We programmers look at difficulties users have dealing with large volumes of data and think, aha! I can solve this by giving them ways to add metadata to those data, so it can be more easily browsed/filtered/faceted/whatever. (We think that because programmers have an analytical cast of mind, which we then extrapolate from -- this would be useful for me, we figure, therefore it would be useful for anyone.) So we build these elaborate systems for storing and managing metadata, and then build elaborate UIs to expose those systems to the user.
Our system goes out to the users, and as long as those users are people like us -- analytical types -- it works OK. But then it goes out beyond that to a general audience, and suddenly we discover a new problem: namely that most users don't want to add metadata and simply will not do it, even if by doing so they could realize huge productivity gains. It doesn't matter how useful the metadata would be to them, to them adding metadata is work, and worse, it's nerd work of a kind they feel fundamentally uncomfortable with and unequipped to deal with. So they just don't do it. And they won't do it, no matter how hard you try to sell them on it.
That was the problem with "circles" in G+: not that it was a bad idea to let people split up their feeds, but that having multiple feeds required them to add metadata (creating circles, naming them, adding people to them, periodically removing people from them, etc.), and people just didn't want to do that. So they didn't.
It's not impossible to build systems that categorize data to make it more comprehensible, but if you want them to work, they have to do so in some way that doesn't put the burden on the user to keep everything correctly tagged/categorized/faceted/whatever. Because they just will not do that, no matter how many carrots you dangle.