Currency in the economy of trolls is attention. When forums are small, trolling is much less rewarding and vice versa. Troll economy also feeds on to itself and

Currency in the economy of trolls is attention. When forums are small, trolling is much less rewarding and vice versa. Troll economy also feeds on to itself and upvoting/downvoting/karma mechanisms would be less and less effective as forum membership grows. To boost the attention economy, trolls would resort to upvote each other more frequently eventually outweighing votes of others. From theoretical perspective this is very similar problem to links and webpages. A troll is equivalent to a spammy web page. Upvotes/downvotes are equivalent to traffic you get on your web page. When UserA upvotes UserB, a link is created between two. The goal of a troll is to get as much traffic as possible. They are incentivised to give each other upvotes in the hope of return favor, or in other words, create as many links to each other as possible. It would be mistake to think that few "good guys" can be used as gate keeper to protect the system against these trolls. In nutshell that is the hope and approach many early search engines had and they failed as the size grew out of their hands. The solution has to be technical and automated. Algorithms like PageRank or machine learning models is highly applicable to trolling issue. For instance, the real value of karma should not be a naïve count of upvotes (in the same way that real important of the page is not how many other pages points to it) but rather who has upvoted it. I think algorithms like PageRank can be easily applied to calculate the value of karma. If trolls upvotes each other 1000 times, their net karma would be much less than 1000 de- incentivising them for putting in the efforts for upvoting. Of course, there are many ways to fool PageRank and but I think algos like this should be sufficient for forums were you don't have to deal with more sophisticated folks like search marketers.