September 9, 2020

EU DisinfoLab joins 28 civil society organisations in calling for for meaningful transparency on all ads

Printable PDF version here

The political campaigning landscape has changed significantly with the digitalisation of our public sphere, which has created new opportunities for political participation, but also poses significant risks to the integrity of elections and political debate. Unlike broadcast political ads shown to the wider public, online ads are tailored to specific homogenous groups of people, which can segment and polarise the voter base and distort political debate. Advertisers can purchase exorbitant amounts of ads and flood people’s social media feeds, thereby buying themselves space in public policy and political debates. The lack of transparency of which ads are shown to whom, why, and who has paid for them, further creates a situation where anyone – from a political party and interest group to a foreign advertising firm like Cambridge Analytica – can distort political debate and easily evade public interest scrutiny. This threatens the credibility of our electoral processes, and ultimately the legitimacy and representativeness of our democracies.

At the source of these problems lies the lack of transparency offered by digital platforms such as social media sites, video apps and search engines. While some platforms have found ways to provide some transparency on political ads (partly due to pressure by the European Commission), their voluntary measures fall short of providing meaningful transparency. One crucial weakness of the status quo is that it leaves platforms to decide what is and is not political advertising – and thus, what advertising will and will not be subjected to platforms’ transparency regimes. To avoid this issue and to recognise the kind of behavioural targeting and algorithmic delivery that underlies all types of social media advertising, it is necessary to require meaningful default transparency for all ads.

Why transparency by default for all ads? 

To allow for public interest scrutiny: Transparency is necessary, first and foremost, to allow for public scrutiny of advertising. As many studies on the implementation of the EU Code of Practice against Disinformation have shown, false negatives and false positives were rife in the political ad libraries of the signatories of the code: non-political advertisements were erroneously included in the libraries, while many political ads were excluded. The lack of a comprehensive repository of all ads made it impossible to verify whether all political ads were included in the libraries, and the political ad libraries and labelling missed a lot of sponsored content. In a situation where it is difficult to police the labelling of political ads, it is ultimately necessary to ensure the transparency of all ads. 

To overcome diverging definitions of political ads: EU member states have diverging definitions of political advertising, and some have no definition at all. The EU Code of Practice on Disinformation distinguishes between political and issue-based advertising, which introduces a distinction that is not reflected uniformly across member states’ electoral laws. Introducing mandatory transparency of all advertising helps to address the difficulty of adopting and applying one common definition of political advertising.

To verify the labelling and disclaimers of political ads: Full, meaningful transparency is the only way to verify if political content is labelled and regulated as such, and it allows civil society and other watchdogs to monitor the grey zone between political and commercial ads. Past experience with the Code of Practice has shown that the platforms often incorrectly categorise and label political ads. As the forms of political advertising online will undoubtedly evolve as the technology changes, full transparency creates enough flexibility to account for such changes. Moreover, “political” ads are not the only ones that should be subject to scrutiny and accountability: false or deceptive advertising, hoaxes, and paid disinformation (notably pertaining to public health, in the current pandemic context) should also be subject to scrutiny.

To better understand malign actors: In addition, full transparency of paid-for content will allow for better identification and a deeper understanding of other malign actors’ strategies. Currently, it is very easy for malign actors to get into the political campaigning eco-system and hijack the political debate to their own ends in ways that are not possible on television or through other advertising channels. Ads can be used to lure people into Facebook groups that are not initially about a political issue, but eventually become focused on a political cause and are used for malign purposes. Investigations show the platforms’ inability to enforce their own policies in this regard. Researchers, civil society and journalists need access to an archive to understand the marketing techniques, networks and origin of these actors. 

