A downarrow doesn't have to work in the exact opposite way as the uparrow. In the trivial case the downarrow would do nothing, and then site wouldn't change at all, except cosmetically. In the more generalized case you would:
-weight it versus upvotes
-weight it based on user history
-weight it based on user's propensity to downvote
-weight it based on the number of distinct "user clusters" it offends
(If one group always votes for butter-side up stories, and another always votes for butter-side down stories, and a story annoys both of them, chances improve that the story is worth voting down.)
Another interesting metric might be contentiousness. If a topic generates an outlying number of downvotes from trustworthy users, and an insignificant number of upvotes (or many upvotes from unpopular users), its participants can be labeled as positive or negative forces. Someone who posts articles that generate a lot of downvoting in the comments might not be picking the best topics. There's probably a nice linear algebra way to generalize this in the same vein as PageRank.
A closed-source, nearly-null downarrow would at least give pessimistic users the feeling that their voice was being heard. (Heuristics is one area where it makes sense to keep things closed-source. Changes should be explained, but not necessarily in exacting detail.)
-weight it versus upvotes
-weight it based on user history
-weight it based on user's propensity to downvote
-weight it based on the number of distinct "user clusters" it offends
(If one group always votes for butter-side up stories, and another always votes for butter-side down stories, and a story annoys both of them, chances improve that the story is worth voting down.)
Another interesting metric might be contentiousness. If a topic generates an outlying number of downvotes from trustworthy users, and an insignificant number of upvotes (or many upvotes from unpopular users), its participants can be labeled as positive or negative forces. Someone who posts articles that generate a lot of downvoting in the comments might not be picking the best topics. There's probably a nice linear algebra way to generalize this in the same vein as PageRank.
A closed-source, nearly-null downarrow would at least give pessimistic users the feeling that their voice was being heard. (Heuristics is one area where it makes sense to keep things closed-source. Changes should be explained, but not necessarily in exacting detail.)