Should Quora ever consider using H-index for reputation?

Probably not - There is no way to answer this question without understanding the intended purpose of a public Quora karma score. Would it be meant to encourage more contributions? Or more upvotes (and therefore vote trading abuse?) See http://buildingreputation.com/doku.php?id=chapter_5%5C#asking_the_right_questions for a more detailed discussion of goal-setting when establishing a reputation system. Also, we've posted http://buildingreputation.com/writings/2010/02/on_karma.html a list of warnings when undertaking a new system of karma, particularly: Karma is user reputation within a context Karma is useful for building trust between users, and between a user  and the site Karma can be an incentive for participation and contributions Karma is contextual and has limited utility globally. [A chessmaster  is not a good eBay Seller] Karma comes in several flavors - Participation, Quality and Robust  (combined) Karma should be complex and the result of indirect  evaluations, and the formulation is often opaque Personal karma is displayed only to the owner, and is good for  measuring progress Corporate karma is used by the site operator to find the very best  and very worst users Public karma is displayed to other users, which is what makes it the  hardest to get right Public karma should be used sparingly - it is hard to understand,  isn't expected, and is easily confused with content ratings Negative public karma should be avoided all together. In karma-math  -1 is not the same magnitude as +1, and information loss is too  expensive. Public karma often encourages competitive behavior in users, which  may not be compatible with their motivations. This is most easily seen  with leaderboards, but can happen any time karma scores are prominently  displayed. [i.e.: Twitter follower count]

Yes Design is all about tradeoffs and every reputation system will have  flaws that can be critiqued. The ideal reputation system should be like  democracy, the worst form of measurement except all the others. On the power-simplicity scale, I think raw h-index just about hits the  sweet spot. There are possible enhancements (g index, higher weighting for more knowledgeable commenter, etc.) but they seem to add too much complexity for the quality gain. There are possibly simpler systems (karma, batting average) but I contend they don't do a good enough job of incentivizing the right things. Advantages  When it comes to Quora, the h-index has a number of advantages over karma systems and weighted  average systems: It's hard to game. Each "trick" you exploit has a finite lifespan since your h-index will rise until the point where the trick becomes ineffective. Also, since it's possible to display all the answers that comprise your h-index, it's very easy for a outside observer to spot whether your high score was due to gaming or genuine high quality answers. Like averaging systems, it quickly rises to reflect a person's  "natural ability". Karma systems reward "grinding". It's impossible to look at a person's Karma and figure out if they're someone who gives a small number of high quality answers or a large number of low quality numbers. H-indices reflect the "natural ability" of the commenter much more accurately since additional effort won't increase it without an increase in the quality of the answers. It biases for quality over  quantity. For the majority of people, the only way to increase their H-index is to write more high quality answers. It  doesn't discourage people from answering long tail questions. One of the downsides of weighted average systems is that they discourage people from contributing where there is the risk of reducing their reputation. Since H-indices can never go down, this is not a worry It motivates people to become better. Once you have topped out your H-index, the most effective way to increase it is to increase your "natural ability" as an answerer. This means stuff like being more widely read, thinking more carefully about answers, experiencing more of  the world etc. Criticisms it under-measures the performance of people who don't participate enough (aka: The Galois problem. Galois would only have a h-index score of 2 since he only ever published 2 papers). In this respect, it is like the ELO system in Chess. Ultimately, I don't believe this to be a major drawback. People who only participate a little and then never come back probably don't care what their reputation is in the first place. it may be too complex conceptually for the average user to grasp. While the math behind the h-index can be somewhat hard to imagine, I think the mental model is rather simple: *) If I have a h-index of 15, that means posting an answer that gets more than 15  upvotes will increase your h-index by 1 *)  My internal ranking scale allows me to interpret that 5 = noob or amateur, 10 =  someone worth listening to, 20 = great commenter, 50 = one of the  superstars of the site. it de-incentivizes people to post comments they know will be below their h-index. I'm open to the possibility that this may be true but I think this ultimately won't be the case. Most notably, because the h-index is monotonically increasing, nothing you do can ever decrease your h-index. Also, the nature of the h-index is that it will rise rapidly at the beginning and then become increasingly harder to increase so once you reach that point, the best strategy is to no longer worry about how to increase it and just participate as you normally would. Ultimately, I think if you tally up the pros vs cons with respect to Quora specifically, the h-index wins out over every existing reputation system (including no reputation system) and it should be considered a strong candidate.

I think this would move in the right direction, but it would need refinement for use on Quora. The following is an exploration of the possibilities; skip to the bulletted paragraph for the metric I think would be best. What are the flaws in h, and what other measures might provide a view without such flaws? There are two possible problems with any karma measure. It can be lower than it should be for some people, or, more seriously, it can be higher. Undeserved low H - the "Galois" effect - is not a big problem, because if people care, they can correct it, and in doing so, improve Quora. So the problem is undeserved high H. Why would that happen? One way is if people bring their Twitter posse to Quora. That is, they have many "followers" (in Quora, Twitter, wherever) who see all their answers. Even if these followers are judicious about upvoting, such a person will naturally get some fraction of their followers to upvote any given answer. A partial remedy for this problem would be the "deep-H-index", which is the max H such that you have H answers with H upvotes from people with a (simple) h-index of H. A legion of naive fans would not help you in this regard; you would only be pursuing upvotes from people whose h is higher than your deep-H. Another way to handle the "famous people posse problem" would be to somehow measure posse size (answer page views?), and highlight people whose h-index is larger than you'd expect given their posse size. For instance, instead of an overall leaderboard, have a "deserves more followers" page, where you'd see people whose number of followers was less than their h. This would be a self-correcting imbalance, because those people would pick up more followers. But what if, instead of their twitter posse, people brought their Facebook clique to Quora? With mutual upvotes, or even explicit vote trading, a group of H people could all attain an h-score of H. To fix this, one could use the "broad-H", your h-index counting no voter more than once. A broad-h of 10 would mean you had 10 answers with 10 upvotes from a total of at least 100 unique users. A mutually-upvoting clique would have a broad-H which was only the square root of their size. Similarly, your twitter posse would only be good for an broad-H of the square root of its size. (Checking your broad-h exactly is an O(N^2) operation, but a quick-and-imperfect O(N) algorithm would find h, then only check how many of the voters for the lowest-voted question in that set hadn't voted for other questions in that set.) Deep-H would encourage people to seek votes from elites; broad-H would encourage them to seek new voters. Since you already are rewarded for votes from elites (all their followers see, and possibly upvote, your answer), I'd say broad-H would be the best single number in value added. Still, as a numbers geek, I'd enjoy being able to see more than one metric for myself or others.