Title: Discrimination through optimization: How Facebook's ad delivery can lead to skewed outcomes
Speaker: Piotr Sapieżyński, Postdoctoral Researcher at Northeastern University.
Abstract: Online social networks - such as Facebook - have become the gateway to the Internet for billions of people. In turn they have accumulated rich data of user behavior which forms the basis of their powerful advertising platforms. Some of the information that Facebook has for users comes directly from the users themselves, but also data collected from strangers, as well as from third party offline data brokers. What are the consequences of online giants collecting huge amounts of personal information without allowing users to delete or correct it?
In this presentation Piotr will talk about gender and race bias in ad delivery. He shows that users of different races and genders see different ads, even in the protected categories, like jobs and housing. This bias occurs even without the advertisers' intent, may lead to reinforcing societal inequalities and is a consequence of the pursuit for maximizing the click-through rates. You can read the paper here: https://arxiv.org/abs/1904.02095
The work has received quite a bit of attention from the press, see also:
https://www.wired.com/story/facebooks-ad-system-discrimination/
https://www.theatlantic.com/ideas/archive/2019/04/facebook-targeted-marketing-perpetuates-discrimination/587059/
https://www.theverge.com/2019/4/4/18295190/facebook-ad-delivery-housing-job-race-gender-bias-study-northeastern-upturn
https://www.cnbc.com/2019/04/04/facebook-targets-ads-based-on-race-and-gender-stereotypes-study.html