Do Best Answers affect AskMe life? October 28, 2011 7:17 AM Subscribe
Does the presence of a marked Best Answer in an AskMe keep it alive shorter or longer?
That is, is there a distinct difference between the amount of time from posting the question to the last answer in questions with a marked best answer and a question without a marked best answer?
That is, is there a distinct difference between the amount of time from posting the question to the last answer in questions with a marked best answer and a question without a marked best answer?
It might depend on the type of question. A math or science or movie trivia question with a marked BA is probably dead in the water; a relationshipfilter or other subjective-type thing would be more interesting, but you'll have to massage the infodump to get down to that set.
posted by shakespeherian at 7:32 AM on October 28, 2011
posted by shakespeherian at 7:32 AM on October 28, 2011
The causal question is unanswerable: questions which can be best-answered (because they are well formed) are very different from those that can't be. I think you'll find that BA questions are different in measurable respects before the BA happens. At the least I'd compare timestamp lives of matched questions which have best/non-best answers at the same age, or maybe matching on some other features of the thread.
Some people BA answers long after the thread is dead, that might be a useful contrast to look at since it partially isolates BA as a poster behavior from questions which are likely to generate good answers.
posted by a robot made out of meat at 7:35 AM on October 28, 2011 [1 favorite]
Some people BA answers long after the thread is dead, that might be a useful contrast to look at since it partially isolates BA as a poster behavior from questions which are likely to generate good answers.
posted by a robot made out of meat at 7:35 AM on October 28, 2011 [1 favorite]
I don't think the infodump records a timestamp of when the best answer was designated. There is probably a difference between posting in a thread that already has a designated best answer to posting in a thread that doesn't have a best answer designated yet but already has that "future best answer" already.
posted by demiurge at 7:37 AM on October 28, 2011
posted by demiurge at 7:37 AM on October 28, 2011
I thought about that, but I had no idea if the time-of-Best-Answer-marking was tracked. Plus I didn't want to make it any more complicated for, well, whoever ends up doing this before I get a chance to screw around with it. Plus all my buckets have been confiscated and my massage license revoked ever since the incident.
posted by griphus at 7:37 AM on October 28, 2011
posted by griphus at 7:37 AM on October 28, 2011
This would actually be a pretty simple one to calculate, I think ... calculate the time difference from post to last comment for the two classes.
No, as is so often the case, the statistics alone wouldn't tell the whole story about what's really going on. For instance, if there's a decisively correct "best answer" early on, people often take advantage of the chance to write stray comments that don't really answer the question, since these seem not to get deleted. ("Hey, as long as you found the exact YouTube video you were looking for, here's another video that's loosely related to it.")
posted by John Cohen at 7:41 AM on October 28, 2011
No, as is so often the case, the statistics alone wouldn't tell the whole story about what's really going on. For instance, if there's a decisively correct "best answer" early on, people often take advantage of the chance to write stray comments that don't really answer the question, since these seem not to get deleted. ("Hey, as long as you found the exact YouTube video you were looking for, here's another video that's loosely related to it.")
posted by John Cohen at 7:41 AM on October 28, 2011
metafilter: you'll have to massage the infodump
posted by elizardbits at 7:44 AM on October 28, 2011 [4 favorites]
posted by elizardbits at 7:44 AM on October 28, 2011 [4 favorites]
Oh my god the incident.
*puts on galoshes*
posted by shakespeherian at 7:50 AM on October 28, 2011 [1 favorite]
*puts on galoshes*
posted by shakespeherian at 7:50 AM on October 28, 2011 [1 favorite]
I thought about that, but I had no idea if the time-of-Best-Answer-marking was tracked.
Oh ho! That's a good point. It's certainly not expressed in the Infodump files; I can't recall offhand if that's actually tracked in the db itself or if we just flip a bit when it happens and leave it at that. But that does throw a wrench in the gears a bit, yes.
No, as is so often the case, the statistics alone wouldn't tell the whole story about what's really going on.
Yes, I'm well aware are of that. As an attempt to do some basic correlation tomfoolery, it would be pretty simple to perform some calculations, is the entire thrust of my earlier comment. People can be enthusiastic about fiddling with data without implied pretensions to scientific rigor or whatever; I'd rather people dig in, have fun, and talk about the limits of their methods than sit around worrying about whether its even worth it to bother. We can worry about the rigor when anything more serious than a community having fun with some self-reflecting amateur data-mining is at stake.
posted by cortex (staff) at 7:51 AM on October 28, 2011 [1 favorite]
Oh ho! That's a good point. It's certainly not expressed in the Infodump files; I can't recall offhand if that's actually tracked in the db itself or if we just flip a bit when it happens and leave it at that. But that does throw a wrench in the gears a bit, yes.
No, as is so often the case, the statistics alone wouldn't tell the whole story about what's really going on.
