I've struggled with these since start. I think they are a good idea, but it is hard to set them up effectively (via website or clients) and also use them effectively from anywhere (it's just too convenient to default to the main timeline).
I thought about defining lists by interest: Art, Code, Cycling, but then decided against it as at any point in time I'm going to want to check all three anyway; neither is of more importance. Plus, categorising peoples tweets is a bit pointless as tweets often span a huge variation of topics (as life does). I myself would tweet in all three of these categories as well as others.
So I tried thinking of lists as more like filters. As I think this is really what I'm looking for when using a client, etc, a way to filter out the cruft and just show people i'm most interested in when I'm in a rush. So the first level of 'noise' reduction would be anything that's not art/code/cycling (but how do I categorise as them!?). The second level of noise reduction would be discounting accounts that aren't people (organisations, etc). The third level of noise reduction would be removing people I don't know - celebs, etc. And so on, and so on. Assuming this is the correct way to define levels of noise reduction, which it might not be.
But as I started setting this up I thought "this is a royal pain in the arse" (and maintaining them would be too). So I've pretty much given up.
Also, I'm not too sure as to the point of public lists. As I might create a list to contain all the people I am following because of flickr, but this list is useless to everyone else... isn't it? Perhaps of very slight relevance to those on the list who might be able to draw some insight as to why I am following them based by their appearance on a particular list.
But, at long last, I've finally got the point of them.