Create a witness approved list of bots that are blacklisted from contributing to user-rep in any way.
This is a consensus rule, so it must be enforced by a hardfork. Which means we would have to do a hardfork every time we add somebody on that list. Not feasible.
Couldn't you just have witnesses (and/or the account owner) cryptographically consent to additions to the list after the use of the list is added in a single HF?
Consensus doesn't work like this. It's a more complex question that atomically accept / reject changes. Also, involving witnesses in this thing could lead to competition - some of them may consider some bot legits, other not. It's opening Pandora's box.
No, you would definetely need consensus on the idea that bot votes shouldn't count towards reputation and that account holders should at least be able to do some kind of irreversable discretionary drop humanness operation, and that self votes and related (proxy/recovery link) accounts shouldn't count to rep either. The pandora's box part I think, would be limited to agreement that an account that didn't discretionarily drop its humanness should be on the list nevertheless. I could imagine that if someone were to make a hybrid bid-bot that in fact includes some form of owner curation (Ok, if the #1 bid is crap and the #2 bid is not, I'll upvote #2 instead), there might be some strive.