Stop using real accounts in suggested edit audits

Question

Well, thanks for making me harass a user over attempting to vandalize a post!

Got a flag on a question notifying us that a new, low-rep user submitted a bogus edit suggestion. Okay, so I went and took a look at it.

Well, that certainly looks spammy.

I then checked out the user's account, which looked normal, having no signs of spam whiff. Since this person is new, I contacted the user asking what happened with the edit.

me:

We just got a flag that you attempted to deface a question with an edit

[snip details]

Can you explain?

user:

No i didn't do anything

me:

This isn't a horrible nightmare of a mess or anything, its just odd. Your account appears to be that of a real person, but this kind of behavior is indicative of a fake account.

Can you please just explain what happened?

etc etc until the point where Brad Larson figured out that this was probably a Suggested Edit audit, which was then confirmed to be the case, and I sincerely apologized to the user for harassing them over bullshit.

How is using actual accounts of random users acceptable in this situation? Why are these audits recorded in the post history just as if they are real?

This is kinda bad, guys. You gotta stop using unsuspecting users as vandals and stop recording these incidents in the post history.

Answer 1

Woops, that's not good. I suppose we always suspected this would come back to bite us, but perhaps not so spectacularly.

Showing fake users on audits is non-ideal, because it creates an easy way to quickly detect audits: Does the display name look fake? Even if not, does clicking the display name bring me to a real user profile?

We just pushed a change (that will go live shortly) that will hopefully alleviate the problem:

Once the audit is passed/failed, then we'll always show the edit as having been created by the Community User.

But while a suggested edit audit is still reviewable, the behavior is unchanged – an unsuspecting user is randomly selected and displayed as the proposer of the edit.

Answer 2

I couldn't agree more. This is impersonation in an explicitly negative scenario, which amounts to libel if you look at it too closely. I thought it was kinda creepy when I first figured it out (which took a while), but got used to it. You've reminded me that I shouldn't have done that.

On a practical note, if you're a user who accepts edits blindly (which the audit is supposed to weed out), it seems unlikely that you're gonna spend time investigating the editor. Thus it seems that there is little benefit in co-opting an authentic account, other than the dev time required to present a fake one that passes even just cursory examination.

Answer 3

Wait, merely having an account on a site that uses audits means I could have actions falsely attributed to me? That's bad, and waiting until after the audit is complete to reveal the truth -- to anybody who's still looking at that point -- is insufficient. People sometimes look at the profile during the review, not just after it's resolved. You're setting people up for people to think "oh, Skeet -- that was that spammer I saw in the review queue" when they next encounter the user. You can update the review event but you can't update people's memories.

The audit code impersonates users. Our terms of service don't allow people to do that, yet our own code does it? That's not cool.

Surely we have, or can create, some real accounts to use in audits. They don't need to show site activity; users with no posts sometimes do suggest edits, after all. If you're worried that people will learn to recognize the fake users over time, change the names for each audit -- most people won't click through. If you're concerned that somebody will click through, notice the name difference, and say "aha, this was an audit", maybe somebody who does that level of investigation isn't the type that you're most worried about catching with audits in the first place.

Answer 4

I agree with the other answerers so far that this behavior is disconcerting. A user's identity should be sacrosanct. There's probably no way to know whether this has caused any real problems (and if so, how severe they were), but it's a matter of principle.

I have a somewhat different idea for a solution: while an edit suggestion is still pending, whether a real one or an audit, don't show the edit suggestor's name or other information at all.

I want to be explicit here that I am not proposing completely removing credit for suggested edits. After a given edit suggestion is judged, the relevant post edit history and suggestor profile should be updated as normal.

From Emmett's answer, the concern seems to be that people may try to use edit suggestor information as a kludgey way to determine whether edits are audits. I submit that such behavior is an anti-pattern. Proper reviewers should judge edit suggestions based on their content, not the identity of their authors. Regardless, my proposal eliminates the problem; you can't try to analyze profile information that isn't there in the first place. And removing this from all profiles rather than just audits prevents reviewers from having any hint at all.

The main complaint I anticipate to my idea is that people won't want to lose visibility on real suggestors' information. I don't think there's a big cost here, though. No information is being destroyed; just hidden for the short period when suggestions are under review. And showing names during review leads mostly to negative behaviors, not positive ones (e.g. introducing bias, and distracting from content). The most legitimate use case I can think of is that higher-rep users might be more trustworthy editors than anonymous or low-rep users; however, once a user reaches 2000 rep (lower on betas) their edits stop being suggestions, so even this is of limited utility.