This is a complicated game, because they are three players: you, the customer paying for the advertisement, and the ad agency—in this case the social network. Each of them is an agent with separate goals: you want some utility from reading the web, the customer wants to increase sales, the ad agency wants to extract money from the customer. Things that don’t make sense from your perspective may point to a conflict of interest between the remaining two players.
For example, if the customer is willing to pay for the number of ads you see (because they don’t know better, or because this is the only thing they can verify somehow), it makes sense for the ad agency to show you more ads even if it meant that you will be less impressed with them, because they are paid for the former and don’t really care about the latter.
Honestly, I have no idea how the customers evaluate whether it is true when Facebook tells them “we have shown your ad to 100 000 people”. Do they trust the numbers blindly? Do they have some way to check… for example by creating a few fake profiles, checking whether the ad was displayed to any of them, and doing some statistical reasoning? No idea. Perhaps understanding what they measure could shed some light on what is the maneuvering space for Facebook for goodharting their measures.
The idea is that the ad agency does not optimize for the impact of the ads on you. They optimize for the customers’ belief in the impact of the ads on you. That is sometimes easier achieved by cheating the customer.
This is a complicated game, because they are three players: you, the customer paying for the advertisement, and the ad agency—in this case the social network. Each of them is an agent with separate goals: you want some utility from reading the web, the customer wants to increase sales, the ad agency wants to extract money from the customer. Things that don’t make sense from your perspective may point to a conflict of interest between the remaining two players.
For example, if the customer is willing to pay for the number of ads you see (because they don’t know better, or because this is the only thing they can verify somehow), it makes sense for the ad agency to show you more ads even if it meant that you will be less impressed with them, because they are paid for the former and don’t really care about the latter.
Honestly, I have no idea how the customers evaluate whether it is true when Facebook tells them “we have shown your ad to 100 000 people”. Do they trust the numbers blindly? Do they have some way to check… for example by creating a few fake profiles, checking whether the ad was displayed to any of them, and doing some statistical reasoning? No idea. Perhaps understanding what they measure could shed some light on what is the maneuvering space for Facebook for goodharting their measures.
The idea is that the ad agency does not optimize for the impact of the ads on you. They optimize for the customers’ belief in the impact of the ads on you. That is sometimes easier achieved by cheating the customer.