9

It seems like it's a sure bet that if you see:

n min Soandso 1

next to a question and Soandso is an unregistered user, the post Soandso made is either:

  • spam,
  • gibberish,
  • a reply to an existing user, or
  • a "me too!" or follow-up question

Because of this, I'm interested to find out how often one-rep unregistered users provide actual answers. Because nearly all non-answers are deleted, I'm having trouble figuring out a way to compare the ratio of useful content from one-rep unregistered users to junk content: it seems like something SE would have to provide to us. Nevertheless, anyone (coughKevinMontrosecough) interested in taking a stab at data analysis for this?

Assuming my hunch is correct and one rep unregistered users are exceedingly unlikely to contribute meaningfully to the site in the form of answers, is there something we can do about it?

10
  • Data on this would be very interesting, however the amount of bad posts we're getting really is very low, so taking action wouldn't be necessary IMO. – fredley Jul 5 '12 at 10:20
  • @QAdley Not to imply that it's inaccurate, but what's the basis for saying the amount of bad posts we get is very low? And very low compared to what? – user3389 Jul 5 '12 at 10:21
  • @MarkTrapp What's to say the amount of bad posts is high and warrants attention? – ediblecode Jul 5 '12 at 10:24
  • 7
    @danRhul The purpose of this question is to get the necessary data to make that determination. – user3389 Jul 5 '12 at 10:24
  • @MarkTrapp I mean low enough to handle. Posts that are bad are very quickly dealt with, our mod/active user capacity is much higher than the level of bad posts we're getting. – fredley Jul 5 '12 at 10:25
  • Related: meta.stackexchange.com/questions/133117/… – badp Jul 5 '12 at 13:17
  • 5
    +1 not because I agree with the proposed solution, which feels draconian, but because I'm curious about the scope of the problem. It does feel like we have a very out of proportion problem (relative to the network) with answers that are purely noise, trolling or vandalism. I'd love to see stats to back that up or disabuse me of that notion. – LessPop_MoreFizz Jul 5 '12 at 14:28
  • @LessPop_MoreFizz I removed the possible solution: should've known it'd cloud the question. I'm in the same boat as you, and it seems like it might be confirmation bias on my part. – user3389 Jul 5 '12 at 17:16
  • I disagree with 'the post Soandso made is either:, etc', I often plow through the front pages clicking on the activity of all of the 1-10 rep users I see to review them, and about half the time, yes, it's already been deleted, but there have actually been quite a number of decent, or even good posts from new users over the last little while. But yes, there is a lot of garbage as well. – GnomeSlice Jul 5 '12 at 17:22
  • @GnomeSlicE The reason why I'm looking for data analysis instead of agree/disagree answers is that everyone's experience at any specific time is no doubt going to be different. For example, right now, there are 8 posts by 1-rep unregistered users on the front page, and all but 2 are already deleted, with one more on its way to deletion. – user3389 Jul 5 '12 at 18:19
18

We've all Very many of us have been 1 rep unregistered users at some point in our "Stack Exchange" career. One of the main points about the engine is to make it as easy as possible for anybody with knowledge to share it.

Just because I don't want to name big names, Agent86 for example only registered after posting his fifth answer on the site. Today he's our second top user ever and a community moderator. Who knows what compulsory registration would've changed?

Sure, for every Agent86 there's a thousand will's answering people with poorly-formatted trollfaces. But our is a site for experts; our is a site that thrives on that one in a thousand people who actually care to be experts in their field. We can't afford losing any of them a priori. The barrier to entry must be as low as possible - that's how the engine is designed.

Stack Overflow, which has much much worse big city problems than we do, blocks users from asking but not answering questions without registering. That's because questions are worth a dime a dozen, but it's answers are the whole point of a Q&A site. It doesn't make sense to make somebody choose if they want to login with Google or Twitter or Facebook or Yahoo or whatever (what's the difference?) before they can start writing an answer.

I really don't think it would be wise to require people to ask questions before they can answer them.

