These days, were I writing an RSS reader (I’m not), I think I’d skip developing an algorithm based on the user’s attention — instead, I’d focus on making it really easy to filter out the things you don’t care about, and to highlight the things you’re more likely to want to see.
I think the problem with Facebook's (and soon Twitter's) approach, isn't the algorithm based timeline itself, but the fact that it is effectively the only option.
I've been thinking a lot about that lately. Were I writing an RSS Reader (I might be), I think I'd give it a go. That wouldn't be the only way of reading in the app though. By all means, the user would have the option to see the unfiltered queue. That option would be at least as prominent as the automatically filtered queue and the app would never try to force the user into using the filtered version. Under these constraints, I believe a filtered queue that takes into account past user behavior may be useful in suggesting what to read next.