January 23, 2015

Discrimination in Online Ad Delivery

Have you been arrested?  Imagine this question appears when someone searches for your name on the Internet.  The Discrimination in Online Ad Delivery report by Professor Latanya Sweeney from Harvard University indicated that an Internet search by black-identifying names was more likely to yield an advertisement suggestive of an arrest record than a search by white-identifying names (download report here).  This racially biased advertisement distribution could inadvertently impact a person’s reputation and the affected person might not know about the potential negative consequences. 

However, advertisers are protected under the First Amendment of the Constitution that advertisements are commercial free speech.  But what if the advertisements suggestive of a criminal background appear more frequently for one racial group than for another’s, is that still free speech or has it become racial discrimination?  As this research yielded more questions than answers, perhaps Internet search companies should yield to moral high grounds and refine their search logic to provide a user experience free of racial bias.  

This viz was influenced by the New York Times’ Why Is Her Paycheck Smaller graph, which has been well-received in the data visualization community for its elegant use of annotation to deliver the thought-provoking message.  In order to draw the diagonal lines in the scatter plot, I sought help from Tableau Zen Master extraordinaire, Joe Mako.  Then I inserted the ‘% Arrest Ads’ annotations to the viz.  The final result highlights the unfair issue of discrimination in online ad delivery.

In addition, the box plots to the right show the differential on the arrest ads display when searching for names according to race and gender.  

This viz was made possible with the permission of Professor Latanya Sweeney from Harvard University and the contribution of Joe Mako, Tableau Zen Master.  I'm grateful to both for their generosity.

No comments:

Post a Comment