Facebook and Google: This is What an Effective Ad Archive API Looks Like

Mozilla and a cohort of independent researchers have detailed the key traits that make for an effective ad archive API — and more transparent elections

On March 28 — after urging from dozens of civil society organizations — Facebook is set to launch its advertising archive API. This tool is intended to provide researchers, journalists, and users with transparency into political ads and audience targeting on Facebook.

Google also pledged to launch a similar tool ahead of the May 2019 EU Parliamentary elections (but postponed their initial March launch date.) As disinformation continues to spread across online platforms with the potential to interfere with democratic elections, it’s critical that these tools are accessible and effective.

So today, Mozilla and a cohort of 10 independent researchers are publishing five guidelines that these APIs must meet in order to truly support election influence monitoring and independent research.

Says EU Code of Practice on Disinformation. The guidelines are also being shared with Facebook and Google.

The guidelines + our letter to policymakers and the tech companies:

To: Google, Facebook, Twitter

Cc: Vice President of the European Commission Andrus Ansip, European Commissioners Mariya Gabriel, Vera Jourova, Julian King

We, the undersigned, are independent researchers investigating the wide variety of issues that are crucial to understanding the impact of disinformation on our societies. This includes research into:

• inauthentic accounts and behaviour on social media,
• political and issue-based advertising,
• micro-targeting practices and ad placements by political parties and other entities,
• the effectiveness of self-regulation measures to counter disinformation
.

To do this work effectively, there must be fully functional, open APIs that enable advanced research and the development of tools to analyse political ads targeted to EU residents. This requires access to the full scope of data relevant to political advertising, and that access must be provided in a format that allows for rich analysis. Tools provided often lack the necessary data or, due to limited functionality, do not allow for analysis.

A functional, open API should have the following:

[1] Comprehensive political advertising content. The APIs should include paid political ads and issue-based ads, without limiting access on the basis of pre-selected topics or keywords. “Political” ads might include, but are not limited to:

• direct electioneering content
• candidates or holders of political office
• matters of legislation or decisions of a court
• functions of government

Non-paid, public content that is generated by users who are known political content purveyors should also be available.

[2] The content of the advertisement and information about targeting criteria, including:

• The text, image, and/or video content and information about where the ad appeared (newsfeed, sidebar, etc.).
• The targeting criteria used by advertisers to design their ad campaign, as well as information about the audience that the ad actually reached.
• The number of impressions that an ad received within specific geographic and demographic criteria (e.g. within a political district, in a certain age range), broken down by paid vs. organic reach.
• The amount of engagements that an ad received, including user actions beyond viewing an ad.
• Information about how much an advertiser paid to place the ad.
• Information about microtargeting, including whether the ad was a/b tested and the different versions of the ad; if the ad used a lookalike audience; the features (race, gender, geography, etc.) used to create that audience; if the ad was directed at platform-defined user segments or interests, and the segments or interests used; or if the ad was targeted based on a user list the advertiser already possessed.

[3] Functionality to empower, not limit, research and analysis, including:

• Unique identifiers associated with each advertisement and advertiser to allow for trend analysis over time and across platforms.
• All images, videos, and other content in a machine-readable format accessible via a programmatic interface.
• 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.

[4] Up-to-date and historical data access, including:

• Availability of advertisements within 24 hours of publication.
• Availability of advertisements going back 10 years
• APIs should be promptly fixed when they are broken
• APIs should be designed so that they either support or at least do not impede long-term studies

[5] Public access. The API itself and any data collected from the API should be accessible to and shareable with the general public.

In the spirit of upholding the EU Code of Practice on Disinformation, we expect you to empower the research community by implementing open, functional APIs of the quality outlined in this letter — just as we expect elected officials and public authorities to fully support the release of such data in a privacy-compliant fashion to enable independent research and inform public debate. Your action on this is essential to ensuring the integrity of the upcoming European Parliamentary elections — as well as elections happening all around the globe — is upheld.

