4 ms·
> the 95% confidence interval is much wider for them If you change this to an 80% confidence interval, it'd become narrower and might actually somewhat favor n
by KajMagnus 14y ago
> the 95% confidence interval is much wider for them
If you change this to an 80% confidence interval, it'd become narrower and might actually somewhat favor new comments, with only a few votes (upvotes)? So this might be configurable?
> whereas an opinion with 1000 upvotes and 500 downvotes is probably more interesting to read than one that only gathered 50 upvotes within the same interval
Isn't it more likely that the 1000 upvotes and 500 downvotes is a cute kitten photo? Or something similar (a very short but strong and popular comment)? :-) And that the one that gathered 50 upvotes and no downvotes is the truly interesting read?
However! If you're able to estimate how many people actually read the comment that got 50 upvotes — then you'll know if it's truly interesting, but too long to read — only 50 people have read it. Or if it's boring — 1500 people have read it.
Perhaps a good sorting algorithm would be: Lower bound of confidence interval for:
(upvotes - downvotes) / num-people-who-read-the-comment
- robryan 14y agoLetting different comments reach the top wouldn't be such a bad thing most of the time, as long as there is more than 1 solid comment. Only time this would be bad is if there is something important, such as a correction or a key reply by the author or the person they were calling out etc where it makes sense that everyone should read first.
- Dave9k 14y agoKnowing how many read a comment would help with comment assessment but how do you get that information. I suppose we could track scroll positions, I'm not sure how reliable it would be. It'd be cool if websites could use eye tracking, then we could easily tell what got read. Maybe in the future.
- kansface 14y agoThe problem is that kittens are quite popular. They are more popular than a well formed discussion. What algorithm separates kitten posts from quality content? How could you tell the difference between a great photo and a lolcat? I think the only way to do this is to empower moderators. Any other scheme is just a race to the middle for bland, unoffensive and easily digestible content to maximize points.