These are very valid points. I would personally not be able to tell if your scientific paper is good or garbage because I do not have the knowledge required to make that distinction.
I think it could be interesting to use the mechanism of Steem but on a more "close" system. If, there were high requirements in order to join coupled with compartmentalization, peers could "curate" a topic, respond to it point out issues with the research, etc. Maybe upvote would play as big a part than valid comments who are upvoted. There are a lot of different way the algorithm could work.
The trap is to fall into sensationalizing science when sometime the most important stuff is the kind of research the majority of people find unreadable and boring.
I definitely agree! One needs a pool of experts of a given field to get a chance the control the content. But ho to attract them here (for work)?