simulating a market with honest and deceptive advertisers
11 June 2018
At Nudgestock 2018 I mentioned the signaling literature that provides background for understanding the targeted advertising problem. Besides being behind paywalls, a lot of this material is written in math that takes a while to figure out. For example, it's worth working through this Gardete and Bart paper to understand a situation in which the audience is making the right move to ignore a targeted message, but it can take a while.
Are people rational to ignore or block targeted advertising in some media, because those media are set up to give an incentive to deceptive sellers? Here's a simulation of an ad market in which that might be the case. Of course, this does not show that in all advertising markets, better targeting leads to an advantage for deceptive sellers. But it is a demonstration that it is possible to design a set of rules for an advertising market that gives an advantage to deceptive sellers.
What are we looking at? Think of it as a culture medium where we can grow and evolve a population of single-celled advertisers.
The x and y coordinates are some arbitrary characteristic of offers made to customers. Customers, invisible, are scattered randomly all over the map. If a customer gets an offer for a product that is close enough to their preferences, it will buy.
Advertisers (yellow to orange squares) get to place ads that reach customers within a certain radius. The advertiser has a price that it will bid for an ad impression, and a maximum distance at which it will bid for an impression. These are assigned randomly when we populate the initial set of advertisers.
High-bidding advertisers are more orange, and lower-bidding advertisers are more pale yellow.
An advertiser is either deceptive, in which case it makes a slightly higher profit per sale, or honest. When an honest advertiser makes a sale, we draw a green line from the advertiser to the customer. When a deceptive advertiser makes a sale, we draw a red line. The lines appear to fade out because we draw a black line every time there is an ad impression that does not result in a sale.
So why don't the honest advertisers die out? One more factor: the norms enforcers. You can think of these as product reviewers or regulators. If a deceptive advertiser wins an ad impression to a norms enforcer, then the deceptive advertiser pays a cost, greater than the profit from a sale. Think of it as having to register a new domain and get a new logo. Honest advertisers can make normal sales to the norms enforcers, which are shown as blue squares. An ad impression that results in an "enforcement penalty" is shown as a blue line.
So, out of those relative simple rules—two kinds of advertisers and two kinds of customers—we can see several main strategies arise. Your run of the simulation is unique, and you can also visit the big version.
What I'm seeing on mine is some clusters of finely targeted deceptive advertisers, in areas with relatively few norms enforcers, and some low-bidding honest advertisers with a relatively broad targeting radius. Again, I don't think that this necessarily corresponds to any real-world advertising market, but it is interesting to figure out when and how an advertising market can give an advantage to deceptive sellers, and what kinds of protections on the customer side can change the game.
Ben Miroglio, David Zeber, Jofish Kaye, and Rebecca Weiss. 2018. The Effect of Ad Blocking on User Engagement with the Web. In WWW 2018: The 2018 Web Conference, April 23–27, 2018, Lyon, France. ACM, New York, NY, USA, 9 pages. https://doi.org/10.1145/3178876.3186162