In general LW admins do not look at drafts, except when a user has specifically asked for help debugging something. I indeed care a lot about people feeling like they can write drafts without an admin sneaking a peak.
The exceptions under discussion are things like “a new user’s first post or comment looks very confused/crackpot-ish, to the point where we might consider banning the user from the site. The user has some other drafts. (I think a central case here is a new user shows up with a crackpot-y looking Theory of Everything. The first post that they’ve posted publicly looks sort of borderline crackpot-y and we’re not sure what call to make. A thing we’ve done sometimes is do a quick skim of their other drafts to see if they’re going in a direction that looks more reassuring or “yeah this person is kinda crazy and we don’t want them around.”)
I think the new auto-rate-limits somewhat relax the need for this (I feel a bit more confident that crackpots will get downvoted, and then automatically rate limited, instead of something the admins have to monitor and manage). I think I’d have defended the need to have this tool in the past, but it might be sufficiently unnecessary now that we should remove it from our common mod toolset.
...
I also want to emphasize since @Dagon brought it up: we never look at DMs. We do have a flag for “a new user has sent a lot of DMs without posting any content”, but the thing we do there is send the user a message saying approximately “hey, we have observed this metadata, we haven’t read your DMs, but just want to encourage you to be careful about spamming people in DMs”. In cases where we suspect someone is doing flagrant DM spam we might disable their ability to send future DMs until they’ve persuaded us they’re a reasonable real person, but still not actually read the DM.
I apologize if I implied that the mods were routinely looking at private data without reason—I do, in fact, trust your intentions very deeply, and I’m sad when my skepticism about the ability to predict future value bleeds over into making your jobs harder.
I wonder if the missing feature might be a status for “post approval required”—if someone triggers your “probably a crackpot” intuition, rather than the only options being “ban” or “normal access” have a “watchlist” option, where posts and comments have a 60-minute delay before becoming visible (in addition to rate limiting). The only trustworthy evidence about future posts is the posts themselves—drafts or deleted things only show that they have NOT decided to post that.
Note that I don’t know how big a problem this is. I think that’s a great credit to the mods—you’re removing the truly bad before I notice it, and leaving some not-great-but-not-crackpot, which I think is about right. This makes it very hard for me to be confident in any opinions about whether you’re putting too much work into prior-censorship or not.
I want to clarify the draft thing:
In general LW admins do not look at drafts, except when a user has specifically asked for help debugging something. I indeed care a lot about people feeling like they can write drafts without an admin sneaking a peak.
The exceptions under discussion are things like “a new user’s first post or comment looks very confused/crackpot-ish, to the point where we might consider banning the user from the site. The user has some other drafts. (I think a central case here is a new user shows up with a crackpot-y looking Theory of Everything. The first post that they’ve posted publicly looks sort of borderline crackpot-y and we’re not sure what call to make. A thing we’ve done sometimes is do a quick skim of their other drafts to see if they’re going in a direction that looks more reassuring or “yeah this person is kinda crazy and we don’t want them around.”)
I think the new auto-rate-limits somewhat relax the need for this (I feel a bit more confident that crackpots will get downvoted, and then automatically rate limited, instead of something the admins have to monitor and manage). I think I’d have defended the need to have this tool in the past, but it might be sufficiently unnecessary now that we should remove it from our common mod toolset.
...
I also want to emphasize since @Dagon brought it up: we never look at DMs. We do have a flag for “a new user has sent a lot of DMs without posting any content”, but the thing we do there is send the user a message saying approximately “hey, we have observed this metadata, we haven’t read your DMs, but just want to encourage you to be careful about spamming people in DMs”. In cases where we suspect someone is doing flagrant DM spam we might disable their ability to send future DMs until they’ve persuaded us they’re a reasonable real person, but still not actually read the DM.
I apologize if I implied that the mods were routinely looking at private data without reason—I do, in fact, trust your intentions very deeply, and I’m sad when my skepticism about the ability to predict future value bleeds over into making your jobs harder.
I wonder if the missing feature might be a status for “post approval required”—if someone triggers your “probably a crackpot” intuition, rather than the only options being “ban” or “normal access” have a “watchlist” option, where posts and comments have a 60-minute delay before becoming visible (in addition to rate limiting). The only trustworthy evidence about future posts is the posts themselves—drafts or deleted things only show that they have NOT decided to post that.
Note that I don’t know how big a problem this is. I think that’s a great credit to the mods—you’re removing the truly bad before I notice it, and leaving some not-great-but-not-crackpot, which I think is about right. This makes it very hard for me to be confident in any opinions about whether you’re putting too much work into prior-censorship or not.