tea.mathoverflow.net - Discussion Feed (Statistically "correcting" votes?)

grp comments on "Statistically "correcting" votes?" (12506)

grp — Mon, 10 Jan 2011 22:56:45 -0800

Harry, I am composing a reply to send to (what I believe to be) a gmail address of yours. I will welcome further discussion by email. Briefly though, your imagining is a crucial part of the "System Design" picture.

Gerhard "Making Change, Person by Person" Paseman, 2011.01.10

Harry Gindi comments on "Statistically "correcting" votes?" (12469)

Harry Gindi — Mon, 10 Jan 2011 09:23:20 -0800

Gerhard, I am asking you about system design right this second. What does it mean specifically? I would imagine that the term "System design" can really mean almost anything, from pure math to civil engineering and beyond. I can't find your e-mail address on usenet, so just explain it here!

Just had to get that off of my chest.

grp comments on "Statistically "correcting" votes?" (12468)

grp — Mon, 10 Jan 2011 08:04:03 -0800

I agree the question is interesting, but am against the idea of implementation. This is mainly because I think there is little meaning to the actual numbers and not much more to, say floor(log(rep)) or other rescaling.

At the panel on metrics in research at ICM2010, I suggested that such online voting systems be studied so that people who need such metrics could adapt them to their needs with full knowledge of the systems' shortcomings. If such adaptations ensue, the meanings behind the votes will be crafted/adjusted/invented for each adaptation, and I doubt that one system's meaning will be any more than a rough guide to the meaning associated with an adaptation.

I think that this question is closer to meta than MathOverflow, and should focus not on correcting votes but on interpreting them and applying meaning. In which case, I add the idea that a response which resonates well with many readers gets upvoted. This may be more applicable to comments than to answers, but people who find an answer already crafted that they may have wish they had written (which may contain common misintepretations of the original question) are likely to vote up that answer, I believe.

Gerhard "Ask Me About System Design" Paseman, 2011.01.10

Steve Huntsman comments on "Statistically "correcting" votes?" (12467)

Steve Huntsman — Mon, 10 Jan 2011 07:58:56 -0800

See also http://tea.mathoverflow.net/discussion/641/deciding-what-to-vote-up/#Item_15 and preceding comments.

Timothy Chow comments on "Statistically "correcting" votes?" (12466)

Timothy Chow — Mon, 10 Jan 2011 07:38:10 -0800

The number of votes an MO question receives is influenced by several factors other than the intrinsic appeal of the answer to the readership. The two factors that come to mind are the timing of the answer (early answers are more likely to get more votes) and the reputation of the answerer (both MO reputation and real-world reputation). Perhaps there are other factors as well.

I'm curious as to whether there is any plausible way to "adjust" votes to "correct" for these factors. It's not immediately obvious to me how to do so, but at the same time I sort of think there might be a way.

Of course, I'm not suggesting that if someone were to come up with a proposed adjustment algorithm that we should try to implement it (even assuming that the software allowed it, which I'm sure it doesn't). However, apart from the fact that this seems to be an interesting question in its own right, it could become important one day if some kind of MO-like voting system gets implemented for, say, the mathematical literature in general as a sort of informal peer-review system, and people seriously want to know what to make of the votes.

I was tempted to make this an MO question rather than a meta question, but I chose to follow Linus's example.