Don't know if this is really helpful...
Steemit is a hugely complex system and it is impossible to design right from the get go.
So it is important to experiment how incentives work to shape behaviour.
I think it will take a while to get to the right formula.
Telling people how to behave simply does not work!
People react to incentives, real or perceived.
This is where psychology is so important. You can't program it.
From other comments ion this post it seems that people are afraid to lose "voting power".
Fear of losing something (voting power, however little it is)
is a much stronger psychological driver than the hope to win something (curation benefits) So that is one potential explanation.

Solutions?
The self voting issue is easily fixed. Just change it so that you can't upvote yourself! I can hear the howls of protest now... not in the least from the whales...

I think tweaking the upvoting system is besides the point.
The whole upvoting is too easy to manipulate and automate. To really incentivize constructive content and engaging behaviour, the comments should be weighted a lot more.

If the purported goal of steemit is to let "good content" rise to the top, a comment should be a much better signal than an upvote if something is worthwhile.

I don't think upvotes should go away entirely but there should be some modifier effect because of comments.
Of course that can be gamed too. So to be considered "good comments" they should have a few upvotes and sub-comments too.

I have posts with great comments, 22 of them, (half are mine of course and those should not be counted) but 7or so people found it worth their while to contribute (some multiple times )a sensible opinion and left sizeable comments.
Post earned 6 SBD. Is that good or bad ?
Dunno.
But from the steemit platform perspective, that should be a better indicator that the post is "quality" rather than a cat video with 50 upvotes and no substantial comments.