Avoiding the Fark Effect

By the Fark effect I mean - at first a social news site is really cool, then over time there are more and more political posts, top ten lists, nsfw links, and pictures of cats. Inside their algorithmic brains, every social news site wants to turn into Fark.

In the wake of being linked on techcrunch a lot of people seem worried the same thing will happen here.

Here's a suggestion - build a bayesian filter specifically to prevent Hacker News from turning into reddit / digg / fark. Positive training data can be hacker news articles before today. Scrape the contents of each link before it goes live. Negative training data can be the current top n articles from those other sources. Don't apply this filter in a boolean way; instead do something like

score *= similarity-to-classic-hacker-news

By making it non-boolean, we will still get articles like "Google announces huge new product". It will just cost you a 0.1x in score whenever the article mentions Ron Paul.

Any other good ideas out there?