DIGITAL HUMAN RIGHTS

By Lucas Alcici

December 19, 2020

Issues arising from content moderation on social media

The internet, and social media in particular, has become the main platform for the exchange of information and ideas. Every single day an enormous amount of content is published and shared on social media, on which more than 3.6 billion users are registered. Due to this ever-growing flow of information and ideas on their platforms, social media companies are increasingly implementing content moderation systems in order to scrutinize what users publish online.

Those systems may take down, flag or limit the reach of a particular content or even suspend or block users’ accounts and, therefore, risk imposing unlawful restrictions on freedom of expression. There is a greater risk of this happening when artificial intelligence (AI) technologies are used to help moderate content online, especially when dealing with controversial content, as algorithms lack the capability of evaluating context, spotting irony or conducting the complete analysis which is necessary to identify when a particular content should be restricted. For this reason, content moderation systems that rely heavily or exclusively on AI tools are more likely to block or restrict content by default, possibly violating the right of users to seek, receive and impart ideas and information, as already pointed out by the UN Special Rapporteur on the promotion and protection of the right to freedom of opinion and expression.

The risk of content moderation systems which rely on AI tools imposing unlawful restrictions on freedom of expression can be illustrated by Instagram’s AI tool DeepText identifying the word “Mexican” as a slur, as its datasets were populated with data in which the word “Mexican” was often associated with the word “illegal”, probably due to hateful comments found on the internet.

Online content moderation may involve a complex process, in which the context of the content posted on social media should often be taken into account. For this reason, AI tools used for online content moderation shall be designed and employed in a way that respects users’ right to freedom of expression.

International human rights law and restrictions to freedom of expression

The right to freedom of expression is essential to democracy and to the protection of all human rights. Universal and regional human rights treaties, such as the International Covenant on Civil and Political Rights (ICCPR), the European Convention on Human Rights (ECHR) and the American Convention on Human Rights (ACHR), guarantee a special protection to freedom of expression.

Regardless of the fundamental role played by freedom of expression in a democratic society, this right is not absolute and may be subject to limitations provided certain conditions are met. According to international human rights law, a restriction to freedom of expression must meet, cumulatively, the requirements of legality, legitimacy, necessity and proportionality.

What does that mean? It means that restrictions to freedom of expression shall be provided by law, pursue a legitimate aim and be necessary to achieve this particular legitimate aim. Furthermore, the restriction shall be the least restrictive measure capable of achieving the legitimate aim pursued.

Each of the above-mentioned treaties presents a list of the legitimate aims which may justify a restriction on freedom of expression in its own system. For instance, the ICCPR and the ACHR establish as legitimate aims: the respect of the rights or reputation of others, the protection of national security and public order and the protection of public health and morals. The ECHR presents a more expansive list, as it includes the prevention of the disclosure of information received in confidence and the maintenance of the authority and impartiality of the judiciary as additional legitimate objectives capable of justifying a restriction to freedom of expression in the European system of human rights.

Take the protection of public health as an example. Freedom of expression might be restricted, provided all cumulative requirements are met, to meet the overriding public interest of limiting misinformation and fake news about the pandemic, as there is an imperative public interest that people must have correct and precise information about the pandemic.

Two social media companies took different approaches towards a particular content posted by Donald Trump related to the current pandemic. This content was regarded as “misleading and potentially harmful information” by Twitter and “incorrect information” by Facebook. While Facebook decided to delete Trump’s post which claimed COVID-19 was “less lethal” than the flu, Twitter decided to hide the same message behind a warning label but did not delete Trump’s tweet. If Twitter's restrictive measure of flagging the post and adding a warning label proves to be sufficient to the protection of public health, we may conclude that this is the least restrictive measure capable of achieving this particular aim. Therefore, Facebook's restrictive measure would not meet the proportionality requirement. Twitter also opted to add a warning label to a tweet from Donald Trump in which he claimed to be immune and no longer contagious after having been infected with the coronavirus, saying the tweet contained “misleading and potentially harmful information related to COVID-19”. Twitter believes the tweet should remain accessible because it "may be in the public's interest".

Social media companies shall comply with international human rights law when moderating content

International human rights treaties create legal obligations to sovereign states, but do not directly impose obligations upon private companies, such as social media companies. For this reason, the adoption of the Guiding Principles on Business and Human Rights by the United Nations in 2011 was an important milestone. Not only do they reinforce the duty of states to “protect against human rights abuse within their territory and/or jurisdiction by third parties, including business enterprises”, but also call upon private companies to respect internationally guaranteed human rights, amongst which the right to freedom of expression, wherever they operate.

In order to comply with the standards of protection laid down by international human rights law to freedom of expression, social media companies shall take a human rights-based approach to content moderation. They shall also carry out human rights due diligence in order to identify, prevent and mitigate adverse impacts their content moderation system might have on freedom of expression.