Statistically speaking it would be flawed to use this to compare and prioritize a list of urls in order to rank them by interestingness/attention/etc.
The reason is that pages that have Facebook like buttons embedded on them are statistically speaking going to have a higher chance of getting shared than those that don't (such as Hacker News - which doesn't have such buttons).
I guess it depends what you might use this for, but to me it seems to only have limited value.
We've been doing that on Maemo News (http://maemo.org/news/) for a few years now. Works well, especially if you couple it with local, HN-style voting and assign a separate weight for each relevancy source
The reason is that pages that have Facebook like buttons embedded on them are statistically speaking going to have a higher chance of getting shared than those that don't (such as Hacker News - which doesn't have such buttons).
I guess it depends what you might use this for, but to me it seems to only have limited value.