As you may have seen, last weekend, Robert Scoble worked himself into a tizzy because an answer of his was moderated into a collapsed state on Quora. The whole bitchmeme surrounding the incident was rather humorous. But there was at least somewhat of a point behind all of it: people began asking more questions about the methods behind Quora’s ranking system. Today, we have some of those answers.
(2) We’re developing an algorithm to determine user quality. The algorithm is somewhat similar to PageRank but since people are different from pages on the web and the signals that are available on Quora are different from those on the web, it’s not exactly the same problem. We’ll use this to help decide what to show in feeds, when to send notifications, and how to rank answers.
Today, fellow co-founder Adam D’Angelo posted some of the details about the new Quora “Answer Ranking Scheme“. As he notes, the exact details are a bit complicated, and are going to change, but the main principles for this Quora “PageRank” are as follows:
- Answers with more upvotes are ranked higher.
- Answers with more downvotes are ranked lower.
- A vote from a user who has written good answers in the past carries more weight (both upvotes and downvotes).
- Answers written by users who have written good answers in the past will be ranked higher.
- Votes from people detected to be gaming the system (vote collusion, spam, bullying, etc.) will be ignored or minimized.
- Whether a user is an admin or reviewer does not affect the scores in this system.
So, essentially, the methods are pretty obvious. Answers with the most votes will be ranked higher (though gaming will not count). Votes from people with a history of good answers will be ranked higher. And answers from people with a history of good answers will be ranked higher. Makes sense.
And yes, anonymous answers are included in the above system because Quora will know who you are (they just won’t surface that information).
[image: Walt Disney Pictures]