4
  • 1
    Not to ask before answering, merely to register before answering. A process which, I'll note, is pathetically simple thanks to OpenID. – LessPop_MoreFizz Jul 5 '12 at 11:38
  • 4
    @LessPop_MoreFizz If there's one historical complaint about the site, it's that OpenID is "difficult" to understand. :/ at any rate asking users to register before answering goes against the general philosophy of Stack Exchange in general. – badp Jul 5 '12 at 13:09
  • 5
    I was never a 1 rep unregistered user, I registered right away, which indicated at least some level of commitment to the site before contributing. – Resorath Jul 5 '12 at 16:04
  • 1
    I removed the potential solution as that's not what I'm interested in focusing on here, but whether or not there's an actual thing with unregistered users or not. But your assertion that everyone was a one-rep unregistered user at some point is easily disproved. – user3389 Jul 5 '12 at 17:20
6

A detailed statistical analysis requires more tools than my pay grade allows, so we might end up waiting a bit on that. I thought about some ways to potentially find answers by low-rep users that ended up being quality, but that's kind of a problematic way to slice data and get anything useful out of it. Perhaps one of our SE overlords will grace us with some detailed data, which would probably be definitive.

I did, however, ask around with the mods over on SO to see what their volume is like and how it compares to ours. I found that:

SO started requiring stricter controls to post questions when they reached a daily question volume of ~3-4k. ("Registration required" was their solution) Right now our answer volume hovers between 100-200 per day, which is less than a tenth of what was considered 'critical mass' - requiring a policy change - on SO. (We peak to 4-500 during the early phase of contests, based on my API analysis)

SO mods delete around 10x the amount of content (per mod) that we do. They're electing a few more mods, so presumably they're at their current "moderator event capacity." That's not to say that everything that gets deleted could be prevented by requiring registration, just that it seems like we've got considerable moderator event capacity remaining by SO standards.

From these somewhat rough numbers, I don't really think we have a serious problem with stuff getting posted that requires deletion, unregistered users or not.

In thinking about this question, my mind jumped to something that I think you also thought about (judging by this comment). It's a little bit cart-before-the-horse, but I think it's still relevant given that gathering definitive metrics is likely to be a time consuming process.

What's the tipping-point ratio of bad to good content, that would make it "worth it" to raise the bar for people to post answers?

I think of this somewhat like a spam filter (or other machine learning algorithm) - increasing the bar means probably causing some "false positives" - ie, your credit card bill being marked as spam. How many v14gr4 ads is it worth looking at to ensure that an important message gets through?

As one of the primary filters of such content on the site, I can say I'm honestly not bothered by the volume. The type of clear "not an answer" type stuff you've noted is pretty easy to deal with. Usually there's an auto "low quality" flag or one or two community members have flagged it, and it takes me 2 or 3 clicks to banish it to the great Recycle Bin in the sky. Whether others feel like flagging said content is a bothersome chore is beyond my knowledge, but hopefully they find it just as easy to flag as I do when I handle the flag.

Based just on my personal opinion, I'd rather have fewer barriers and deal with a thousand "not an answer" flags/deletions a month (which is far greater number for me than it is currently) than potentially miss out on even one really worthy answer.

2

Yes, they can. I just scanned the list of recent answers from new users, and found

How do you beat Belial in Diablo 3 as a Monk?

which is a perfectly reasonable answer. Now I'm sure that many of them are junk. But I see lots of new users that aren't registered, so obviously registering is more work for a lot of them than they want to do, and I don't think the junk is so overwhelming that we should be willing to sacrifice good answers (and probably users) in order to prevent it.

5
  • 1
    Just for the sake of adding to this, here was an excellent answer by an unregistered user – Sterno Jul 5 '12 at 16:11
  • 1
    Interesting anecdotal evidence and it at least demonstrates that posts by unregistered users are not universally bad, but the reason I'm asking for hard data is to determine where our line is: if we only get one good or acceptable answer in a hundred, is that okay? What about one in a thousand? One in several thousand? If good posts by unregistered users are really rare and the ultimate solution is to require users to register, is that a large enough hurdle such that the rare gem of a post won't be made at all (instead of just taking a few minutes longer as that user registers)? – user3389 Jul 5 '12 at 17:24
  • This is a post from a 1-rep registered user. They registered 2 days ago and posted this answer 19 hours ago. – fredley Jul 6 '12 at 8:50
  • @QAdley I'm not sure where you're looking, but the user page for both the answer I posted and the one Sterno posted says "Unregistered". – bwarner Jul 6 '12 at 12:00
  • @bwarner My mistake, I saw 'member for' and assumed that meant registered. – fredley Jul 6 '12 at 12:01

You must log in to answer this question.