My main thought is that without a ton of data points in the topic (like user interaction), it will be hard to accurately create a summary view. It’s one of those things that would work great on a site like ArsTechnica or even Slashdot or the Gawker sites, just because their commenters are crazy about participation.
On a site like How-To Geek where the discussion rarely goes over 20 posts per topic, it’s probably not the most useful thing. I think it’s one of those “cool” features that will probably work well about a year or two from now.
What is the algorithm behind this? That should be up for discussion.
Personally on a giant topic I’d like to see posts from the following:
- Consistently interesting people. Like that @codinghorror guy, he always writes good stuff.
- Posts from new(er) users that are Liked a lot. Liked by the more interesting people, not by drive-by users.
- Posts that are critical to the discussion, like if there is a chain of a couple of replies.
You almost need two algorithms, one for smaller sites, and one for larger sites, that factor in the participants a bit more.
EDIT: Testing to see if adding a link and text will push this to the top: How we’re addressing misleading apps in Windows Store - Windows Developer Blog