I think this might be the most salient point of all.
What would the data look like if we were to look at the correlation between the timing of a comment (compared to the timing of the original post) and it’s vote rank?
I’d guess it is likely that the first “good” comment drives the discussion, as it will be read by more people, voted on more often, responded to more frequently, etc.
What if no comments were displayed on articles for 48 hours after publication of the article? At the 48-hour mark, all the comments recieved would be displayed simultaneously and made available for vote rank & response...you have ~20 comments, displayed in random order (is it possible to code for unranked comments to display in random order on a per-login basis?) Would this help remove the (hypothesized) early-comments-win bias?
I think this might be the most salient point of all.
What would the data look like if we were to look at the correlation between the timing of a comment (compared to the timing of the original post) and it’s vote rank?
I’d guess it is likely that the first “good” comment drives the discussion, as it will be read by more people, voted on more often, responded to more frequently, etc.
What if no comments were displayed on articles for 48 hours after publication of the article? At the 48-hour mark, all the comments recieved would be displayed simultaneously and made available for vote rank & response...you have ~20 comments, displayed in random order (is it possible to code for unranked comments to display in random order on a per-login basis?) Would this help remove the (hypothesized) early-comments-win bias?