I don't see the point, because of this: "Already possible with off-chain logic. If you write a good bot to oppose every action of a particular user, you can achieve the same effect, so why add an on-chain solution?" If the bot-owner is worried about offending the post's author, they could program the bot to leave a comment to explain that it's nothing personal.

Also, ideally, I think the vote should have some relationship to the content that is being voted upon. Yes, stakeholders are free to choose how to vote their stake, but the blockchain shouldn't encourage them to vote based on any factors other than to appraise the value of a post. Intuitively with vote negation, negated votes would probably be cast in the wrong direction 50% of the time (as compared to the "true" value of the post). Conceptually, it's a step away from "proof of brain", not towards it.

If anything, I think the problem now is the opposite. Whether downvote or upvote, there's no penalty for casting a wildly extreme vote. Which is why I have been a long-time proponent of an algorithm based upon a second price auction. Under rules like that, the voter would be penalized for being out of step with the voting consensus - on upvotes and downvotes. This would motivate voters to consider not just their own appraisals, but also their own perception of how others will appraise the post..