Communities fall apart, especially large ones. Maintaining the same level of quality would be unprecedented, even if you had the exact same people they would change over time.
Also the minimalist computer science inspired upvote/downvote sorting system is a bit of a mismatch to human communication. Someone has to try a more complicated paradigm. I think tags on posts and comments would work better. Have a menu/autocomplete thing for common tags like [agree, troll, disagree, interesting, this, politics, hostile] and weigh submissions and comments through them. Should also be able to combine multiple tags and have a rating option for each tag. Label severity of disagreement, politics or suspected trolling.
This way when you need to tune the knobs you have more data to work with than just upvote/downvote, total karma and average karma. I don't expect pg to do these kinds of experiments though, he has minimalist tastes and limited time.
Also many times when communities fall apart, we have limited data to work with to see where things started to go wrong. Was it too many n00bs that chased experts away? Was it hostility? Did too many new people start voting on new submissions? I've seen some attempts at analysis of HN but don't remember them making clear conclusions.
Also the minimalist computer science inspired upvote/downvote sorting system is a bit of a mismatch to human communication. Someone has to try a more complicated paradigm. I think tags on posts and comments would work better. Have a menu/autocomplete thing for common tags like [agree, troll, disagree, interesting, this, politics, hostile] and weigh submissions and comments through them. Should also be able to combine multiple tags and have a rating option for each tag. Label severity of disagreement, politics or suspected trolling.
This way when you need to tune the knobs you have more data to work with than just upvote/downvote, total karma and average karma. I don't expect pg to do these kinds of experiments though, he has minimalist tastes and limited time.
Also many times when communities fall apart, we have limited data to work with to see where things started to go wrong. Was it too many n00bs that chased experts away? Was it hostility? Did too many new people start voting on new submissions? I've seen some attempts at analysis of HN but don't remember them making clear conclusions.