7
$\begingroup$

Judging by my own experience, reputation counting is far from perfect. Mine has been growing mostly thanks to answers to easy questions which I've been lucky to be first to read. On the other hand I have a couple of questions which might be too technical/specific for the majority, so I suspect they got upvotes only from few specialists in a narrow field. Also, I've seen lots of lightweight questions with many upvotes seemingly caused by superficially sensational/cool appearance of the formulation.

Would not it make sense to weigh votes by weights depending on (a combination of "ages" and) reputations of the voters?

$\endgroup$
15
  • 11
    $\begingroup$ Only somewhat facetiously, weighing votes in such a way might misleadingly suggest we were trying to make 'reputation' match some other quantity more perfectly, when it already perfectly reflects the stated rules for calculating 'reputation'. $\endgroup$ Mar 14, 2014 at 13:32
  • 15
    $\begingroup$ My advice: if you think "reputation" as now computed is not useful, then ignore it. $\endgroup$ Mar 14, 2014 at 14:14
  • $\begingroup$ @ScottMorrison You may understand my question precisely as asking whether the quantity measured by votes in the current way matches the word 'reputation' well enough. $\endgroup$ Mar 14, 2014 at 18:32
  • $\begingroup$ @GeraldEdgar I like the idea itself very much and I think measuring reputation is very useful, I am just asking whether everybody agrees that what is currently measured is really reputation or closer to something else like luck or ability to make impression. $\endgroup$ Mar 14, 2014 at 18:35
  • $\begingroup$ Let me add that if it is indeed the case that what is measured here is not exactly reputation but something else, then calling it reputation can be not just misleading but also harmful, I believe. $\endgroup$ Mar 14, 2014 at 19:04
  • 5
    $\begingroup$ While I would prefer a word other than reputation such as contribution to make this clearer, I would still insists on "reputation" of course meaning reputation as a user of this site. One can also get points (not too many and not too easily but still) by things that have nothing whatsoever to do with math in a strict sense via making successfull suggestions for edits. Conversely if I would start to post advertisments on the site I might get spam flags that kick in which would make me loose reputation. And rightly so as "a citizen of this site" but unrelated to my math competence. $\endgroup$
    – user9072
    Mar 14, 2014 at 19:40
  • 15
    $\begingroup$ I think it's very important we don't try to make MO 'reputation' reflect any kind of real world reputation. MO 'reputation' is the running total of a system of awarding points designed with the goal of encouraging good contributions to the site. I'm happy to think about that goal, as long as everyone realizes that's not really what the word reputation is meant to mean. $\endgroup$ Mar 14, 2014 at 23:05
  • 1
    $\begingroup$ @quid I certainly did not mean restricting the meaning of reputation to mathematical abilities, and I find measuring them by any kind of numbers impossible. Nice edits improving readability are also relevant here and are rightly included as appropriate part of local reputation. As all of those who commented here I am talking about reputation relevant to MO only. I just want to say that (a) things like luck or ability to impress are not relevant and should be ruled out; (b) after improving it this way, local reputation would be more commonly recognized and so trusted to have higher weight. $\endgroup$ Mar 15, 2014 at 8:26
  • $\begingroup$ @ScottMorrison Well MO is part of the real world anyway, right? :) And this reputation is indeed reputation in this part of the real world. And I think that currently it encourages contributions which can impress the majority, which does not necessarily imply that they are good. One of the things local reputation should be about is the ability to judge which contributions are really good and not just superficially likable by majority. $\endgroup$ Mar 15, 2014 at 8:32
  • 1
    $\begingroup$ And as I said there is the opposite side - the case of contributions which might be good but relatively technical/specialized and because of that get few votes since only relatively advanced users realize that they are interesting. $\endgroup$ Mar 15, 2014 at 9:30
  • 3
    $\begingroup$ @მამუკაჯიბლაძე in some way I agree quite a bit that in numerous ways the way these points are computed is not optimal. Here and also on Stack Overflow (its meta) they have numerous discussion like this, see things like for a 'historical' example meta.stackexchange.com/questions/9731/… but this got discussed over and over. Just to me starting to tweak it feels like opening a can of worms. In the end some will be a bit happier but likely more a bit unhappier, not because it will be worse but mainly as many do not like changes for such things. $\endgroup$
    – user9072
    Mar 15, 2014 at 10:26
  • $\begingroup$ @quid Very interesting link! If you are willing to turn this comment into an answer I will gladly contribute to raising your reputation :) $\endgroup$ Mar 15, 2014 at 10:41
  • $\begingroup$ One more thing. I cannot speak in the name of others (neither here nor anywhere else), but speaking of me I think it is good to try to prefer improving myself over making myself happier. $\endgroup$ Mar 15, 2014 at 11:04
  • 5
    $\begingroup$ This previous meta question also covers related ground: meta.mathoverflow.net/q/697 $\endgroup$
    – j.c.
    Mar 15, 2014 at 13:43
  • 5
    $\begingroup$ @მამუკაჯიბლაძე I expanded to an answer. Not that this is very important, but note: on meta.MO there is no reputation there is participation (computed in a quite different way, namely number of posts, comments, edits, votes cast [not received!] over a moving window of I think 6 weeks). $\endgroup$
    – user9072
    Mar 15, 2014 at 14:18

3 Answers 3

14
$\begingroup$

There have been discussions and ideas surrounding decaying votes, along with weights of votes themselves. I've got mixed feelings about both.

