By Paul De Hert* and Andrés Chomczyk Penedo**
* Professor at Vrije Universiteit Brussel (Belgium) and associate professor at Tilburg University (The Netherlands)
** PhD Researcher at the Law, Science, Technology and Society Research Group, Vrije Universiteit Brussel (Belgium). Marie Skłodowska-Curie fellow at the PROTECT ITN. The author has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813497
1. Dealing with online misinformation: who is in charge?
Misinformation and fake news are raising concerns for the digital age, as discussed by Irene Khan, the United Nations Special Rapporteur on the promotion and protection of the right to freedom of opinion and expression (see here). For example, during the last two years, the COVID19 crisis caught the world by surprise and considerable discussions about the best course of action to deal with the pandemic were held. In this respect, different stakeholders spoke up but not all of them were given the same possibilities to express their opinion. Online platforms, but also traditional media, played a key role in managing this debate, particularly using automated means (see here).
A climate of polarization developed, in particular on the issue of vaccination but also around other policies such as vaccination passports, self-tests, treatment of the virus in general, or whether the health system should focus on ensuring immunity through all available strategies (see here). Facebook, YouTube, and LinkedIn, just to name a few, stepped in and started delaying or censoring posts that in one way or another were perceived as harmful to governmental strategies (see here). While the whole COVID19 crisis deserves a separate discussion, it serves as an example of how digital platforms are, de facto, in charge of managing online freedom of expression and, from a practical point of view, have the final say in what is permissible or not in an online environment.
The term 'content’ has been paired with adjectives such as clearly illegal, illegal and harmful, or legal but harmful, just to name the most relevant ones. However, what does exactly each of these categories entail, and why are we discussing these categories? What should be the legal response, if any, to a particular piece of content and who should address it? While content and its moderation is not a new phenomenon, as Irene Khan points in her previously mentioned report, technological developments, such as the emergence and consolidation of platforms, demand new responses.
With this background, the European Union is currently discussing at a surprisingly, very quick speed the legal framework for this issue through the Digital Services Act (the DSA, previously summarised here). The purpose of this contribution is to explore how misinformation and other categories of questionable content are tackled in the DSA and to highlight the option taken in the DSA to transfer government-like powers (of censorship) to the private sector. A more democratic alternative is sketched. A first one is based on the distinction between manifestly illegal content and merely illegal content to distribute better the workload between private and public enforcement of norms. A second alternative consists in community-based content moderation as an alternative or complementary strategy next to platform-based content moderation
2. What is the DSA?
The DSA (see here for the full text of the proposal and here for its current legislative status) is one of the core proposals in the Commission’s 2019-2024 priorities, alongside the Digital Markets Act (discussed here), its regulatory ‘sibling’. It intends to refresh the rules provided for in the eCommerce Directive and deal with certain platform economy-related issues under a common European Union framework. It covers topics such as: intermediary service providers liability - building up from the eCommerce Directive regime and expanding it -, due diligence obligations for a transparent and safe online environment -including notice and takedown mechanisms, internal complaint-handling systems, traders traceability, and advertising practices-, risk management obligations for very large online platforms and the distribution of duties between the European Commission and the Member States. Many of the these topics might demand further regulatory efforts beyond the scope of the DSA, such as political advertisement which would be complemented by sector-specific rules as, for example, the proposal for a Regulation on the Transparency and Targeting of Political Advertising (see here).
As of late November 2021, the Council has adopted a general approach to the Commission’s proposal (see here) while the European Parliament is still dealing with the discussion of possible amendments and changes to that text (see here). Nevertheless, as with many other recent pieces of legislation (see here), it is expected that its adoption is sooner rather than later in the upcoming months.
3. Unpacking Mis/Disinformation (part1): illegal content as defined by Member States
We started by discussing misinformation and fake news. If we look at the DSA proposal, the term 'fake news' is missing in all its sections. However, the concept of misinformation appears as disinformation in Recitals 63, 68, 69, and 71. Nevertheless, both terms are nowhere to be found in the Articles of the DSA proposal.
