained through cooperation between national authorities, civil society organizations (CSOs), IT platforms, and the European Commission. The commission conducts monitoring exercise to determine the effectiveness of the code of conduct with the most recent conducted from 4 November to 13 December 2019 in six weeks by 34 CSOs and five public bodies.

Figure 1 : Pictorial presentation of the call to end hate speech

source: European Commission 2020, All rights reserved 

The social media websites are required by the law to remove any materials that are ate related within one day with complex cases taking a week to act on it. For instance, Facebook reportedly employed several staff with the responsibility of dealing with reports of content not adhering to the NetzDG and to increase the effectiveness of the platforms in monitoring posts by individuals (BBC,2018). Additionally, the Justice ministry in Germany said it would develop forms on its site which allows the users of the social platforms to report any content going against the NetzDG or has not been removed on time (BBC,2018). The NetzDG also offers guidelines for the social media platforms to develop a comprehensive complains system where staff can be quickly notified on content associated with hate. The code of conduct by the European commission’s provides the necessary guidelines to help the social platforms to respond rapidly in spotting and removing hateful content.

 

Automated Content Moderation Tools

Natural Language Processing:

The process of parsing text in this content employs the NLP techniques where the text is parsed to predict the text; for instance, the sentiment the text indicates. Parsing develops structures of syntax from the text-based on PoS models and tokens. The parsing algorithm takes into consideration the grammar of the text for syntactic structuring (Nallagatla, 2019). The social platforms can use the NLP tools in content filtering in the case of hate speech and illegal content posted online. The NLP classifiers are utilized in the case of detecting extremist content, hate speech and conduct the analysis of the sentiment on content determining the case, and ultimately red-flagging them to the staff (New America, n.d.). The recruitment of NLP classifiers focuses on training them regarding the text example called documents which have been annotated by human to indicate and categorize them to a specific category such as hate content versus not hateful content. A set of documents called a corpus is provided to a model which operates with identifying and patterns linked with every annotated class (New America, n.d.).

The corpora are pre-processed to enable the numerical representation of specific attributes in a text, for example, the missing of a particular word (New America, n.d.). The pre-processed and annotated text documents are essential in training machine learning models used in classifying new documents (New America, n.d.). Furthermore, the classifier is tested on a different sample of data used in training to determine the degree the classifications of the models match with those of human coders. However, there are some limitations associated with this tool of content moderation. They are specific to a domain where they address a particular form of objectionable content which makes it challenging to address all kinds of hate content and illegal content (New America, n.d.). Also, gathering and compiling enough comprehensive sets of data to be used in training the NLP classifiers in a tedious, expensive, and challenging process which makes it different for the online platforms for adopting its use in handing hate content (New America, n.d.). Sentence tokenization separates a collection of written language into sentences components illustrates in figure 2 and its output (Yordanov, 2019).

Figure 2: Application of sentence tokenization nltk.sent_tokenize function

source: Yordanov, 2019, Some rights reserved.

The output provides three components in sentences separately

source: Yordanov, 2019, Some rights reserved.

Word tokenization/segmentation creates component words by separating a string of written language. Figure 3 demonstrates word segmentation which can be useful in detecting words which target a specific group of individuals.

Figure 3: Word segmentation using the nltk.word tokenized function

source: Yordanov, 2019, Some rights reserved.

The output of the process is explained below with the word segmentation easily analyzed

source: Yordanov, 2019, Some rights reserved.

 

Findings on Regulating online Hate

eSafety findings on the support of people regarding the regulation of online hate speech were positive. According to Table 1, more than 60 percent of Australians support the introduction of further specific regulations to curb the spread of online content that is harmful to an individual (eSafety,2020). In the same manner, the NetzDG offers guideline for the social media platforms to develop a comprehensive complains system where staff can be quickly notified on content associated with hate, the Australian government collaborating with eSafety can ensure the social media websites put in additional effort. Reports indicate that social media platforms are not doing enough to deal with the increase in online hate speech. From their report, around one in seven which is 14% of adults aged between 18 to 65 are estimated to be targeted by online hate speech during the research period in 12 months ending in august 2019 (eSafety, 2020).

Furthermore, individuals identifying as Torres Strait Islander, Aboriginal or LGBTQI expensive incidences of hate in social platforms twice as much as the national average. Specifically, individuals targeted by online hate speech usually cite the reason to be their nationality, ethnicity, race, gender, religion and political views (eSafety, 2020). Established social media systems, including Facebook and Instagram, are the areas prone to online hate. Considering the impacts of the hate has on the individuals, the social media platforms need to more and also the government need to develop a framework to deal with the issue.

Table 1: The attitudes of Australians towards online hate speech as of August 2019

 

source: eSafety.gov.au, All rights reserved.

Table 2: Individual reasons for online hate speech

source: eSafety.gov.au, All rights reserved.

 

References

Alkiviadou, N. (2018). Hate speech on social media networks: Towards a regulatory framework? Information & Communications Technology Law, 28(1), 19-35.

https://www.tandfonline.com/doi/full/10.1080/13600834.2018.1494417

Artical 19.(2018). Self-regulation and ‘hate speech’ on social media platforms. Article 19: Free Word Centre.

https://www.article19.org/wp-content/uploads/2018/03/Self-regulation-and-%E2%80%98hate-speech%E2%80%99-on-social-media-platforms_March2018.pdf

Bakalis, C.(2018). Rethinking cyber-hate laws. Information & Communications Technology Laws, 27(1), 86-110.

https://www.tandfonline.com/doi/full/10.1080/13600834.2017.1393934

BBC. (2018, 1 January). Germany starts enforcing hate speech law. BBC News.

https://www.bbc.com/news/technology-42510868

Echikson, W., & Knodt, O. (2018). Germany’s NetzDG: A Key Test for Combatting Online Hate. CEPS Policy Insight.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3300636

eSafety. (2020). Online hate speech: Findings from Australia, New Zealand, and Europe.

https://www.esafety.gov.au/sites/default/files/2020-01/Hate%20speech-Report.pdf

Nallagatla, D. (2019, 3 June). Using text analytics and NLP: An introduction. Transforming Data with Intelligence.

https://tdwi.org/articles/2019/06/03/adv-all-introduction-to-using-text-analytics-and-nlp.aspx

New America. (n.d.). Everything in moderation.

https://www.newamerica.org/oti/reports/everything-moderation-analysis-how-internet-platforms-are-using-artificial-intelligence-moderate-user-generated-content/how-automated-tools-are-used-in-the-content-moderation-process/

Press Corner. (2020, 22 June). Commission publishes EU Code of  Conduct on countering illegal hate speech online continues to deliver results. European Commission – European Commission.

https://ec.europa.eu/commission/presscorner/detail/en/ip_20_1134

Yordanov, V. (2019, 13 August). Introduction to natural language processing for text. Medium.

https://towardsdatascience.com/introduction-to-natural-language-processing-for-text-df845750fb63