Yes, I'm well aware are of that. As an attempt to do some basic correlation tomfoolery, it would be pretty simple to perform some calculations, is the entire thrust of my earlier comment. People can be enthusiastic about fiddling with data without implied pretensions to scientific rigor or whatever; I'd rather people dig in, have fun, and talk about the limits of their methods than sit around worrying about whether its even worth it to bother. We can worry about the rigor when anything more serious than a community having fun with some self-reflecting amateur data-mining is at stake.
posted by cortex (staff) at 7:51 AM on October 28, 2011 [1 favorite]
John Cohen: "as is so often the case, the statistics alone wouldn't tell the whole story about what's really going on"
This is blasphemy.
posted by Plutor at 7:54 AM on October 28, 2011 [3 favorites]
This is blasphemy.
posted by Plutor at 7:54 AM on October 28, 2011 [3 favorites]
We can worry about the rigor when anything more serious than a community having fun with some self-reflecting amateur data-mining is at stake.
Thinking about identification strategies is fun, too :(
posted by a robot made out of meat at 7:56 AM on October 28, 2011
Thinking about identification strategies is fun, too :(
posted by a robot made out of meat at 7:56 AM on October 28, 2011
A quick look at the schema confirms that, yeah, no datestamp for when a best answer was marked. So that's out. It'd be interesting to try and nail down theories for guessing when a best was marked, but nothing great immediately springs to mind for me. One limited approach would be to review follow-up comments from askers looking for explicit statements that they'd gone and marked things as best—can't know for sure that those markings were closely proximal to said followup, but it'd be a reasonable assumption at least. But that doesn't cover much ground since most folks don't I think get that explicit about it.
As a jaunt to the side on this, it'd be interesting to examine specifically the content and shape of best answers vs. all other answers; setting aside the pretty big caveats about the significant per-user subjectivity of "best" assignments, there might be at least some interesting things to put together.
Another potential project, tied to that: identify potential clusters of besting behavior—users who best not at all, users who best sparingly or sporadically, users who best generously. Look at the shape and content of best answers in those different categories to see if there is some difference in kind from what different groups tend to qualify as "best".
Without comment text in the infodump those latter ideas aren't easily doable, but if someone actually wants to have a crack at playing with that idea I'd be happy to put together a one-off dump.
posted by cortex (staff) at 8:02 AM on October 28, 2011
As a jaunt to the side on this, it'd be interesting to examine specifically the content and shape of best answers vs. all other answers; setting aside the pretty big caveats about the significant per-user subjectivity of "best" assignments, there might be at least some interesting things to put together.
Another potential project, tied to that: identify potential clusters of besting behavior—users who best not at all, users who best sparingly or sporadically, users who best generously. Look at the shape and content of best answers in those different categories to see if there is some difference in kind from what different groups tend to qualify as "best".
Without comment text in the infodump those latter ideas aren't easily doable, but if someone actually wants to have a crack at playing with that idea I'd be happy to put together a one-off dump.
posted by cortex (staff) at 8:02 AM on October 28, 2011
Oh, and a scatter plot of volume-of-besting vs. volume-of-favoriting may or may not be more interesting than just an overlapping bunch of dots. Hmm.
posted by cortex (staff) at 8:04 AM on October 28, 2011
posted by cortex (staff) at 8:04 AM on October 28, 2011
We can worry about the rigor when anything more serious than a community having fun with some self-reflecting amateur data-mining is at stake.
Something more serious is at stake: if people make statistical fallacies about one topic, that'll carry over to how they think about other topics as well. So it's always important to notice when statistical fallacies are being committed, no matter what the subject matter is.
posted by John Cohen at 8:07 AM on October 28, 2011 [1 favorite]
Something more serious is at stake: if people make statistical fallacies about one topic, that'll carry over to how they think about other topics as well. So it's always important to notice when statistical fallacies are being committed, no matter what the subject matter is.
posted by John Cohen at 8:07 AM on October 28, 2011 [1 favorite]
Does tag addition have a timestamp? Could you use the "resolution" of a question to get a timestamp for identifying comments after a question is marked as resolved? (though I recognize that not everyone marks their questions thusly and the data might be skewed with the big back-tagging efforts that have gone on.)
posted by Jacob G at 8:08 AM on October 28, 2011
posted by Jacob G at 8:08 AM on October 28, 2011
What's the effect of a Keep 'em coming! comment from the asker?
posted by pracowity at 8:13 AM on October 28, 2011
posted by pracowity at 8:13 AM on October 28, 2011
This is as good a thread as any to comment on the recent person who, after getting the answer to their question, marked their own "Yes, that's it!" comment as the only best answer. One of those small things that shouldn't bug me, but does.
posted by inigo2 at 8:14 AM on October 28, 2011
posted by inigo2 at 8:14 AM on October 28, 2011
I don't use AskMe much, but when I do I tend to use 'Best Answer' to mean 'correct answer' (insofar as most of my questions are pretty objective).
posted by shakespeherian at 8:16 AM on October 28, 2011
posted by shakespeherian at 8:16 AM on October 28, 2011
Does tag addition have a timestamp?