In literature, the terms are used interchangeably or are distinguished, with disinformation defined as the intentional and purposive spread of misleading information, and misinformation as ‘unintentional behaviors that inadvertently mislead’ (see here). But that distinction does not help in recognizing either mis- or disinformation, from other categories of content.
Ó Fathaigh, Helberger, and Appelman (see here) have pointed that disinformation, in particular, is a complex concept to tackle and that very few scholars have tried to unpack its meaning. Despite the different policy and scholarly efforts, a single unified definition of mis- or disinformation is still lacking, and the existing ones can be considered as too vague and uncertain to be used as legal definitions. So, where shall we start looking at these issues? A starting point, so we think, is the notion of content moderation, which according to the DSA proposal, is defined as follows:
'content moderation' means the activities undertaken by providers of intermediary services aimed at detecting, identifying, and addressing illegal content or information incompatible with their terms and conditions, provided by recipients of the service, including measures taken that affect the availability, visibility, and accessibility of that illegal content or that information, such as demotion, disabling of access to, or removal thereof, or the recipients' ability to provide that information, such as the termination or suspension of a recipient's account (we underline);
Under this definition, content moderation is an activity that is delegated to providers of intermediary services, particularly online platforms, and very large online platforms. Turning to the object of the moderation, we can ask what is exactly being moderated under the DSA? As mentioned above, moderated content is usually associated with certain adjectives, particularly illegal and harmful. The DSA proposal only defines illegal content:
‘illegal content’ means any information, which, in itself or by its reference to an activity, including the sale of products or provision of services is not in compliance with Union law or the law of a Member State, irrespective of the precise subject matter or nature of that law;
So far, this definition should not provide much of a challenge. If the law considers something as, it makes sense that it is similarly addressed in the online environment as in the physical realm. For example, a pair of fake sneakers constitute a trademark infringement, regardless of if the pair is being sold via eBay or by a street vendor in Madrid’s Puerta del Sol. In legal practice, regulating illegal content is not black and white. A distinction can be made between clearly illegal content and situations where further exploration must be conducted to determine the illegality of certain content. This is how it is framed in the German NetzDG, for example. In some of the DSA proposal articles, mainly Art. 20, we can see the distinction between manifestly illegal content and illegal content. However, this distinction is not picked up again in the rest of the DSA proposal.
What stands is that the DSA proposal does not expressly cover disinformation but concentrates on the notion of illegal content. If Member State law defines and prohibit mis- or disinformation -which Ó Fathaigh, Helberger and Appelman have reviewed and found to be inconsistent across the EU- , then this would fall under the DSA category of illegal content. Rather than creating legal certainty, this further reinforces legal uncertainty and pegs the notion of illegal content to be dependent on each Member State's provisions. But where does this leave disinformation that is not regulated in in Member State laws? The DSA does not like it, but its regulation is quasi hidden.
4. Unpacking Mis/Disinformation (part2): harmful content non defined by the DSA
The foregoing brings us to the other main concept dealing with content in the DSA, viz. harmful content. To say that this is a (second) 'main' concept might confuse the reader, since the DSA does not define it or regulate it at great lengths. The DSA’s explanatory memorandum states that `[t]here is a general agreement among stakeholders that ‘harmful’ (yet not, or at least not necessarily, illegal) content should not be defined in the Digital Services Act and should not be subject to removal obligations, as this is a delicate area with severe implications for the protection of freedom of expression’.
As such, how can we define harmful content? This question is not new by any means as we can trace back policy documents from the European Union dating back to 1996 (see here) dealing with this problem. Since then, little has changed in the debate surrounding harmful content as the core idea remains untouched: harmful content refers to something that, depending on the context, could affect somebody due to it being unethical or controversial (see here).
