Number of flags per deleted post?

April 4, 2013 12:31 PM

I was wondering if the currently available piles of data support certain metrics...

1. Median number of flags per post
2. Median number of flags per deleted post
3. Median number of deleted comments per post
4. Median number of flagged comments per post
5. What percentage of flagged comments get deleted.

I am in no way interested in knowing the specific posts in question; just the aggregates.

I apologize if this is documented somewhere. I didn't know where to look.

posted by DWRoelands to MetaFilter-Related at 12:31 PM (20 comments total)

There may indeed be some data about this, but my hunch is that the mods' feeling about that data is "there is no way in all the nine hells we want to make that information public".
posted by EmpressCallipygos at 12:39 PM on April 4, 2013 [3 favorites]

Flag data isn't something we're going to make public even in the aggregate. The deletion data is sort of in the info dump via gaps in comment/post numbers but isn't really right up top. Maybe someone has put together a thing that can summarize it.
posted by jessamyn (staff) at 12:44 PM on April 4, 2013 [1 favorite]

The flags aren't covered in the Infodump, and they won't be; if they were public, the mods would be criticized for deleting ill-advised threads with a relatively low flag count, and also for not deleting threads with a relatively high flag count. It's a headache with much upside.
posted by Pronoiac at 12:46 PM on April 4, 2013 [4 favorites]

Deleted comments are tough as all the infodump tells us is that a comment was deleted, not what post it was associated with.

There are two ways to deal with this, both of which assume that a fast moving thread is the source of deletions:

1. Assume the previous comment was made in the same thread that the deleted comment was in and/or
2. Assume the comment was made in the thread that had the most comments over the previous N minutes.

Neither of these are very satisfactory unfortunately. Definitely open to ideas for different heuristics...
posted by Tell Me No Lies at 12:50 PM on April 4, 2013

Yeah, a little more detail since this is something we've talked about a lot in the past (and I've thought about in particular):

Flag data's always been behind the scenes and we don't really have any upside for changing that since it's not something we use to automatically trigger actions. Things that have been flagged more than the median can and do stick around; things with less do get deleted; judgement calls end up driving most of this stuff, and it'd be hard to even say much that wasn't just sort of massively overbroad about flag-vs-deletion correlations without digging down into the fine row-by-row details of the raw flag data. It's kind of go datadump or go home, really.

And the biggest problem with that is that flags are attached to user contributions, and any detailed look at what's being flagged is necessarily the publicizing of data about who has been flagged. Which is something I think a lot of people would not be comfortable with, for themselves or for others, even though I know there are other folks who have said they'd be fine with or even interested in that sort of info.

So, we're pretty much staying totally clear of putting flag data out there.

Deletion rates over time are derivable from the Infodump comment data—we don't include actual deleted comments as such in those tables but you can check for gaps in the comment-id column to figure out how much has gone. Doesn't give you super fine data at a per-thread level, but providing that level of detail would, per above with flags, also mean providing a dump of info about who is getting their comments deleted, which again feels like not a great thing to do.
posted by cortex (staff) at 12:50 PM on April 4, 2013

I don't think we can get much info on deleted comments from the Infodump. We can glean the times, but nothing about which threads they were in. So we could get overall stats like, say, average (mean) comments deleted per day, or percentage of comments deleted. We can't get median number of deletions per thread from the Infodump, but I'd guess it's zero, based on a vaguely-recalled low-to-mid-single-digit deletion rate.
posted by Pronoiac at 12:57 PM on April 4, 2013

Yeah, you could throw a (granted, deeply speculative) model at median comment deletions per thread question just by looking at the deletions per day rate, threads per day, and some sort of poisson distribution for the deletions. If I get inspired on some rainy day I might try looking at that in a little more detail since I don't really have a problem with talking about deletion data in aggregate.
posted by cortex (staff) at 1:01 PM on April 4, 2013

I do like the idea of deleted posts/comments as the dark matter/energy of the MetaFilter universe — only detectable through their influence on the things we can see.
posted by benito.strauss at 1:11 PM on April 4, 2013

Information on deleted posts and comments is justifiably obscure. "Sometimes dead is better."
posted by justsomebodythatyouusedtoknow at 1:18 PM on April 4, 2013

Deleted posts are easy, actually. I seem to recall from the last time I ran numbers that a surprisingly large number of posts (close to 15%?) get deleted.

I just realized there there is another likely metric for determining which thread a comment was deleted from, which is the longest thread in that general time period. Not only would a long thread be statistically more likely to be the source of the comment, but very long threads often are very fractious threads.
posted by Tell Me No Lies at 1:58 PM on April 4, 2013

I want to know what the most favorited deleted comment ever was.
posted by nathancaswell at 3:47 PM on April 4, 2013 [3 favorites]

Any chance "flagged as fantastic" could ever be made public or would it be impossible to separate it out? I'd love, for example, to see the correlation between that and number of favorites.
posted by Kid Charlemagne at 3:50 PM on April 4, 2013 [1 favorite]

How about this question - what is the greatest number of flags a post ever received? No need to identify the post, just looking for a number.
posted by Chrysostom at 3:52 PM on April 4, 2013

I think we should start a grass roots movement to make the portabello mushroom thread the most flagged thread of all time, everyone get in there and flag it as fantastic, which it is.
posted by nathancaswell at 3:55 PM on April 4, 2013 [1 favorite]

How about this question - what is the greatest number of flags a post ever received? No need to identify the post, just looking for a number.

I'd love to know the answer to that, too, but "more than one" is probably the closest we'll get to a real answer.
posted by double block and bleed at 5:11 PM on April 4, 2013

Well, I recognize the concerns about flag data that cortex relates above, but it seems to me that saying simply, "The most flagged post ever had 1,378 flags" doesn't tell us anything that could be used in a negative way against anyone. It's merely trivia.
posted by Chrysostom at 6:40 PM on April 4, 2013

Isn't a flag is a just a way for users to request the mods to take a look at something that may be problematic? They aren't meant to be "please delete this" requests. So, a flag count doesn't have much meaning beyond that. Sometimes mods get to problematic things before they have attracted many/any flags and sometimes they are busy with putting out a different fire and there might be tonnes of flags.
posted by vidur at 8:09 PM on April 4, 2013

Isn't a flag is a just a way for users to request the mods to take a look at something that may be problematic?

Um, yes? I think you're looking for this to actually MEAN something, but I'm just curious. It doesn't necessarily tell you anything important any more than knowing what the largest county in Pennsylvania is.

It's Lycoming County.
posted by Chrysostom at 8:50 PM on April 4, 2013

Sorry for the confusion. My comment wasn't aimed at you or anyone else in particular. Just generally on topic.
posted by vidur at 11:45 PM on April 4, 2013

I flag a ton of stuff I don't necessarily want to see deleted. My flagging threshold is "could a reasonable person have a problem with this?" If I strongly feel something should be deleted, I generally use the contact form and argue my case.
posted by ryanrs at 2:17 AM on April 5, 2013