I can't believe nobody has mentioned collaborative filtering.
Simple collective voting and new/hot/controversial are very unsophisticated measures of how much any particular user will like something.
For an example of how powerful collaborative filtering is, check out movielens.org
The key is to take each user's votes and use them to figure out based on other peoples' votes what stories and comments to display. Then the community doesn't really "grow". Sure you may get some niche communities that grow, such as people who all tend to upvote cat pictures, but the "thinking person" doesn't have to worry, b/c as long as there is a weak (or negative) correlation between cat pictures and your upvotes, the system will never recommend any to you.
Imagine rather than one massive community, an infinite number of beautifully overlapping ones. I think the ideal system would consider votes for stories and comments, and show each person mostly content that had a high probability of being enjoyed.
The "new" tab might show things with less of a strict filter to prevent false negatives.
The research group that built Movielens has some papers on the web and the math isn't that tough. With a properly design collaborative filtering system there simply will not be the digg/reddit degradation where the site gets so big and noisy and low quality. Sure there may be stories that are hugely popular, but as long as the algorithm gets enough quality input (votes) on a variety of things, no user would get too many duds.
Ironically Reddit's redesign only increases noise by tending people away from the subreddit niche communities that had existed -- I would more likely click on a weak title if in a quality subreddit than if it's the #4 item on the screen.
At the rate it's going, news.yc will surely become the next reddit, and eventually the next Digg. As others have pointed out, we'll all have probably moved on by then to the next site (or hopefully one that implements CF!).
It seems odd to me that nobody has implemented a site like this with CF... As with most of my posts, if anyone wants to collaborate to do an experimental site like this just let me know. There are some cool ruby libraries out there and of course gsl.
Comments
I can't believe nobody has mentioned collaborative filtering.
Simple collective voting and new/hot/controversial are very unsophisticated measures of how much any particular user will like something.
For an example of how powerful collaborative filtering is, check out movielens.org
The key is to take each user's votes and use them to figure out based on other peoples' votes what stories and comments to display. Then the community doesn't really "grow". Sure you may get some niche communities that grow, such as people who all tend to upvote cat pictures, but the "thinking person" doesn't have to worry, b/c as long as there is a weak (or negative) correlation between cat pictures and your upvotes, the system will never recommend any to you.
Imagine rather than one massive community, an infinite number of beautifully overlapping ones. I think the ideal system would consider votes for stories and comments, and show each person mostly content that had a high probability of being enjoyed.
The "new" tab might show things with less of a strict filter to prevent false negatives.
The research group that built Movielens has some papers on the web and the math isn't that tough. With a properly design collaborative filtering system there simply will not be the digg/reddit degradation where the site gets so big and noisy and low quality. Sure there may be stories that are hugely popular, but as long as the algorithm gets enough quality input (votes) on a variety of things, no user would get too many duds.
Ironically Reddit's redesign only increases noise by tending people away from the subreddit niche communities that had existed -- I would more likely click on a weak title if in a quality subreddit than if it's the #4 item on the screen.
At the rate it's going, news.yc will surely become the next reddit, and eventually the next Digg. As others have pointed out, we'll all have probably moved on by then to the next site (or hopefully one that implements CF!).
It seems odd to me that nobody has implemented a site like this with CF... As with most of my posts, if anyone wants to collaborate to do an experimental site like this just let me know. There are some cool ruby libraries out there and of course gsl.