In this respect, the discussion on this kind of content does not tackle a legal problem but rather an ethical, political, or religious one. As such, it is a valid question to be asked if laws and regulations should even mingle in this scenario. In other words, does it make sense to talk about legal but harmful content when we discuss new regulations? Should our understanding of illegal and harmful content be construed in the most generous way to accommodate for the most amount of situations possible to avoid this issue? And more importantly, if the content seems to be legal, does it make sense to add the adjective of ‘harmful’ rather than using, for example, ‘controversial’? Regardless of the terminology used, this situation leaves us with three types of content categories: (i) manifestly illegal content; (ii) illegal, both harmful and not, content; (iii) legal but harmful content. Each of them demands a different approach, which shall be the topic of our following sections.
5. Illegal content moderation mechanisms in the DSA (content type 1 & 2)
The DSA puts forward a clear, but complex, regime for dealing with all kinds of illegal content. As a starting point, the DSA proposal provides for a general no monitoring regime for all intermediary service providers (Art. 7) with particular conditions for mere conduits (Art. 3), caching (Art. 4), and hosting service providers (Art. 5). However, voluntary own-initiative investigations are allowed and do not compromise this liability exemption regime (Art. 6). In any case, once a judicial or administrative order mandates the removal of content, this order has to be followed to avoid incurring liability (Art. 8). In principle, public bodies (administrative agencies and judges) have control over what is illegal and when something should be taken down.
However, beyond this general regime, there are certain stakeholder-specific obligations spread out across the DSA proposal also dealing with illegal content that challenge the foregoing state-controlled mechanism. In this respect, we can point out the mandatory notice and takedown procedure for hosting providers with a fast lane for trusted flaggers notices (Arts. 14 and 19, respectively), in addition to the internal complaint-handling system for online platforms paired with the out-of-court dispute settlement (Arts. 17 and 18, respectively) and, in the case of very large online platforms, these duties should be adopted following a risk assessment process (Art. 25). With these set of provisions, the DSA grants a considerable margin to certain entities to act as law enforcers and judges, without a government body having a say in if something was illegal and its removal was a correct decision.
6. Legal but harmful content moderation mechanisms in the DSA (content type 3)
But what about our third type of content, legal but harmful content, and its moderation? Without dealing with the issue of content moderation directly, the DSA transfers the delimitation of this concept to providers of online intermediary services, mainly online platforms. In other words, a private company can limit apparently free speech within its boundaries. In this respect, the DSA proposal grants all providers of intermediary services the possibility of further limiting what content can be uploaded and how it shall be governed via the platform’s terms and conditions and, by doing so, these digital services providers are granted substantial power in regulating digital behavior as they see fit:
‘Article 12 Terms and conditions
1. Providers of intermediary services shall include information on any restrictions that they impose concerning the use of their service in respect of information provided by the recipients of the service, in their terms and conditions. That information shall include information on any policies, procedures, measures, and tools used for content moderation, including algorithmic decision-making and human review. It shall be set out in clear and unambiguous language and shall be publicly available in an easily accessible format.
2. Providers of intermediary services shall act in a diligent, objective, and proportionate manner in applying and enforcing the restrictions referred to in paragraph 1, with due regard to the rights and legitimate interests of all parties involved, including the applicable fundamental rights of the recipients of the service as enshrined in the Charter.’
In this respect, the DSA consolidates a content moderation model heavily based around providers of intermediary services, and in particular, very large online platforms, acting as lawmakers, law enforcers, and judges at the same time. They are lawmakers as the terms and conditions lay down what is permitted as well as forbidden in the platform. While there isn't a general obligation to patrol the platform, they must react to notices from users and trusted flaggers and enforce the terms if necessary. And, finally, they act as judges by attending to the replies from the user who uploaded illegal content and dealing with the parties involved in the dispute, notwithstanding the alternative means provided for in the DSA.