To protect consumers and strengthen businesses: For commercial advertising, transparency by default benefits both brands and users. Universal ad transparency will help combat discriminatory and potentially illegal advertising practices, and help ensure compliance with privacy and data protection laws as they apply to ad targeting. At the same time, transparency also helps protect consumers – particularly those from vulnerable groups – from advertising for illegal and harmful products, and potentially increases trust in brands and in the platforms. Businesses that act in good faith and comply with regulation (including the GDPR) also benefit, as transparency levels the playing field by preventing bad-faith advertisers from breaking the law with impunity. Transparency on the advertiser, engagement and targeting criteria is only one part of a company’s advertising strategy and therefore does not imply a disclosure of trade secrets. Transparency mechanisms would have to be built into the platforms in an easy-to-use format so that it doesn’t prove a burden for advertisers. 

Such public-facing transparency is a necessary yet in itself insufficient first step for enhancing the accountability of platforms and advertisers. While the measures described below will enhance the transparency of advertisers, this needs to go alongside transparency of the ad optimisation processes on the part of the platforms, as well as user-level transparency explaining why exactly an ad is reaching them individually. Transparency in and of itself is only instrumental to accountability and needs to therefore be backed up with further action to safeguard rights and democratic processes online. For example, transparency may reveal widespread harmful practices that may in fact be prohibited but have escaped meaningful enforcement, or even novel practices that should be regulated.

What universal transparency by default looks like

Mandatory, functional ad libraries: The European Commission should foster the development of and issue minimum technical standards for advertisement libraries for digital platforms, covering both the design and functioning of ad libraries. These minimum technical standards should be developed through a multistakeholder process, and help overcome the numerous problems and bugs that render the existing ad libraries meaningless as transparency tools. The ad repositories should comply with well-defined accessibility and technical standards standards. We suggest the following starting point for these standards: 

  • Libraries provided by each platform should be compatible with each other. There should be a unique set of standards and protocols provided by the Commission that all platforms are required to use.
  • Platforms should assign unique identifiers to each advertisement and advertiser to allow for trend analysis over time and across platforms. Advertisers should keep the same unique identifier no matter what platform they’re using. 
  • All images, videos, and other content should be provided in a machine-readable format accessible via an application programmatic interface. This should include any words used in images or in audio provided as searchable text.
  • The ability to download a week’s worth of data in less than 12 hours and a day’s worth of data in less than two hours.
  • Bulk downloading functionality of all relevant content. It should be feasible to download all historical data within one week.
  • Search functionality by the text of the content itself, by the content author or by date range.
  • Up-to-date and historical data access, including the availability of advertisements within 24 hours of publication; the availability of advertisements going back 10 years. In addition, APIs should be promptly fixed when they are broken and APIs should be designed so that they either support or at least do not impede long-term studies
  • The API itself and any data collected from the API should be accessible to and shareable with the general public. 
  • The ad libraries must be free of charge and shared under a permissive open source licence.  
  • The ad libraries should include clear audit trails for content which has been removed, including the reasoning for its removal while maintaining data on the advertiser, funder, spend, and targeting. 

Such ad libraries should become mandatory for platforms from a set number of users onwards, to be decided by a European-level regulator or coordination platform between national regulators, and reviewed on a yearly basis. 

Such public advertising libraries, which again should include commercial advertising as well as “political advertising” (however defined), must disclose the following information at minimum

