In the “top 10” aggregate, you are at risk of the following Simpsonian problem: you have two posters A and B; one writes longer comments than the other and also happens to be cleverer / more interesting / funnier / better at appealing to the prejudices of the LW crowd. So in the whole group there is a positive correlation between length and quality, but actually everyone likes A’s shorter comments better and everyone likes B’s shorter comments better. (Or, of course, likewise but with “longer” and “shorter” switched.)
It’s an interesting possibility. But I have looked at the data and for all ten users the comments above 1000 characters get higher average ratings than shorter comments.
In the “top 10” aggregate, you are at risk of the following Simpsonian problem: you have two posters A and B; one writes longer comments than the other and also happens to be cleverer / more interesting / funnier / better at appealing to the prejudices of the LW crowd. So in the whole group there is a positive correlation between length and quality, but actually everyone likes A’s shorter comments better and everyone likes B’s shorter comments better. (Or, of course, likewise but with “longer” and “shorter” switched.)
It’s an interesting possibility. But I have looked at the data and for all ten users the comments above 1000 characters get higher average ratings than shorter comments.
Aha, excellent.