I think zealous voting is indeed representative of the current userbase: For a long time the quality and diversity of links in Reddit was quite stable, reflec

I think zealous voting is indeed representative of the current userbase: For a long time the quality and diversity of links in Reddit was quite stable, reflecting the strong core userbase. Major systemic issues [1] led to more junk-food like submissions edging in, and slowly the forces involved drove out the original core audience, long-september style. This process continued gaining momentum, until fairly quickly lurching to the new stable state we find it in today. A similar process can be observed in wikipedia: those who care the most, have the most time, and the most tolerance for internet drama naturally dominate: the young and the passionate. 1. a) Reddit conflated two distinct actions into a single vote: training the recommendation engine (vote on what you like/dislike) and judging quality (what is good/bad). This lead to a culture of voting on taste, tending the system away from voting on quality, partly because of: b) The nature of these systems, like the mainstream media, is to tend to the lowest common denominator as audience size increases. This should not have been surprising. How to avoid these issues? Paul's suggestion (somewhere else on this page) about automated topic clustering is a good one. A similar approach I favor is encouraging clusters of users to form naturally, using a collaborative classification/filtering approach; even something as simple as allowing three tags per link could do this. Once clusters are established, it should be much easier to allow users to opt-in or out of the clusters they like/dislike. The editor/oracle approach and/or weighted voting should also work, but seems like a lot more effort. [ahem. I feel like I should mention I'm somewhat biased, as I'm now working on something experimental in this domain]