lol, well there you have it, on a site with thousands of users and thousands of voting buttons. seems undemocratic. seems sometimes all the populist appearances/ claims of SE/ mod service to users amount to just a sham... guess we all now know whos in charge around here

> Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics and astronomy. It's built and run by you as part of the Stack Exchange network of Q&A sites.

> At the highest levels, you'll have access to special moderation tools. You'll be able to work alongside our community moderators to keep the site focused and helpful.

↑ "community" lol!

There is no democratic process involved in issuing suspensions. Nobody is claiming otherwise. Please drop it.

← lol, ok, my mistake! o_O

@DavidZ You explicitly mentioned that judging the size of the flag history without context isn't something the Physics mods do. I proposed the most liberal alternative - that they do the full factor analysis.

Well, what I explicitly said was we don't suspend based solely on the size of the flag history.

I'm not sure I understand the point you were making there, where you talked about that "full factor analysis".

Sure, and the antonymous claim would be that you'd do the factor analysis, taking into all the multivariable relations which determine the qualities of a user that garners a suspension. My point was that even with that liberal hypothesis, there's still a systematic statistical bias on the flag variable, which you are denying the Physics moderation has.

Well, not a point, but just a demonstration of how there can be such a bias.

I'm not sure what's unclear about my argument.

Honestly, most of it is still unclear to me. I don't follow the logic.

For one thing, what do you mean by "systematic statistical bias"?

Or more broadly, it seems like you have some preconceived notion of the role which flag history should play in moderation decisions, and you're claiming that it plays more of a role than that. But I'm not sure what your preconceived notion is.

It's kind of hard to talk about anything statistical in a useful way without knowing that.

Ok, here's a piecemeal approach. The moderation policy or system, by definition, is a general algorithm to determine if certain users are suspension-worthy. Consider your model, where there's an unknown real variable X associated to each individual which determines how problematic they are. There are many variables Y_i with which X has nonzero correlations with, the flag variable F is one. So you can write X = a_1 F + a_2 Y_2 + ... + a_n Y_n + (error term).

That's factor analysis. Intuitively, this is what a moderator would do to determine X, which is the best possible approach - because it doesn't simply focus on one single factor that has high correlation with X, but all such variables.

My claim is that any intuitive determination of the coefficient of F, a_1, produces drastic biases on the right hand side.

OK. I understand what you've laid out in that last block of messages, but I don't see how your claim is supported.

(We should also not take for granted that a linear model is accurate, for whatever that's worth.)

OK, glad this is making sense. The claim can very well be wrong, but it seems to me that how problematic users get moderation attention in the chat is fundamentally based on the flags he'd acquire. Namely, to even start computing X a moderator would go through the user list and catalog the users via the number of flags. Is this incorrect?

Off the cuff, I believe that is incorrect.

Mmk.

I mean, looking at the user list sorted by flag count is certainly not something we actually do in the normal course of moderation. I'm not sure I've ever done it, actually.

It may or may not be similar in effect to the process we actually do use, and I'm inclined to say it's not, but I can't be sure.

Yeah I wasn't actually describing what you do, it was more of an internal model of mine for what you'd do. So take it as an abstraction.

Fair enough, I figured so.

I'm just saying, I kind of doubt whether it's even a useful abstraction.

Gotcha.

Of course, flags are one of the primary ways that our attention gets drawn to certain messages (and indirectly to the users who post those messages), so it definitely is the case that users which are more frequently flagged tend to get more attention from the mods. That means that your abstraction isn't totally useless. I just think it's quite a bit less realistic than you may have been thinking.

I see. Yeah it's totally possible I'm over-invested in this interpretation.

Yeah, it does kind of seem that way. But I do understand that it's hard to get good information about how the moderation process really works.

What happens in practice is that we have a private moderator chat room with an ongoing discussion, of sorts, about noteworthy things that happen on the site. For example, when a message in the h bar gets flagged and the flag seems valid or potentially valid, someone would leave a note about the message in there, and it typically kicks off a discussion about whether the message is actually inappropriate.

Right, I'm aware of that.

Sometimes everyone agrees the message was fine and we don't give it a second thought.

In cases where the message is particularly bad, or we notice a pattern of messages posted by a particular user which have previously been discussed in the room, the discussion might move on to whether it's appropriate to suspend that user.

We typically search through the private room transcript, as well as possibly the public room transcript, and try to evaluate the overall contribution history of the user when deciding on possible suspensions.

