Andy Baio has an interesting post about a recommendation algorithm he and a partner created to color a writer's words red or blue to reflect political leanings. A writer's bias is weighted based off of links to sites deemed appealing to either the political Right or Left. The program is site specific, having been developed for use on MEMEorandum.
This is a great example of social feedback technology. At a glance, a user can make a judgment about the content in question. I don't know if it's desirable for people to make rush judgments of this type, but damned if the program doesn't make them easy!
I would like to know if general usage promotes negative or positive feedback. If writers desire to be seen as objective, they may seek to balance their links. A negative feedback loop could drive the content towards a calm middle ground. Or perhaps bias drives traffic and writers will strive for Red or Blue through and through? MEMEorandum could become a totally bipolar site. Here's a situation where an outsider's tool could shift the output of a system*. Fun stuff!
*As I understand it, the program doesn't use live information. So "shifting the output" is only theoretical.