3 ms·
More like the latter, but you don't get to choose. The system automatically clusters you with others whose historical votes are like yours. So what you get is "
by eevilspock 12y ago
More like the latter, but you don't get to choose. The system automatically clusters you with others whose historical votes are like yours. So what you get is "true karma" where the collaborative filter you get is a direct result of the quality of your votes, as opposed to karma being a score as it is on HN. You may think you are an open and fair-minded thinker, but if you vote ideologically, or vote with things you want to be true whether or not they are, you will be clustered with the same. It would be like Apple fanboys and Android fanboys getting put into their own clusters, getting out of the way of others who engage in "agnostic" fair-minded discussions. [Typo in prior reply. Should be "A lot of people... who think like you think this item is garbage."]
Astro-turfers and bots trying to game the system naturally don't cluster with anyone except those who happen to agree with them, in which case who cares? There really is no difference between fake votes and dumb votes, so the system doesn't make that distinction. Trust is asymmetrical, so even even though your filter is influenced by those in your cluster, you may have no influence on your cluster. Trust is earned by voting on something before others do, and then those others agreeing with your vote. So it is impossible to game trust by mimicking the votes of others. The system distinguishes between leading votes and following votes.
The clustering is entirely transparent... You can "flip" your view at any time to one of the other clusters. I have a number of ideas how to surface a meaningful label for each cluster so that, for example, a self-professed fair-minded user can realize he is actually rather close minded and ideological. I call this "Mirror". One's cluster would never be publicly displayed (not in the business of shaming), but you can look in the mirror in private, and gain self-awareness.
The system would support the maximum privacy and anonymity guarantees possible (allowing people the option to register with no password recovery email, for example).
Before the system has enough users to do effective clustering, it would simply flag garbage that has widespread agreement as such. That in and of itself is a valuable service.
My original project name for this is "Garbage" (How much of the web do you think is garbage?), but I've settled on "Common Karma" as the official name. This is my first sharing of the above idea and its names beyond my friends. You heard it here first!
- walterbell 12y agoSounds quite promising, look forward to seeing your algorithms take on professional astro-turfers (human or bot) who may have algorithmic assistance. This 1997 essay on collaborative filtering has the baggage of an "extropian" perspective, from Sasha Chislenko (RIP). Some of the social speculation remains relevant, http://www.lucifer.com/~sasha/articles/ACF.html http://www.lucifer.com/~sasha/articles/ACF.html This was in the early days of ACF, http://www.strategy-business.com/article/19707?gko=acda4 http://www.strategy-business.com/article/19707?gko=acda4