In this presentaiton, I will present several works we did on making political advertising safer: 

(i) Behavior analysis: We have analyzed the political ad campaign sent by the Russian Internet Research Agency (IRS) before the U.S. Presidential elections to understand what mechanisms have been exploited to manipulate voters.
(ii) Auditing: To make advertisers more accountable, Facebook asks advertisers to self-declare if they are sending political ads and provides a central repository (the Facebook Ad Library) that contains information about the political ads that have run on their platform. Ad Libraries are a positive step forward as they allow researchers and journalists to scrutinize political ads, but come with limitations. One big problem is that advertisers need to self-declare whether their ad is political, and we do not know what enforcement mechanisms are put in place by Facebook to check compliance. Hence, we have no idea how many (undetected) political ads are incorrectly missing from the Ad Library? We have proposed an auditing system to assess how many political ads are missing from the Ad Library.
(iii) Content moderation: In a follow-up study, we wanted to understand, at a more fundamental level, if there is hope to reliably distinguish political from non-political ads (e.g., given an ad, can users agree whether it is political or not). This is important because political ads are subject to higher levels of transparency and restrictions both from ad platforms and country specific legislations; and unreliability can lead to discrimination and circumvention. 

From a legal perspective, I will present how we articulated these findings and contributed to European law through the Digital Services Act and the European Democracy Action Plan.