It does indeed! And yeah, that's a neat idea for another potential sideways estimation of besting time.
though I recognize that not everyone marks their questions thusly and the data might be skewed with the big back-tagging efforts that have gone on
True, and the tag data doesn't identify the tagger explicitly. That said, you could probably do a decent job of separating out most of the back-tagged "resolved" tags (if any, even; that may not have been something people were generally doing during that project) from the poster-added ones by looking at relative time of addition of the tag—resolutions marked on long-closed questions would likely be back-tagging activity rather than way-after-the-fact followup tagging by the asker, for example.
That gets slightly more complicated because there's also some admin-side backtagging we had done when we saw resolution-noting followups by the poster in old askmes. (I don't know if Matt or Jess do thisanymore, I don't at least, since we started sending reminder mails automatically to suggest followups and "resolved" markings).
posted by cortex (staff) at 8:16 AM on October 28, 2011
It does indeed! And yeah, that's a neat idea for another potential sideways estimation of besting time.
though I recognize that not everyone marks their questions thusly and the data might be skewed with the big back-tagging efforts that have gone on
True, and the tag data doesn't identify the tagger explicitly. That said, you could probably do a decent job of separating out most of the back-tagged "resolved" tags (if any, even; that may not have been something people were generally doing during that project) from the poster-added ones by looking at relative time of addition of the tag—resolutions marked on long-closed questions would likely be back-tagging activity rather than way-after-the-fact followup tagging by the asker, for example.
That gets slightly more complicated because there's also some admin-side backtagging we had done when we saw resolution-noting followups by the poster in old askmes. (I don't know if Matt or Jess do thisanymore, I don't at least, since we started sending reminder mails automatically to suggest followups and "resolved" markings).
posted by cortex (staff) at 8:16 AM on October 28, 2011
This is as good a thread as any to comment on the recent person who, after getting the answer to their question, marked their own "Yes, that's it!" comment as the only best answer. One of those small things that shouldn't bug me, but does.
I'm more annoyed that I never got my bonus points.
posted by theodolite at 8:33 AM on October 28, 2011
I'm more annoyed that I never got my bonus points.
posted by theodolite at 8:33 AM on October 28, 2011
On my most recent askme, I decided not to mark any of the answers as "best" because I figured doing so would limit anyone adding more answers. But in the end, I suspect the number of people who could answer it was totally limited anyway, and the answers dried up anyway.
posted by crunchland at 9:07 AM on October 28, 2011
posted by crunchland at 9:07 AM on October 28, 2011
If anyone is actually going to work on this, don't forget that Anon questions are going to seriously skew the data since they average more answers and have no "best answers" marked.
posted by auto-correct at 9:44 AM on October 28, 2011
posted by auto-correct at 9:44 AM on October 28, 2011
If anyone is actually going to work on this, don't forget that Anon questions are going to seriously skew the data since they average more answers and have no "best answers" marked.
Except that sometimes they do. I even have a best answer to an anonymous question myself.
But yes, weird outliers like that aside, pretty much any datawankery exercise involving AskMe is likely to have to do something different with anonymous questions.
posted by FishBike at 11:05 AM on October 28, 2011
Except that sometimes they do. I even have a best answer to an anonymous question myself.
But yes, weird outliers like that aside, pretty much any datawankery exercise involving AskMe is likely to have to do something different with anonymous questions.
posted by FishBike at 11:05 AM on October 28, 2011
Yeah, to be more explicit, some questions that had best answers assigned were later anonymized.
posted by cortex (staff) at 11:06 AM on October 28, 2011
posted by cortex (staff) at 11:06 AM on October 28, 2011
Based on my experience, which is largely in the relationship/babies/etiquette/recovery categories, an early Best Answer will indeed stymie further answers. My totally unsubstantiated instinct is that best practice is to let 24 hours go by before handing out little green checkmarks for these type of opinion rather than One True Answer questions.
posted by DarlingBri at 5:06 PM on October 28, 2011
posted by DarlingBri at 5:06 PM on October 28, 2011
I'm with you, DarlingBri. I don't see a lot of upside to marking best answers early, unless it's some very specific question ("What movie am I thinking of?"). Heck, I might wait to mark even in those cases just on the off chance someone drops in to add something else related but interesting (as John Cohen alludes to above).
posted by Conrad Cornelius o'Donald o'Dell at 6:24 PM on October 28, 2011
posted by Conrad Cornelius o'Donald o'Dell at 6:24 PM on October 28, 2011
You are not logged in, either login or create an account to post comments
posted by cortex (staff) at 7:26 AM on October 28, 2011