I wonder if the number of comments might be a better heuristic for measuring the variance in people’s perspective on the article. If you look at those 3 examples, the first had the most comments, but the least upvotes and lowest percentage positive.
If someone feels that they are in agreement and their viewpoint is already present in the discussion they might have a lower likelihood of adding another comment, but if there is a larger variance in the viewpoints on an issue than people would be more likely to have what they feel is unique information to add to the discussion.
As a continuation of that idea though. One of the prerequisites of factionalization / triblization is the existence in enough variance in viewpoints to create distinct independent clusters. Others in the same cluster become the in group, and those outside of the cluster become the out group.
However, while variance is required for clustering, clustering isn’t always present with high variance. You can still have more uniform distributions with large spreads.
Being aware that clustering effects are more likely in areas of high variance seems to me to a a good heuristic to internalize.
I wonder if the number of comments might be a better heuristic for measuring the variance in people’s perspective on the article. If you look at those 3 examples, the first had the most comments, but the least upvotes and lowest percentage positive.
If someone feels that they are in agreement and their viewpoint is already present in the discussion they might have a lower likelihood of adding another comment, but if there is a larger variance in the viewpoints on an issue than people would be more likely to have what they feel is unique information to add to the discussion.
As a continuation of that idea though. One of the prerequisites of factionalization / triblization is the existence in enough variance in viewpoints to create distinct independent clusters. Others in the same cluster become the in group, and those outside of the cluster become the out group.
However, while variance is required for clustering, clustering isn’t always present with high variance. You can still have more uniform distributions with large spreads.
Being aware that clustering effects are more likely in areas of high variance seems to me to a a good heuristic to internalize.