The problem I have with this sort of scheme is the rubber-necking aspect. When a car hits the wall you get a lot of people slowing down to look but its not "interesting" information its more morbid curiosity.<p>Something I'm looking to play with a bit more is trying to measure discovery 'reaction.' The first test will be trying to score 'reaction' when new blog posts are discovered. By crawling the blogosphere and looking at signals along the lines of activity generated when discovered (as opposed to over all activity) I'm wondering if we can characterize "interesting" information from "oh and here is another example" type information.
The Drip Stream is kinda cool, since it effectively prevents "everything" from becoming popular, but also makes it easy to bootstrap if there aren't a lot of users playing. I am very interested in how these approaches affect user return rates, since it would be very easy to try a bunch of algorithms on different users and see which gets more people to come back.
This looks awesome, I've been curious to how we might be able to programatically solve for "surfacing interesting content". Upvote solutions are on a decline, and curation/moderation is on the rise as we realize those solutions don't scale all that well. Apply to music, videos, photos, links, comments, etc...
Great article. Unlike most blog posts on this topic, it offers practical approaches to computing the score in real time without resorting to cron jobs and the like.
I found the article to be well-written and informative, but I take issue with the title:<p>Surface isn't a verb unless you're a submarine.<p>Or rather, outside of some narrow technical uses, it sounds jargony and self-important.