A study [PDF] published in a journal of the Association for the Advancement of Artificial Intelligence found that sites that have a "downvote" button to punish bad comments lock the downvoted users into spirals of ever-more-prolific, ever-lower-quality posting due to a perception of having been martyred by the downvoters.
What's more, positive attention for writing good posts acts as less of an incentive to write more good stuff than the incentive to write bad stuff that's produced by negative attention.
Leskovec and his colleagues don't look at Reddit, but instead at four other prominent information-sharing Web sites with roughly similar mechanics (these sites, like Reddit, allow users to "upvote," or "downvote" posts and comments). They use some complicated statistical and experimental techniques to reach two key findings:
(1) People who write low quality posts are more likely to write again when they get negative attention. Furthermore, the quality of their posts deteriorates. This goes beyond the simple adage that you shouldn't feed the trolls by giving them attention. The evidence suggests that negative feedback can perhaps actually create trolls. It also suggests that people getting negative feedback are more likely to give others negative feedback, too, spreading the infection.
(2) People who write high quality posts are encouraged by positive attention to write more. However, they aren't as encouraged by positive attention as bad posters are by negative attention. Furthermore, the quality of their posts does not go up. Broadly speaking, encouragement doesn't seem particularly effective.
How Community Feedback Shapes User Behavior [Justin Cheng, Cristian Danescu-Niculescu-Mizil, Jure Leskovec]
Why Reddit sucks: some scientific evidence [Henry Farrell/Washington Post]
(via Making Light)
(Image: /disapprove, hobvias sudoneighm, CC-BY)