4 ms·
Multiply the reward by (1 / (flags_post / (flags_all_posts / n_posts))) * (1 / (downvotes_post / (downvotes_all_posts / n_posts)))
by allendoerfer 11y ago
Multiply the reward by
(1 / (flags_post / (flags_all_posts / n_posts))) * (1 / (downvotes_post / (downvotes_all_posts / n_posts)))
- dang 11y agoCan you define these terms? I don't understand what they mean exactly. At first glance, this expression looks like a positive number, meaning you could still get a reward by upvoting everything. If we ever roll out something like this, it has to not be possible to write a bot to gain karma—except, of course, by writing a bot the identifies the best stories, in which case you deserve it.
- allendoerfer 11y agoSure. What I wanted to say is just, that you can create a measurement for low-quality posts through dividing the number of flags and/or downvotes the post got by the average number the other posts had had. Take the inverse of that product and you have a nice factor that approaches 0 for low-quality posts and 1 for high quality posts. I chose a simple example, but of course you could use a more appropriate statistical method, apply a logarithm afterwards or just choose a threshold and reward only posts with 0.9 to 1. Edit: I just realized I made a mistake. Of course, posts do not have downvotes. Flags would still work, but to a lesser extend, I guess. There has to be another method to identify low-quality posts. Maybe some measurement based on the number of comments, their upvotes and so on. That would get really complex, because these metrics again would have to be ungameable.