Are you being targeted with political ads based on personal data on Social Media? This is the subject of our paper “Automatically Identifying Political Ads on Facebook: Towards Understanding of Manipulation via User Targeting”. We are happy to share that the paper was accepted to the MISDOOM 2020 conference (International Symposium on Disinformation in Open Online Media) and will be published on Oct 26-27.

A pre-print version of the paper can be found on Arxiv.

To manifest a better insight into the properties of the data, we provide a
series of interactive illustrations by leveraging the targeting attributes, in addition to election information collected from online resources:

Top 15 Most Important Targeting Attributes. SHAP values show the distribution of the impacts each feature has on the model output.

The research has been in collaboration with ProPublica, a non-profit investigative newsroom, and researches from the George Washington University.  What is unique about the report is that it uncovers private data on  Facebook – specifically political advertising targeting information – that was collected by a network of volunteers. Following are a few highlights of the paper

  1. “Despite the growing public interest, the effect of political ad targeting on social media has not been explored yet due to the lack of publicly available data. In an effort to promote ad transparency and hold advertisers including political groups accountable, ProPublica has collected a dataset of political ads in the period before the 2018 US midterm elections.”

  2. “An important first step towards understating of political manipulation via user targeting is to identify politically related ads, yet manually checking ads is not feasible due to the scale of social media advertising. Consequently, we address the challenge of automatically classifying between political and nonpolitical ads, demonstrating a significant improvement compared to the current text-based classifier”

  3. “Our evaluation sheds light on questions, such as how user attributes are being used for political ads targeting and which users are more prone to be targeted with political ads”

As part of the AdVerif.ai Research Program we aim to develop state-of-the-art Natural Language Processing and advance scientific knowledge on fighting misinformation. It is the second research paper published to an ACM conference in the last year. Our first paper, “Identifying Nuances in Fake News vs. Satire: Using Semantic and Linguistic Cues” was presented in November 2019 as part of the prestigious EMNLP conference and was covered by VentureBeat and ZDNET.