I think there is also a distinction in the granularity of the data harvested and what it was used for.
Knowing someone's location, age, etc and using that to classify whether or not someone should reach out to them to remind them to vote is one thing.
Harvesting everything they've liked on facebook to classify them by big five personality type to predict how to use disinformation to manipulate someone (often by scaring them into thinking they needed a gun to defend themselves, or that they needed to deport immigrants because they were gangerous, or into thinking islamic extremism was a real threat to their lives, as people high in neuroticism are more heavily motivated by arbitrary negative emotional impulses) into skewing their political preferences is an entirely separate game.
Knowing someone's location, age, etc and using that to classify whether or not someone should reach out to them to remind them to vote is one thing.
Harvesting everything they've liked on facebook to classify them by big five personality type to predict how to use disinformation to manipulate someone (often by scaring them into thinking they needed a gun to defend themselves, or that they needed to deport immigrants because they were gangerous, or into thinking islamic extremism was a real threat to their lives, as people high in neuroticism are more heavily motivated by arbitrary negative emotional impulses) into skewing their political preferences is an entirely separate game.