Rather than using the distinction between manifestly illegal content and ordinary illegal content and refraining from regulating other types of content, the DSA creates a governance model for moderation of all content in the same manner. While administrative agencies and judges can request content to be taken down, under Art. 8, the development of the further obligations mentioned above poses the following question: who is the main responsible to define what is illegal and what is legal? Are the existing institutions subject to checks and balances or rather private parties, particularly BigTech and very large online platforms?
7. The privatization of content moderation: the second (convenient?) invisible handshake between the States and platforms
As seen with many other areas of the law, policymakers and regulators have slowly but steadily transferred government-like responsibilities into the private sector and mandated their compliance relying on a risk-based approach. For example, in the case of financial services, banks, and other financial services providers have turned into the long arm of financial regulators to tackle money laundering and tax evasion rather than relying on government resources to do this. This resulted in financial services firms having to process vast amounts of personal data to determine whether a transaction is illegal (either because it is laundering criminal proceedings or avoiding taxes) with nothing but their planning and some general guidelines; if they fail in this endeavor administrative fines (and in some cases, criminal sanctions) can be expected. The result has been an ineffective system to tackle this problem (see here) yet regulators keep on insisting on this approach.
A little shy of 20 years ago, Birnhack and Elkin denounced the existence of an invisible handshake between States and platforms for the protection and sake of national security after the 9/11 terror attacks (see here). At that time, this invisible handshake could be considered by some as necessary to deal with an international security crisis. Are we in the same situation as we speak when it comes to dealing with disinformation and fake news? This is a valid question. The EU policy makers seems to be impressed by voices such as Facebook’s whistleblower Frances Haugen who wants to align 'technology and democracy' by enabling platforms to moderate post. The underlying assumption seems to be that platforms are in the best position to moderate content following supposedly clear rules and that 'disinformation' can be identified (see here).
Content moderation presents a challenge for States given the amount of content generated non-stop across different intermediary services, in particular, social media online platforms (see here). Facebook employs a sizable staff of almost 15,000 individuals as content moderators (see here) but also relies heavily on automated content moderation, authorized by the DSA proposal under Arts. 14 and 17, in particular, to mitigate mental health problems to those human moderators given the inhuman content they sometimes have to engage with. To put this in comparison, using the latest available numbers from the Council of Europe about the composition of judiciary systems in Europe (see here), the Belgian judiciary employs approximately 9200 individuals (-the entire judiciary dealing with issues about commercial law up to criminal cases-), a little more than half of Facebook’s content moderators.
As such, one can argue that courts could be easily overloaded with cases that demand a quick and agile solution for defining what is illegal or harmful content if platforms didn't act as a first-stage filter for content moderation. Governments would need to heavily invest in administrative or judicial infrastructure and human resources to deal with such demand from online users. This matter has been discussed by scholars (see here). The available options they see either (i) strengthening platform content moderation by requiring the adoption of judiciary-like governance schemes, such as social media councils as Facebook has done; or (ii) implementing e-courts with adequate resources and procedures suited to the needs of the digital age to upscale our existing judiciary.
8. The consequences of the second invisible handshake
The DSA seems to have, willingly or not, decided on the first approach. Via this approach, -the privatization of content moderation-, States do not have to deal with the lack of judicial infrastructure to deal with the amount of content moderation that digital society requires. As shown by our example, Facebook has an infrastructure, just on raw manpower available, that doubles that of a country’s judiciary, such as Belgium. This second invisible handshake between BigTech and States can be situated in the incapacity of States to deal with disinformation effectively with the current legal framework and institutions.
If the DSA proposal is adopted ‘as is’, then platforms would have a significant power over individuals. First, through the terms and conditions, they would in position to determine what is allowed to be said and what cannot be discussed, as provided for by Art. 12. Not only that but also any redress before decisions adopted by platforms would have to be first channeled through the internal complaint handling mechanisms, as provided for by Arts. 17 and 18, for example, rather than seeking judicial remedy. As it can be appreciated, the power scale has clearly shifted towards platforms, and by extension to governments, in detriment of end-users.
