Can Flag Weight be available on data-dump and data-explorer?

Just curious :-)

  • 8
    Why? And to make this rather obvious question exceed fifteen characters again I ask, why?! – Shog9 Apr 26 '11 at 14:51
  • 3
    @Shog9 Why not? – waiwai933 Apr 26 '11 at 15:07
  • 1
    Why? Because it's information. The fact that it's not interesting to you does not mean that it won't be useful for somebody else. Edit: I would like to see accepted/declined flags ratios for different flags also. – Aleadam Apr 26 '11 at 15:12
  • @Aleadam: so it boils down to idle curiosity then? Normally, feature-requests carry a lot more weight when there's at least speculation as to how they might benefit someone. Features start out with -100 points - rarely do serious developers waste precious time developing features just for the heck of it. And BTW: the accepted/declined state of flags is not public information, and unlikely to become public any time soon. – Shog9 Apr 26 '11 at 16:30
  • @Shog9 I do not know how you got from "useful for somebody else" to "idle curiosity". One example on top of my head of information that could be extracted from that is a correlation between flag weight and rep, to answer questions such as: Is the 3k mark a good threshold to allow closing votes? Should it be increased/decreased? Is flag weight dependent on rep? Or it is more linear with respect to the time the user has been in SO? – Aleadam Apr 26 '11 at 16:44
  • @Aleadam: because "useful for somebody else" is useless for evaluating the need for a feature, unless "somebody else" comes along to enlighten us. Correlating flag weight and reputation is a specific use, and can thus be judged on its merits. I encourage you to post this as an answer, along with any other specific uses you might imagine for such information. – Shog9 Apr 26 '11 at 16:50
  • 1
    due to recent changes flag weight is no longer public ... will not be added to the data dump – waffles Jan 20 '12 at 8:25

There could be interesting applications to that data. As per Shog9 request, I will attempt to add some possible uses here. Sorry if it takes a strong research focus, but that's the scientist in me talking :)

Some of it may be already know by mods or the dev team. Also, please, take them only as examples of what could be done:

  • Plain statistical reports as Lance proposed in the previous answer. Mean, variation, mode of the general population to understand a little better what users groups there are in SO.
  • ANOVA (or nonparametric analyses if required) to compare the flag weight between different population groups (e.g., divided by rep count, age, history in SO, even subscription to other SE sites).
  • Correlation analyses flag weight vs rep or flag weight vs time in SO. Does it increase linearly in a particular range? If so, where does it reach a plateau? Where does it start increasing? That could help improve the threshold for closing votes privileges, for example. The weight vs time correlation could provide insights into whether rep alone is a good indicator of flagging capacity.
  • Analysis of the flag weights in different SE sites. Are meta sites flag weights better than the parent sites? Should be flag weight be handled differently for small sites compared to big ones?

It seems that is decided that flag accept/decline ratio will not be public. That is problematic because it would allow to discriminate between bad/mediocre flaggers and non-flaggers. If not a ratio, then a simple number of the total flags should be sufficient.

  • 1
    Currently, the only thing that exposes users with high flag weight is the 'deputy' badge. So, if someone has many flags, but no badge, you could pretty much discern that most of their flags were rejected. I don't think that's something everyone would want known about them in a database dump. – Tim Post Apr 27 '11 at 1:30
  • 1
    @TimPost: You can already see it on people's profiles, though. – Ullallulloo Apr 27 '11 at 2:19
  • 1
    @Tim I would understand your point if there's some privacy problem I'm unable to see here, but being able to see flag weights of 20 on the profile is not very different from what I propose in that respect. Furthermore, the same could be applied to the number of questions/answers vs total rep: having 300 questions and a rep of 150 would not speak well about the user, but the data is freely available for anyone to see. Anyhow, that has been discussed and decided, so I do not intend to advocate too much for it. – Aleadam Apr 28 '11 at 4:47

That is a good idea to be able to make some statistical reports on flag weight and other user variables.

You must log in to answer this question.

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