17

Internally, we are discussing making it easier for us to tell if post activities came via the mobile version of the site. We already have a column in the PostHistory table called ApplicationId that indicates whether the activity (mostly creating and editing) on a post was from the Android or iOS app. We use this data to track things like whether questions posted with an app are shorter than average.* This seems like useful data, so I'm proposing adding this column to the public data.

In order to make the data user-friendly, I'd also like to create a ApplicationTypes table with these columns:

  1. Id int = ApplicationId from PostHistory
  2. Name varchar(10) = 'ios', 'android', 'mobile', 'full'
  3. Description varchar(300) = whatever detailed description might be helpful

One potential concern is that we would be revealing some information about our users that was not public in the past. You could, for instance, identify the 17 people who used Apple devices on Android Enthusiasts and the 101 Android users who have infiltrated Ask Different. You could also determine which of those people did it for the hat. But I think the value of having this information outweighs the potential cost.


* The answer on SO:

source  avg_body N       
------- -------- ------- 
web         1319 8718569 
android      723    2295    
ios          693     898   

  • 2
  • 2
    I was almost going to suggest this a while back, but for some user's couldn't it also give you a fair idea when they were away from home from the post history timestamp? – PeterJ Jan 23 '15 at 0:54
  • 7
    What uh...what is the value of having this data (besides auditing)? – Tim Stone Jan 23 '15 at 1:10
  • 1
    Can you explain why you think having this information outweighs the potential cost of being able to identify how a user is interacting with the network and then sharing that identifying information with everyone? I am less concerned that Stack Exchange can see that information than I am with that information being part of the data dump. – Andy Jan 23 '15 at 2:21
  • well @Andy, in Winterbash everyone knew that you posted using a Android|iOS|mobile device. There's actually more information about you right in the network (SE), like at what hours/days you interact with the site, what kind of topics interest you, etc. This peculiar information (interacting from mobile) don't help to narrow down your identity more than what it already is. – Braiam Jan 23 '15 at 2:36
  • 2
    @Braiam "You could, for instance, identify the 17 people who used Apple devices on Android Enthusiasts and the 101 Android users who have infiltrated Ask Different.". That is pretty specific. Much more so than being a user that frequents a specific tag or posts during work hours in a certain time zone. – Andy Jan 23 '15 at 2:41
  • 3
    @TimStone: Well, for one thing, it would give people access to the data to answer questions like: Should new users be asking questions on Stack Overflow from their mobile devices? – Jon Ericson Jan 23 '15 at 4:30
  • 1
    Ah, that's good. Do you have any other examples? I figured there were indeed valid reasons, but I couldn't think of them and needed to have an idea what the usage scenarios would be to see if I could propose some sort of middle ground that would make useful data available without being too pry-y. – Tim Stone Jan 23 '15 at 4:33
  • @Jon wouldn't the data from the public dump be too heavily skewed by deleted posts assuming there is a marked quality difference? – PeterJ Jan 23 '15 at 8:48
  • @PeterJ: Yes, deleted content skews the data dramatically. But that's a larger problem to solve than this proposal. – Jon Ericson Jan 23 '15 at 16:15
  • @TimStone: Not necessarily a permanent thing, but I can't give out the queries for the mobile-related hats since they don't work on the public data. Essentially, we have a "default public" approach to information: unless there's some good reason to keep it private, we want to share what we know. User privacy is one reason for not sharing information. Hence the question. ;-) – Jon Ericson Jan 23 '15 at 16:21
5

Absolutely.

Counterargument: there are only use-cases for this, and there are no downsides to doing this. User-driven tests on typos, post quality, followup frequency...

All of these things are useful.


The concern for releasing these is understandable, but I doubt there are any serious use-cases for abuse of this information. If somebody decides to harass a user over a hat, then that's a) silly, and b) a moderator can handle it quickly.

In reality, I don't think this information is abusable in a practical or meaningful way, and the value it could provide is broad enough that we can't even completely foresee its uses. We already know what some of them are, though, and that alone, I think, justifies their presence in a dump.

3

Whenever you're thinking of exposing personal data, the question to ask is not “is there a good reason to exclude it?” but “is there a good reason to include it?”.

The type of computer used by a poster is personal data. What's next? Expose the User-Agent string sent by the poster's browser, so that you can treat people differently depending on whether they're using Firefox or Opera? (Hey, fellow Unix & Linux users, let's gang up on people using Internet Explorer!)

I can't think of a good reason to expose this information to the public. I can definitely think of reasons to collect this information for statistical purposes (do we need to bother with X app at all? is there a correlation between usage of X app and Y behavior that could indicate a usability problem?), so it's reasonable to record it. But keep it internally unless you have a good reason to expose it.

You must log in to answer this question.

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