Besides this, the transfer of government-like powers to platforms contributes to avoiding making complicated and hard decisions that could cost political reputation. Returning to our opening example, the lack of a concrete decision from our governments regarding sensitive topics has left platforms in charge of choosing what is the best course of action to tackle a worldwide pandemic by defining when something is misinformation that can affect the public health and when something could help fight back something that is out of control. Not only that but if platforms wrongfully approach the issue, then they are exposed to fines for non-compliance with their obligations, although particularly very large online platforms can deal with the fines proposed under the DSA.
If the second invisible handshake is going to take place, the least we, as a society, deserve is that agreement is made transparent so that public scrutiny can oversight such practices and free speech can be safeguarded. In this respect, the DSA could have addressed the issue of misinformation and fake news in a more democratic manner. Two proposals:
9. Addressing disinformation more democratically to align 'technology and democracy'
Firstly, the distinction between manifestly illegal content and merely illegal content could have been extremely helpful in distributing the workload between the private and public sector in a manner that administrative authorities and judges would only take care of cases where authoritative legal interpretation is necessary. As such, manifestly illegal content, such as apology to crime or intellectual property infringements, could be handled directly by platforms and merely illegal content by courts or administrative agencies. In this respect, a clear modernization in legal procedures to deal with claims about merely illegal content would still be necessary to adjust the legal response time to the speed of our digital society. Content moderation is not alone in this respect but joins the ranks of other mass-related issues, such as consumer protection, where effective legal protection is missing due to the lack of adequate infrastructure to channel complaints.
Secondly, as for legal but harmful content, while providers of online intermediary services have a right to conduct their business as to how they see fit and therefore can select which content is allowed or not via terms and conditions, citizens do have a valid right to engage directly in the discussion of those topics and determine how to proceed with them. This is even more important as users themselves are the ones interacting on these platforms and that content is exploited by platforms to ensure that controversy remains on the table to ensure engagement (see here).
However, there is a possibility to deal with content moderation, particularly in the case of legal but harmful content, that avoids a second invisible handshake: community-based content moderation strategies (see here) where users have a more active role in the management of online content has proven to be successful in certain online platforms. While categories such as clearly illegal or illegal and harmful content do not provide much margin for societal interpretation, legal but harmful content could be tackled by citizens' involvement. In this respect, community-based approaches, while resource-intensive, allow for citizens to engage directly in the debate about the issue at hand.
While community-based content moderation also has its own risks, it could serve as a more democratic method than relying on platforms’ unilateral decisions and it might serve where judges and administrative agencies cannot go due to the legality of content. As noted by the Office of the United Nations High Commissioner for Human Rights, people, rather than technology, should be making the hard decisions but also States, as elective representatives of society, need to make decisions about what is illegal and what is legal (see here).
Our alternatives are only a part of a more complete program. Further work is needed at policy level to address fake news. Sadly, as it may be, the matter is not matured yet and ripe for regulation. While the phenomena of political actors actively spreading misleading information (the twittering lies told by political leaders) are well-known and discussed, the role of traditional news media, who are supposed to be the bearers of truth and factual accuracy, is less well understood. Traditional news media are in fact a part of the problem, and play a somewhat paradoxical role with respect to fake news and its dissemination. People learn about fake news, not via obscure accounts that Facebook and others can control, but through regular media that find it important for many reasons to report on disinformation. Tsfatie and others (see here) rightly ask for more analysis and collaborations between academics and journalists to develop better practices in this area.
We are also surprised by the lack of attention in the DSA proposal to the algorithmic and technological dimension that seems central to the issue of fake news. More work is needed on the consequences of algorithmic production of online content. More work too is needed to assess the performance of technological answers to technology. How to organize a space of contestation in a digitally mediated and enforced world? Are the redress mechanisms in the DSA sufficient when the post has already been deleted, i.e. "delete first rectify after"?
Art credit: Frederick Burr Opper, via wikimedia commons