Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.
I only browsed this quickly and have no expertise in this, but it seems interesting to me. I do not see a reason to be conerned about too much credit given (there is not that much credit given, it is more an analysis of a couple of existing things, including MO; that is for the longer article the shorter is focused on MO but also in the sense of 'we study this' as opposed to 'this is the only thing to study') or about insufficient historical background.
The main credit given is that MO is described as 'very effective' (as 90/100 of the questions are answered, fully or in part). And, I would share the opinion that MO is very effective at getting a certain type of question answered. (The types of questions are also analysed/described.)
There are some details were I feel the description is slightly off, in particular concerning meta.MO. But in general as said it seems quite accurate. Finally, it seems the 'social machine' does not originate with the authors' of this article but is some sort of technical term in that context.
(as 90/100 of the questions are answered, fully or in part)
Isn't this a reason "easy" questions for which the answer is only a highly voted comment to have real (possibly CW) answer?
Just browsed the papers and suppose the results come from automatic analysis of the public dumps.
I do not think this is (mainly or at all) based on an automatic analysis; by contrast it seems a sample of questions was analysed in detail (100 in group theory or so; see the shorter paper).
I am not sure I understand what you (joro) mean, but there are plenty of reasons why I think (open) questions with comment-answers are not good. That it would affect the quality of an automatic analysis is not high (or at all) on the list of reasons for this, though.
First, is anyone here familiar enough with the work above to explain it better, and perhaps knew about it before it was posted to ArXiv?
I was not aware of these two articles until very recently, but there has been a fair amount of interest in MathOverflow in the social computing community. (More here, for example.) I don't think I can explain the work any better than they do. When I was curious about some details, I contacted the authors directly and they were always very responsive.
Second, although this may be a curmudgeonly viewpoint, I suspect that more credit for the success of MathOverflow is being given than it is due, possibly because of insufficient historical background (E.g. the Manhattan project, older forms of 'crowdsourcing'); does anyone here share the concern that the claims of effectiveness might be exaggerated?
Human-computer interaction is a fairly recent field of study but it didn't spawn from thin air. These studies use specific methodologies to explore different aspects of MathOverflow and all the papers I read describe their methodology in sufficient detail. In particular, the statistics in the two papers are based on a sample of 100 questions within the [gr.group-theory] tag 100 questions drawn from April 2011 and July 2010 to obtain a spread. This is obviously a biased sample, but it's an interesting choice since [gr.group-theory] is currently the fifth most active tag.
I looked at the first of the two papers, and it actually seems very sensible and quite interesting!
1 to 7 of 7