You could set up a comment voting system based on the theory of forum quality that only high-value comments in response to other high-value comments indicate a healthy forum. Make any upvote or downvote on a child comment also upvote or downvote the parent comment it is replying to. So you’d get a bonus of upvotes if someone replies to your upvoted comment with another upvoted comment, but people would be hesitant to reply anything to comments being downvoted since their replies probably wouldn’t be upvoted to keep the parent comment from getting upvotes.
This might be a thing that needs to be plugged in a forum from the start and let the local culture form around it instead of dropping it in an existing forum. Also there might be the problem that votes tend to degenerate to track consensus agreement instead of comment quality, and this system might exacerbate some groupthink failure modes.
We could test this theory by using it on the existing data and selecting the best comments under this theory. I would be interested in reading the “top 20 (or 50) LW comments ever” found by this algorithm, posted as a separate article. It could give us an approximate idea of what exactly the new system would incentiize.
Is there a good canonical source for all LW comments ever? I’m interested in importing the data into Python and playing around with ranking algorithms. (I’m not sure what disclaimer to use to keep others from not doing the same just because I publicly said that I’m interested in it, but yeah, feel free to duplicate work and come up with other interesting analyses)
It’s probably doable to use those to scrape comments and put them into some kind of list or database, but spending time looting LW comments that way seems like wasted effort compared to getting a full dump from an official source.
Ask a lot of good questions so that other people do the real work and say lots of stupid shit to people you don’t like. Would this be sufficient to game this system? :)
Someone should make a top level post introducing a contagious meme for upvoting good questions. Good questions are upvoted too rarely, but I don’t think we need software to fix it.
Absolutely. For my own part, I find that getting my good questions answered provides sufficient incentive, but I’m willing to believe that I’m atypical in that respect and further incentive above that up to a threshold would be beneficial,, and I have no idea where we are relative to that threshold.
You could set up a comment voting system based on the theory of forum quality that only high-value comments in response to other high-value comments indicate a healthy forum. Make any upvote or downvote on a child comment also upvote or downvote the parent comment it is replying to. So you’d get a bonus of upvotes if someone replies to your upvoted comment with another upvoted comment, but people would be hesitant to reply anything to comments being downvoted since their replies probably wouldn’t be upvoted to keep the parent comment from getting upvotes.
This might be a thing that needs to be plugged in a forum from the start and let the local culture form around it instead of dropping it in an existing forum. Also there might be the problem that votes tend to degenerate to track consensus agreement instead of comment quality, and this system might exacerbate some groupthink failure modes.
We could test this theory by using it on the existing data and selecting the best comments under this theory. I would be interested in reading the “top 20 (or 50) LW comments ever” found by this algorithm, posted as a separate article. It could give us an approximate idea of what exactly the new system would incentiize.
Is there a good canonical source for all LW comments ever? I’m interested in importing the data into Python and playing around with ranking algorithms. (I’m not sure what disclaimer to use to keep others from not doing the same just because I publicly said that I’m interested in it, but yeah, feel free to duplicate work and come up with other interesting analyses)
You could ask matt to send you the necessary parts of the database.
There’s this and this. Maybe they allow you to go all the way back to the beginning.
It’s probably doable to use those to scrape comments and put them into some kind of list or database, but spending time looting LW comments that way seems like wasted effort compared to getting a full dump from an official source.
Ask a lot of good questions so that other people do the real work and say lots of stupid shit to people you don’t like. Would this be sufficient to game this system? :)
Incentivizing asking a lot of good questions but not so many that people who might answer such questions get overloaded might be a good thing.
Someone should make a top level post introducing a contagious meme for upvoting good questions. Good questions are upvoted too rarely, but I don’t think we need software to fix it.
Absolutely. For my own part, I find that getting my good questions answered provides sufficient incentive, but I’m willing to believe that I’m atypical in that respect and further incentive above that up to a threshold would be beneficial,, and I have no idea where we are relative to that threshold.