Lakeview Times Case Study: Draft Expert Opinion 1

[This  is an example of a draft expert opinion, written collectively  by the T-Tripods Ethics Case study working group and is a rough draft. People submitting their cases for consideration to the collection being presented in conjunction with the 1st Annual Workshop on Methods for Teaching Ethics in Data Science, have the option of attaching 1) no expert opinions 2) 1-3 draft expert opinions such as this example, or 3) 1-3 finished and polished expert opinions from someone different than the case author with a byline.  The program committee will be judging which cases we will select for inclusion. For cases that are selected that come with < 3 expert opinions of  type 3, the Tufts Undergraduate think tank The Lantern and their faculty advisors will be soliciting appropriate experts to write the expert opinions on the cases. ]

The Lakeview Times has a clear policy of whom it shows ads to: the person for whom they will be  paid the most money for an impression. This is not a discriminatory policy. Lakeview, on the other hand, does allow its advertisers to further target which subset of readers see its ads; we know this from the fact that Rainbow Togs is allowed to target its ads to readers that The Lakeview Times believes are of the female gender, based on whatever demographic information Lakeview has access to. The error here comes from Brookstar: Brookstar did NOT specify what subset of readers it wanted to see its ads; Brookstar implicitly assumed that not specifying a target group meant that the ad would be shown equally to male and female readers, or at least in proportion to which men and women read the Lakeview Times. But the Lakeview Times instead interpreted this “no preference” as “any reader, with any demographics, in any proportion” rather than “want to reach a typical proportion of readers by gender.” Better communication between Brookstar and Lakeview would have uncovered this discrepancy. 

It is worth noting, however, that if Lakeview’s ad policy is simply to always give the ad impression to the company that is implicitly bidding for a particular customer, then even changing Brookstar’s ad buy to insist that the ad be shown half to male and half to female readers may not, absent other changes in Lakeview’s algorithm fix the problem. For example, let’s compare the scenario where Brookstar buys 500 impressions (reader characteristics not specified) to the updated scenario where Brookstar correctly buys 250 impressions to male readers and 250 impressions to female readers. Let us imagine a  scenario where Rainbow Togs has bought sufficient premium ads to show to every  female customer who visits the website for the next 90 days. Then in the first scenario, as in the story, all 500 impressions go in front of male readers. But in the second scenario, 250 impressions go in front of male readers and the remaining 250 impressions are still waiting to be shown when the Rainbow Togs buy gets exhausted! If the scholarship application deadline is sooner than 90 days from now, Brookstar is still going to be unhappy: no female readers will get to see the ad for the Brookstar Scholarship until past the deadline to apply! Thus the simple policy of always giving the ad impression to the highest bidder ignores another “hidden” customer requirement as well, namely that ads that are timely get served in a timely manner: ads about a Fourth of July Sale should be served before the fourth of July Sale is over, ads for Brookstar Scholarship Applicants should be served before the application deadline. Again, if Brookstar writes down its true requirements (ads served in proportion to reader demographic groups in regards to gender, race, age, etc AND all ads bought served before a particular date and time), then it would be good business for Lakeview to try to meet these requirements. An ad assignment algorithm that is not demographic and deadline aware cannot meet these requirements, so Lakeview needs to implement something more complicated. In this particular case, before changing their entire algorithm, we would recommend that Lakeview simply make a charitable contribution to a worthy local cause (Brookstar Scholarship Fund), by treating their ads as premium 8 cent ads so they will be shown half the time instead of the Rainbow Togs ads, even if they are only netting The Lakeview Times three cents an impression.