More specifically to your scenario - the bots are not the only ones who can downvote things

Yes but the main point of your post was that bots, as long as they continue to receive large delegations and represent a large portion of the voting stake, should be playing an important role in downvoting. I question whether this will be the case in the model you laid out, especially over time as people realize they can immunize themselves from those nasty bot downvotes (which you want to encourage) by buying tiny votes from each bot (or at least the ones active in downvoting)

I think your approach needs some iteration here. Overall I agree on not knowing how things will work out, and the iteration can certainly come later.