I've been digging through this in an effort to understand exactly how it preserves user privacy. I'm at a bit of a loss, probably because I lack domain expertis

I've been digging through this in an effort to understand exactly how it preserves user privacy. I'm at a bit of a loss, probably because I lack domain expertise. From digging through their docs[1]: > This adds a new privacy concern, however. Revealing that a person is in a treatment or control group is just a random value, but revealing the fact that they had an opportunity could reveal much more information, such as demographics used for targeting of the advertising campaign. (Note: revealing a person’s treatment/control group assignment isn’t entirely private, as it could reveal membership in a group, which we often want to avoid in general. Opportunity logging, though, can reveal much more information.) I understand that if I were to see that Johnny Individual was served ad campaign X, and X was targeted towards 35 year old male identifying individuals in Colorado, I would know that Johnny Individual fit the bill according to the ad network. What I can't understand is: Who are we trying to keep this info from? The ad network needs to know this to properly run the campaign. In the example they provide, I believe the ad network is running the lift test and reporting aggregate results. Could someone please point out what I'm missing? [1] https://github.com/facebookresearch/fbpcf/blob/master/docs/PrivateRCT.md https://github.com/facebookresearch/fbpcf/blob/master/docs/P...