RE: As a top steem witness, will you support a hardfork implementing the “Economic Improvement Proposal” or “EIP.”
I don't entirely agree with this proposal, but overall it shows promise.
1 (curation rewards) - This is the part I'm mostly concerned about. As @drakos mentioned, the psychological impact on authors having 50% of their post rewards go to other people could negatively impact authors willing to post.
This also aggravates bid bots, because now bid bots will be getting massive curation returns from their votes, and thus getting even more powerful. This in effect counteracts points 2 and 3, since by strengthening the returns for bid bots, it may reduce the cost of them - potentially cancelling out the other changes that would by themselves reduce the viability of bid bots.
As suggested by @anyx - having this be a witness parameter would help avoid future hard forks, witnesses would be able to adjust it over time in response to any positive or negative effects on author activity and curation.
2 (convergent linear rewards) - Overall I agree with this, it's difficult to see how this would affect authors in the long term, since it may deter authors with a smaller following.
At the same time it could help limit the current power that bid bots have over post rewards, keeping more of the reward pool available to authors who don't use bid bots.
Something that we may only find out after it's implemented, is how well this stands up to the fact that there are many bid bots available with high voting power, and when people use several of them on a single post then the curve may simply return to linear.
3 (downvote pool) - This is a definite YES. There is currently no benefit to downvoting a post. Skilled curators who often find the best and worst content, would generally prefer to conserve their voting mana for the good posts rather than waste it on downvoting poor content.
A downvote pool would help to encourage people to downvote illegal, low quality or plagairized content without worrying about it affecting their upvote power. This has been a commonly requested idea for a long time now - it's worth at least trying it out.