-11

I believe it should be possible to score user reputation using an algorithm similar to page rank.

The idea is up-voting from a higher ranked stack user should weigh greater than an up-vote from a lower ranked user. A more important user is one who has received up votes from important people (just like google's page rank, a page is important if it is linked to by important pages)

This will solve problems of the kind where people create a clique and keep up-voting each other within the clique (I don't know if this happens, just possible). It feels like a more equitable system than the current static scoring mechanism. It would also give opportunity for bright and young minds to move up the ranks on stack faster.

|improve this question
  • 1
    I dont think this would fly. voting is equal, not based on rank – Naftali aka Neal Jul 11 '11 at 15:42
  • 4
    I have a highish rep on SO, but that doesn't mean I'm any more qualified to upvote than anyone else. – user7116 Jul 11 '11 at 15:42
  • Should voting be equal? Are people equal? those are philosophical question. I think opinions of people should be weighed based on experience and 'believability'. For example: Guido making a up vote on a python tagged question/answer is more important than me doing the same. – GeneralBecos Jul 11 '11 at 15:46
  • What about those new members asking questions? That would make people much less inclined to answer, among all the other problems. – Ry- Jul 11 '11 at 15:50
  • @GeneralBecos, it's true that people are not all equal in terms of technical ability. Sadly, there just isn't a good algorithm for assigning vote weights to users. Not even close. At least giving every person exactly one vote per post is simple to both understand and implement. – Pops Jul 11 '11 at 15:51
  • 1
    See If a user has good rep on a particular tag, shouldn't his votes on that tag weight more ?, Indicate How Trustworthy the UpVotes Are, Weighted Down-Voting based on Reputation – Tim Stone Jul 11 '11 at 15:54
  • @Popular Demand, the page rank is an excellent algorithm for solving the exact problem you mention. – GeneralBecos Jul 11 '11 at 15:57
  • 1
    @GeneralBecos, I strongly doubt that that's true, for the reasons explained in Bill's answer and the posts Tim linked to. – Pops Jul 11 '11 at 16:01
10

The main problem that I have with this is that I have a fairly high reputation on Stack Overflow, and I know that there are topics where my vote should just not count more than anyone else's. I'd like to give extra upvotes to certain posts, but I don't think it would be fair. I don't think it would be any different than using a sock puppet to vote for things I like.

This will solve problems of the kind where people create a clique and keep up-voting each other within the clique (I don't know if this happens, just possible).

This does happen. We detect it and invalidate the votes.

|improve this answer
  • 1
    You might want to add a link to Vote Fraud and You – Some Helpful Commenter Jul 11 '11 at 15:51
  • It should be possible to identify 'believability'/'expertise' based on the manner in which experience was earned. Like if a person asks a lot of python questions and has a high rank, he is likely a python expert. This means if this very reputable python person up votes an idea around say philosophy, the vote would count for less than a vote for python. Difficult to implement, but fairly automated + you don't need to manually clean up. – GeneralBecos Jul 11 '11 at 15:55
  • Thanks @Conrad, I added the link. – Bill the Lizard Jul 11 '11 at 15:57
  • @GeneralBecos: That would be possible to implement, but I still don't think it would fly. Tagging isn't perfect. I myself have badges in tags where I've barely participated on purpose, simply because questions I've answered had a tangentially related tag (the similarities between c# and java, and c# and vb.net often have this affect). Also, much of the clean up is now automatic, if not most of it. – Bill the Lizard Jul 11 '11 at 16:01
  • @Bill Thanks for the link. Could anyone share how the spam upvotes/downvotes are detected. The reason I ask is to know if it is 'gameable'. If it isn't gameable, my guess is you are already using some algorithm similar to page rank to detect gaming. – GeneralBecos Jul 11 '11 at 16:15
  • 1
    @GeneralBecos: We don't know the exact algorithm used because making it public would certainly make it more gameable. – Bill the Lizard Jul 11 '11 at 16:18
  • Again, in the absence of quality tags, I agree this wouldn't fly. But given the volume of tags, it seems fairly easy to determine a users expertise. More data should solve that problem. – GeneralBecos Jul 11 '11 at 16:19
  • But yeah, I foresee severe cold start problems for new users. – GeneralBecos Jul 11 '11 at 16:20
  • @GeneralBecos: Can you elaborate on what you mean by "severe cold start problems for new users"? Most new users are only interested in getting the answers to a few questions and don't really get into the "competition" aspect of the sites until later. The weekly, monthly, and quarterly views on the users page was intended to help alleviate that "I can never catch Jon Skeet" feeling that some of us had. – Bill the Lizard Jul 11 '11 at 16:23
  • Specifically, the problem I mentioned is for users when there isn't enough data to identify expertise. (since they have few questions with many tags, likely more tags than questions. We want to be in the situation where we have more questions than tags). Given that we can't identify expertise adequately enough early, it would be difficult to gauge what a users upvote/downvote is worth. – GeneralBecos Jul 11 '11 at 16:31
  • @Popular Demand: Given that most users on stack are serious (per your comment), I think this page rank system should work well enough to be practical... hmmm. – GeneralBecos Jul 11 '11 at 16:36
4

This will not fly.

Voting is equal and not based on rank.

SE is a democracy when it comes to voting (or at least it should be) and no one user should have higher power to vote then others.

When it comes to other things like privileges -- yes that is based on rep.

|improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .