I think 10/10 is good, but for votes on articles, there should be an "originality" multiplier, which people can vote up and down, which will multiply the whole thing, so that the score is originality times 10 times (upvotes minus downvotes). The originality multiplier is important for research papers, as without this, accurate but derivative papers will outscore classics which are extremely original, but harder to read, perhaps, or less immediately obviously useful, because they are ahead of their time.
Not rewarding originality is a very big problem in a refereeing community, it rewards derivative work, and encourages second-rate contributions that merely rehash things that are already known. This is understood by most journal editors, who have a high bar for originality. This doesn't need to apply to Q/A of course (but maybe it's good there too).