  • Exact spend: broad spend ranges like 0-100 EUR, 100-1,000 EUR, … are not meaningful information for users and researchers. Enhanced transparency on all aspects of online targeting – including the amount spent – is a necessary price to pay for the increased customer access advertisers gain with behavioural targeting practices. 
  • Advertiser information: this needs to be accurate and complete. Third parties, like advertising agencies, who advertise on behalf of another entity need to be as transparent as the brand or entity that commissions the advertisement. Information on both the third party and the political candidate or party needs to be detailed in the ad library and the disclaimer. Information on the funding entity should also be disclosed and verified. 
  • Advertiser identification: Platforms should facilitate the linkage with other databases that support verification by displaying official identification such as corporate registers, advertisers’ tax ID, political candidates’ electoral court declaration, or any other identification number that facilitates enforcement and verification of the advertisers’ identity. 
  • Targeting mechanism: use of lookalike audiences, and which audiences they chose; use of profiling based on imported datasets and the source of this data (such as a newsletter platform, for instance); or other similar mechanisms to improve targeting by the platforms. 
  • Targeting and delivery criteria, with the same level of granularity as the advertiser can choose from. This must include the optimisation goal selected by the advertiser and general information on the optimisation logic used by the platform (possibly in another layer/interface that is accessible from the ad library).
  • Audience reached
  • Engagement and reach in absolute and relative terms, e.g. likes, shares, comments
  • The ad library should include this information for all ads, including the ads taken down by the platforms because they did not adhere to community standards. With the exception of content judged illegal by the relevant state authorities, banned ads should remain in the ad repository for public scrutiny. For ads that were taken down, information about the kind and category of content, and the reasons and process for take-down should be displayed.
  • None of these measures should reveal personal user information and all of them should be GDPR compliant. 

Real-time transparency disclosures for individual users: There should be clear, consistently applied on-screen designations of ads, distinguishing them from other content. Users should have easy access to easily comprehensible basic transparency information, as well as easy access to the above-mentioned transparency information. Furthermore, access to a personal ad library showing users who is targeting them and how, would allow users to better hold platforms to account. The design of this access and the information presented should be at least as accessible as other platform features. Companies should show evidence of the design process and provide information on user interactions with it on request.

Verification of advertisers: Platforms and political advertisers need to be held to account for verifying all advertisers’ real identity, who’s paying (indirectly) for the ads, contact details and for political advertisers a reference to their declaration to the electoral authorities (when applicable in the country context). Such verification needs to be quick and mandatory. It should not rely on self-declaration by the advertiser but require the platforms to verify the information provided. It also needs to be more closely monitored by authorities, to ensure platforms perform better than they did as part of their efforts for the Code of Practice, with appropriate sanctions available for advertisers and platforms that do not stick to the rules. 

Anonymity where needed to protect safety: We encourage the European Commission to issue guidelines for platforms to protect advertisers in high risk contexts. The European Commission could make suggestions for a mechanism for advertisers to anonymise their identity on the basis of political threats and risk, for public interest actors such as human rights defenders and activists. Such an anonymity mechanism would for instance protect organisations raising awareness on LGBTQ+ rights in countries where those rights cannot be taken for granted. This mechanism, meant to protect those in need of anonymity, could be abused as a loophole by advertisers trying to hide their identity, even though they do not have to fear prosecution. Therefore, consideration should be given to independent mechanisms to oversee the granting of anonymity. Exemption applications should be carefully scrutinised according to a transparent set of criteria and information should be made publicly available on the number of exemptions requested and granted on an annual/quarterly basis.

Binding requirement and enforcement: The Commission should develop a mechanism for ensuring universal transparency for online advertising meets the standards set out above. This should include relevant penalties for non-compliance up to and including preventing a platform from running any ads if their efforts in this area are deemed insufficient.

Signatories 

Access Now

Algorithm Watch

AMO – Association for International Affairs

Civil Liberties Union for Europe

Defend Democracy

ePanstwo Foundation

EU Disinfolab

European Association for Local Democracy (ALDA)

European Citizen Action Service (ECAS)

European Digital Rights (EDRi)

European Partnership for Democracy (EPD)

European Women’s Lobby

Global Forum for Media Development (GFMD)

Global Witness

Homo Digitalis

Irish Council for Civil Liberties

Lie Detectors

Oana Goga (Centre National de la Recherche Scientifique — CNRS)

Open Rights Group

Panoptykon Foundation

Privacy International

Ranking Digital Rights 

Rights International Spain

Stiftung Neue Verantwortung (SNV)

The Democratic Society

Transparency International

Vrijschrift

Who Targets Me

World Leadership Alliance – Club de Madrid