I'd rather not outsource my morality to the arbitration of an algorithm, regardless of its provenance being of a company that claims to "not be evil." This honestly seems like a particularly flagrant application of this feature; we have enough human interaction mediated by coercive tech, the way we communicate personal beliefs about one another shouldn't be the next pillar to fall. That it's "just a suggestion", as others have argued to excuse it, belies how strongly its UI implies authoritativeness -- users reflexively view the squiggly underline as a sign that something is unambiguously wrong with what they've written.