Yours Sincerely,

Mozilla Foundation

Co-written by

1. Jef Ausloos (University of Amsterdam, NL)
2. Jason Chuang (Mozilla)
3. Chloe Colliver (Institute for Strategic Dialogue, UK)
4. Laura Edelson (NYU Tandon School of Engineering, US)
5. Sasha Havlicek (Institute for Strategic Dialogue, UK)
6. Natali Helberger (University of Amsterdam, NL)
7. Stefan Heumann (Stiftung Neue Verantwortung, GER)
8. Sam Jeffers (Who Targets Me, UK)
9. Rasmus Nielsen (University of Oxford, Reuters Institute for the Study of Journalism, UK)
10. Alex Sängerlaub (Stiftung Neue Verantwortung, GER)
11. Michael Veale (University College London and University of Birmingham, UK)
12. Mathias Vermeulen (Vrije Universiteit Brussel, BE + Mozilla Foundation)

Co-signed by

1. Amelia Acker (University of Texas at Austin, US)
2. Alexandre Alaphilippe (EUDisinfoLab, BE)
3. Jonathan Albright (Columbia University, US)
4. Isabelle Augenstein (University of Copenhagen, DK)
5. Dominik Batorski (University of Warsaw, PL)
6. Anja Bechmann (Datalab, University of Aarhus, DK)
7. Reuben Binns (University of Oxford, UK)
8. Tobias Blanke (King’s College London, UK)
9. Kalina Bontcheva (University of Sheffield, UK)
10. Ruben Bouwmeester (Deutsche Welle, DE)
11. Elda Brogi (European University Institute, IT)
12. Axel Bruns (Queensland University of Technology, AUS)
13. Paul-Olivier Dehaye (Personaldata.Io, BE)
14. Leon Derczynski (IT University of Copenhagen, DK)
15. José van Dijck (Utrecht University, NL)
16. Marius Dragomir (Central European University, HU)
17. Charles Ess (University of Oslo, NO)
18. Aline Franzke (University Duisburg-Essen, DE)
19. Erika Franklin Fowler (Wesleyan University, US)
20. Michael Franz (Bowdoin College, US)
21. Frederic Guerrero-Solé (Universitat Pompeu Fabra of Barcelona, ES)
22. Jakub Grue Simonsen (University of Copenhagen, DK)
23. Luca Hammer (University of Paderborn, GER)
24. Jaron Harambam (University of Amsterdam, NL)
25. Anne Helmond (University of Amsterdam, NL)
26. Phil Howard (Oxford University, UK)
27. Vladan Joler (Share Lab, US)
28. Steve Jones (University of Illinois at Chicago, US)
29. Brian Keegan (University of Colorado, Boulder, US)
30. Aleksi Knuutila (Open Knowledge, FI)
31. Aleksandra Korolova (University of Southern California, US)
32. Alex Krasodomski (Demos, UK)
33. Philip Kreißel (Ich bin here, DE)
34. Aleksandra Kuczerawy (University of Leuven, CITIP, BE)
35. Paddy Leersen (University of Amsterdam, NL)
36. Christina Lioma (University of Copenhagen, DK)
37. Sílvia Majó-Vázquez (University of Oxford, UK)
38. James Meese (University of Technology Sydney, AUS)
39. Divina Meigs (Sorbonne Nouvelle University, FR)
40. Marianela Milanes (Asociación por los Derechos Civiles, ARG)
41. Carl Miller (Demos, UK)
42. Lisa-Maria Neudert (Oxford University, UK)
43. Aviv Ovadya (The Thoughtful Technology Project, US)
44. Symeon Papadopoulos (Centre for Research and Technology Hellas, GR)
45. Cameron Piercy (University of Kansas, US)
46. Thomas Poell (University of Amsterdam, NL)
47. Oreste Pollicino (Università Bocconi, IT)
48. Travis Ridout (Washington State University, US)
49. Luca Rossi (IT University of Copenhagen, DK)
50. Yotam Shmargad (University of Arizona, US)
51. Javier Ruiz Soler (European University Institute, IT)
52. Damian Tambini (London School of Economics, UK)
53. Emily Taylor (Chatham House, UK)
54. Rebekah K. Tromble (Leiden University, NL)
55. Claes De Vreese (University of Amsterdam, NL)
56. Abby Wood (University of Southern California Gould School of Law, US)
57. Brahim Zarouali (University of Amsterdam, NL)
58. Amy X. Zhang (MIT CSAIL, US)
59. Arkaitz Zubiaga (Queen Mary University of London, UK)

If you are a researcher working on these issues and want to co-sign this letter please contact us at [email protected], and we will add names on a rolling basis to this letter.

This work is part of Mozilla’s larger effort to combat online disinformation ahead of the upcoming EU elections, and other elections around the world in 2019. We’re closely following the commitments these companies made in the EU Code of Practice on Disinformation, in order to influence the development and assessment of the most effective tools possible.

Earlier this year, Mozilla and dozens of our allies sent a public letter to Facebook, demanding the company make good on their promises to provide more transparency around political advertising. Mozilla also conducted an EU-wide survey on the state of misinformation, and found that nearly 84 percent of people polled suspected (or knew for certain) that they had seen misinformation while using the internet that very week.