whoever relays stuff 🇵🇸🏴☠️🇺🇲 on Nostr: In the long run (after launching with something simple) I would use A/B testing to ...
In the long run (after launching with something simple) I would use A/B testing to determine scores based on relative performance.
An example of how it could work (not the only example) -
By default, every other time the user loads a thread, their sorting could switch back and forth between "best comments" and "random." They would still be able to manually override this.
The "random" tab wouldn't actually be completely random, it would always be trying to show 2 different users 2 different sorting arrangements that can be compared to each other - like, users A and B get post X first and post Y second in the thread, but users C and D get post Y first and post X second, seeing which ranking results in more upvotes vs downvotes for all posts overall.
This is useless with too few users, so there's no point launching with it, but with enough users it gets the data needed to fill the "best answers" tab with the actual best answers. It's very resistant to newer answers being drowned out by older answers (which had more time to collect votes). It's very resistant to popularity contest bullshit, because an unpopular answer that gets a lot of discussion will still be recognized for its contribution to the thread, instead of buried in downvotes (i.e. reddit).
An example of how it could work (not the only example) -
By default, every other time the user loads a thread, their sorting could switch back and forth between "best comments" and "random." They would still be able to manually override this.
The "random" tab wouldn't actually be completely random, it would always be trying to show 2 different users 2 different sorting arrangements that can be compared to each other - like, users A and B get post X first and post Y second in the thread, but users C and D get post Y first and post X second, seeing which ranking results in more upvotes vs downvotes for all posts overall.
This is useless with too few users, so there's no point launching with it, but with enough users it gets the data needed to fill the "best answers" tab with the actual best answers. It's very resistant to newer answers being drowned out by older answers (which had more time to collect votes). It's very resistant to popularity contest bullshit, because an unpopular answer that gets a lot of discussion will still be recognized for its contribution to the thread, instead of buried in downvotes (i.e. reddit).