zlacker

>>pg+(OP)
One thing I wonder about is whether accumulating personal karma is a red herring: trolls don't seem to care about their karma, and good folks may not care either.

Perhaps the most important thing about upvotes and downvotes is how they affect visibility. Everyone wants their voice to be heard, and some people want the opportunity to influence whether other people's voices are heard or not, e.g. by flagging stories or killing comments through downvoting.

If the big deal here is visibility, then I would concentrate on the algorithms that decide when a comment thread is rendered gray or invisible and the algorithms that decide the ranking of comment threads. I would look for patterns of votes or commenting that might help distinguish "popular but fluffy" from "popular and thought-provoking."

>>raganw+1f
Also, and this is really obvious, I would try to quantify the value of each position on the front page. Given an "average" story at #1, how many votes does it get? How long does it stay at #1? The answer is probably a curve of some sort, or a surface, but the general idea is that if you know what the typical #1 story gets, you can look at a story and determine whether it is outperforming stories in #1 position or under-performing them. You can do the same for all the positions on the page. You can determine the historical likelihood that someone looking at the front page will upvote a story in position #5

This data is interesting because now if you randomly perturb the front page, say by putting a new story in position #5 and lying about how many points it has. Show this perturbed page to a small sample of users and see if the story beats the historical averages or not.

If that kind of thing worked, it could help correct for the phenomena where a popular story stays popular just because it's popular. You can now rank stories by how well they take advantage of their "real estate" on the front page. Those that underperform the average sink, those that outperform the average rise.

One hitch. This might not work if you tell people you're doing it.