I personally know a FAANG employee whose full time job is building tools to try and help understand why the company's recommendation algorithm picks the things it does (and more specifically, predict how changes to the algo will affect that).

Even the people who build these systems barely know what the algorithm is going to do, much less why. It will be a herculean task to try and convey that to an average user.