I don't see how the technical point of view is relevant.

I listed three examples that should make it obvious. You refuse for some reason to accept the obvious, that these kind of tests try to alter real world behaviour and not behaviour with regards to the product.

There's a big difference between changing a button's color in a page to see how user engagement changes and trying to find when teenagers are in a more vulnerable emotional state to serve them ads. If you can't see the obvious I'm afraid there is anything more to say.