Discrimination through Optimization: How Facebook’s Ad Delivery Can Lead to Biased Outcomes

The enormous financial success of online advertising platforms is partially due to the precise targeting features they offer. Although researchers and journalists have foundmanyways that advertisers can target—or exclude— particular groups of users seeing their ads, comparatively little attention has been paid to the implications of the platform’s ad delivery process, comprised of the platform’s choices about which users see which ads. It has been hypothesized that this process can “skew” ad delivery in ways that the advertisers do not intend, making some users less likely than others to see particular ads based on their demographic characteristics. In this paper, we demonstrate that such skewed delivery occurs on Facebook, due to market and financial optimization effects as well as the platform’s own predictions about the “relevance” of ads to different groups of users. We find that both the advertiser’s budget and the content of the ad each significantly contribute to the skew of Facebook’s ad delivery. Critically, we observe significant skew in delivery along gender and racial lines for “real” ads for employment and housing opportunities despite neutral targeting parameters. Our results demonstrate previously unknown mechanisms that can lead to potentially discriminatory ad delivery, even when advertisers set their targeting parameters to be highly inclusive. This underscores the need for policymakers and platforms to carefully consider the role of the ad delivery optimization run by ad platforms themselves—and not just the targeting choices of advertisers—in preventing discrimination in digital advertising.1

[1]  A. Agresti,et al.  Approximate is Better than “Exact” for Interval Estimation of Binomial Proportions , 1998 .

[2]  G. Cumming,et al.  Inference by eye: confidence intervals and how to read pictures of data. , 2005, The American psychologist.

[3]  Craig E. Wills,et al.  Understanding what they do with what they know , 2012, WPES '12.

[4]  Toniann Pitassi,et al.  Fairness through awareness , 2011, ITCS '12.

[5]  Latanya Sweeney,et al.  Discrimination in online ad delivery , 2013, CACM.

[6]  Roxana Geambasu,et al.  XRay: Enhancing the Web's Transparency with Differential Correlation , 2014, USENIX Security Symposium.

[7]  Balachander Krishnamurthy,et al.  Measurement and analysis of OSN ad auctions , 2014, COSN '14.

[8]  Michael Carl Tschantz,et al.  Automated Experiments on Ad Privacy Settings: A Tale of Opacity, Choice, and Discrimination , 2014, ArXiv.

[9]  David Lazer,et al.  Measuring Price Discrimination and Steering on E-commerce Web Sites , 2014, Internet Measurement Conference.

[10]  Karrie Karahalios,et al.  Auditing Algorithms : Research Methods for Detecting Discrimination on Internet Platforms , 2014 .

[11]  Balachander Krishnamurthy,et al.  Beyond CPM and CPC: determining the value of users on OSNs , 2014, COSN '14.

[12]  Sean A. Munson,et al.  Unequal Representation and Gender Stereotypes in Image Search Results for Occupations , 2015, CHI.

[13]  Roxana Geambasu,et al.  Sunlight: Fine-grained Targeting Detection at Scale with Statistical Confidence , 2015, CCS.

[14]  Adam Tauman Kalai,et al.  Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings , 2016, NIPS.

[15]  Claude Castelluccia,et al.  MyAdChoices: Bringing Transparency and Control to Online Advertising , 2016, ACM Trans. Web.

[16]  David García,et al.  Bias in Online Freelance Marketplaces: Evidence from TaskRabbit and Fiverr , 2017, CSCW.

[17]  Krishna P. Gummadi,et al.  Quantifying Search Bias: Investigating Sources of Bias for Political Searches in Social Media , 2017, CSCW.

[18]  D. Tambini,et al.  Digital Dominance: The Power of Google, Amazon, Facebook and Apple , 2018 .

[19]  Timnit Gebru,et al.  Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification , 2018, FAT.

[20]  Krishna P. Gummadi,et al.  Investigating Ad Transparency Mechanisms in Social Media: A Case Study of Facebooks Explanations , 2018, NDSS.

[21]  Christo Wilson,et al.  Investigating the Impact of Gender on Rank in Resume Search Engines , 2018, CHI.

[22]  Krishna P. Gummadi,et al.  Potential for Discrimination in Online Targeted Advertising , 2018, FAT.

[23]  Krishna P. Gummadi,et al.  Treads: Transparency-Enhancing Ads , 2018, HotNets.

[24]  David Lazer,et al.  Auditing Partisan Audience Bias within Google Search , 2018, Proc. ACM Hum. Comput. Interact..

[25]  Aleksandra Korolova,et al.  Facebook's Advertising Platform: New Attack Vectors and the Need for Interventions , 2018, ArXiv.

[26]  Krishna P. Gummadi,et al.  Privacy Risks with Facebook's PII-Based Targeting: Auditing a Data Broker's Advertising Interface , 2018, 2018 IEEE Symposium on Security and Privacy (SP).

[27]  Deirdre K. Mulligan,et al.  Discrimination in Online Personalization: A Multidisciplinary Inquiry , 2018, FAT.

[28]  Fabrício Benevenuto,et al.  Measuring the Facebook Advertising Ecosystem , 2019, NDSS.

[29]  Cynthia Dwork,et al.  Fairness Under Composition , 2018, ITCS.

[30]  Krishna P. Gummadi,et al.  Auditing Offline Data Brokers via Facebook's Advertising Platform , 2019, WWW.

[31]  Catherine Tucker,et al.  Algorithmic bias? An empirical study into apparent gender-based discrimination in the display of STEM career ads , 2019 .

[32]  Ben Green,et al.  Disparate Interactions: An Algorithm-in-the-Loop Analysis of Fairness in Risk Assessments , 2019, FAT.

[33]  Piotr Sapiezynski,et al.  Investigating sources of PII used in Facebook’s targeted advertising , 2019, Proc. Priv. Enhancing Technol..

[34]  Piotr Sapiezynski,et al.  Quantifying the Impact of User Attentionon Fair Group Representation in Ranked Lists , 2019, WWW.