It's important to look at the utility of reputation, the degree that the system trusts you, before you consider the extrinsic motivation that it provides. Believe it or not, having someone arrive to your site and actually type something meaningful is quite an achievement. Having someone type something that is both meaningful and technically sound, however simple, is close to magic. I didn't realize how much this was the case until I started looking at tests that we run to try and understand visitor behavior. My point is, don't under-value contributions from knowledge that isn't rare, those are still valuable contributions.

Someone that has a track record of arriving to the site and doing meaningful, technically correct things should be given additional privileges - you'd be silly to not give them more ways to participate. Despite the actual extrinsic motivation, which is having a number associated with you increase as you see more and more people agree with what you wrote - the underlying system of trust is particularly good at giving the more skilled more tasks.

The idea of decaying votes becomes interesting because it would help to ensure that the most relevant, up-to-date answers would (at least, in theory) continue to rise to the top on older questions. On theoretical sites, or sites where domain knowledge is apt to change yearly, this becomes interesting. Your answer that got 400 up votes would need to continue to receive up votes as it aged, or answers that were receiving a greater velocity of votes would overtake it - a lack of votes on your answer being an indication that it isn't as relevant as it once was, and older votes would start falling off.

That would also be the mother of all headaches to implement. It's just an interesting theory.

Weighted votes for clearly demonstrated rare knowledge would be extremely cool, but not exactly accurate - if someone got 10,000 reputation for answering 100 very easy questions, does this qualify them to be a 'better' voter when it comes to those stellar answers that often go without due appreciation because they're difficult to understand? I'd tend to think not. If we did this, we'd have to do a combination of age, tag, rep, views to votes, snake oil and a little hoodoo - and it would still be wildly inaccurate and prone to abuse.

These are theoretical problems that I love to think about, because I do believe that an algorithmic solution could be in the midst.

Your best bet, for now is to share links to spectacular answers, and use the bounty system. If you really do believe that your rep was a little too easily earned, give it to users that tend to leave obsessively composed answers.

But, don't make the mistake of under-valuing your contributions here. Did I mention how hard it can be to get people participating on a meta site too? ;)

$\endgroup$
4
  • 2
    $\begingroup$ I think this is similar to what they do with paper citations where the ordinary PageRank does not work well. There is a TimedPageRank where you exponentially decay older links according to their edge. Obviously it is quite expensive computationally to do it from scratch but the good news is that there are incremental algorithms. One potential problem is that old users have probably already voted for old answer and cannot recast votes on old answers which is quite different from citations and web links, i.e. decaying might give too much influence to new users in determining what is top answers. $\endgroup$
    – Kaveh
    Mar 17, 2014 at 6:47
  • $\begingroup$ Thank you very much! You have formulated very accurately lots of important aspects around my question which I would never dream of! $\endgroup$ Mar 17, 2014 at 8:15
  • 3
    $\begingroup$ This is really a great answer. $\endgroup$
    – Joël
    Mar 25, 2014 at 3:46
  • 2
    $\begingroup$ "and it would still be wildly inaccurate and prone to abuse." An important note. When you complicate a system, you will always increase its "gameability" $\endgroup$ Mar 17, 2015 at 10:37
11
$\begingroup$

In some way I agree quite a bit that in numerous ways the way these points are computed is not optimal.

We had various discussions on MO, also on Stack Overflow (its meta) they have numerous discussion like this basically since 'always' see Fastest Gun in the West Problem for a 'historical' post on this there (but this got discussed over and over later on), and it is not much different for math.SE either.

Just starting to tweak it feels like opening a can of worms to me. In the end some will be a bit happier but likely more a bit unhappier, not because it will be worse in an absolute sense but mainly as many do not like changes for such things. I am also not sure it will be really better; likely one will solve one (perceived) problem and unwittingly create another one.

To give an example. Some (including me) feel that the Great Reputation Recalc change to +5 for question and +10 for answer (from 10/10) was perfectly reasonable and justified. (Disclaimer: I do realize that change favored me personally, but it is really not only or mainly this.) Some others think it is generally not good.

And, still some others have a sort of middle ground position, having some understanding for the change but feeling it should not have been applied retroactively. But then, I would have considered it as extremely problematic had it not been applied retroactively but only taken effect at some point.

Now, for the more specific idea of weighing votes. Personally I would have not much against this and it might have some merit though the precise details would be important to know to have an actual opinion. But then in any case I think it would increase an existing sentiment of some that there is some high-rep clique on the site that dominates everything or some such ideas (which are in my opinion misguided, but then presumably in such users minds I am part of that clique so no surprise there).

A somewhat related idea that also got discussed and which I like would be to have points not depend in a linear way on votes. But, say the effect of an upvote decreases with their number (to damp/cap gains that feel out of proportion for things that happen to get popular) while that of downvotes increases with their number (on the grounds that while everybody can get some downvote on everything if there is an accummulation somewhere there should be an actual reason).

The daily reputation cap at 200 (from votes) goes somewhat in that direction. But, at least for a site like MO this does not work that well in my opinion.

But in the end I feel the current system is somehow 'good enough.' That it is not that perfect or sophisticated can also be considered as a feature to highlight that the points are first and foremost a rough measure necessary as a tool to make community moderation work.

$\endgroup$
3
$\begingroup$

OK I give up :)

Let me just add one more thing only. Voting with no consequences whatsoever for oneself - I think this sort of amplifies the irresponsible part of the person...

Anyway, I hope similar concerns will be raised again by others frequently enough

$\endgroup$

You must log in to answer this question.

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