Great question. I guess I'd start by questioning the premise that Twitter can become a place that produces optimal interactions. That character limit, I suspect, will always lead to less than sincere, and certainly less than complete conversations.

But if I had to implement a change, I'd ban profanity and introduce some kind of downvote system. Profanity, because psychologically, I think it makes a difference if you're not able to rely on swearing to make your point. Even if you're just hurling insults, there's a meaningful difference between, "you're an idiot," and "you're a fu**ing idiot." (This would also make it a little more ad-friemdly.)

And downvoting, because it would provide a way for people to say, "this isn't productive/valuable." Currently, no such system exists on Twitter. What I'd really like is for this to eventually become a way to identify trolls, though, of course, you'd have to find ways to avoid its abuse. Youtube used to have downvotes and it was really helpful for identifying crappy "how-to" videos and things like that. Plenty of kinks to work out there, but those are my knee-jerk ideas.