When that discussion is happening, the pure flag count doesn't play much of a role, if any. We're more focused on content and context.

The history of manually imposed suspensions (if any) does also enter the discussion.

I know I'm being a little vague but I can't really think of a good way to give any specific detail without revealing things I shouldn't.

I understand.

Cool, I appreciate that.

I have heard the narrative that too many flags coming from a certain room gives a bad impression of that room to the community and the authority. So it doesn't seem to be a bad guess that certain users with more auto-suspension history would implicitly be looked upon with a frown.

Yeah, I've heard that as well. That's mostly an issue of how moderators from other sites perceive, say, the h bar. That doesn't factor into our decisions about chat suspensions, except to the slight extent that it helps us calibrate our bar for what kind of behavior deserves a suspension.

I see.

Of course we would also probably escalate to the SE team to investigate why they're getting invalidly flagged so much.

If we get the sense that the SE team is concerned about the level of flagging (or other persistent problems) taking place in the h bar, we'll discuss among ourselves, and also with them, what would be needed to deal with that. That's another case in which we may have to recalibrate our bar for what qualifies for suspension.

That has been happening recently, to some extent. It is correlated with the SE blog post and the drive to be more welcoming across the network.

I noticed that, yes.

Right now my impression is that we're not in a position where the h bar is in any imminent danger of being shut down, but at the same time there are some concerns from SE that we have to take seriously.

Well, on that issue, my impression is that a fraction of hbar's users thinks that conversation policing is a real problem. I do think the moderation is taking fairly extreme measures to ensure welcoming-ness (that's a word apparently?) across the network.

Which is why I think hbar will eventually be either fairly inactive or simply be shut down.

Well, there's a sense in which the measures we're taking are extreme, but I think that's mostly because less extreme measures have been tried and didn't achieve the desired effect.

> "We must be a place where facts matter, where reasoned and respectful discourse and debate serve as arbiters of truth."

@BalarkaSen The shutdown of Mos was a self-fulfilling prophecy. Don't fall into the same trap.

@Loong Ehh, I don't understand this dramatic comparison between the fate of ME and so on. What I said wasn't a prophecy, but an educated guess on my part (I have given several reasons on why I think this) - the goal being to lay out the problem that I think is with the moderator-user interaction.

And potentially what can happen because of it.

whelp... after 2 years worth of chat bans being issued in only about a month, perhaps "self-destroying" prophecy is the more accurate phrase

> “My philosophy is very simple. When you see something that’s not right, not fair, not just, stand up, say something, and speak out.”

@BalarkaSen Honestly, I don't trust your educated guess one bit. I don't think you have anywhere near enough information to come to that conclusion. I'm also somewhat suspicious of the logic you've used to arrive at the conclusion, but that's more a matter of opinion.

Dishonestly, the room is going to be just as much fun without 0cel...

@DavidZ I think I've been quite clear on the logic I have proposed. It's not a matter of "trust", it's a matter of axioms and propositions.

Of course I don't have information on how moderation works internally, hence why I listened to what you had to say on that matter above.

But to conclude that the moderator-user interaction and communication is fundamentally broken one doesn't really need to know that :P

You can of course falsify my conclusion by actually discussing the points I have raised regarding the issue in the past couple days in this chat. To outright "not trust" the argument is equivalent to debunking, which is not part of skepticism.

@BalarkaSen Not that I've seen.

See this, this and on a tangentially relevant note, this

Also point (2) in this summary of events, an answer to which was proposed but it was exactly the message which stands as conflicting with your proposed stance of the importance of flags, so maybe there's something else to say?

Ah, and this was a one-line summary of what I thought was the problem.

I think that's all.

BaSe: from all your comments you seem to care a lot about SE chat/ mod policy & have some affinity for that pov/ concern but maybe the real answer right now is to disengage from it some/ more. the idea of a chat mod algorithm is really off. its fundamentally not algorithmic. its subjective, its imperfect/ fallible humans making imperfect decisions. the refereeing metaphor is apt. DZ is being as candid as possible about what goes on but its also unavoidably vague/ opaque in places...

reading thru that, some observations/ concerns of mine afaik consistent with available knowledge (1) the mods are never going to tell us exactly how 0celo7s suspension was decided. (2) that means it could have been made by 1 or more mods, possibly even SE mgt. (3) as far as outsiders know, it could have been a single "outside room" mod who decided/ unilaterally issued the 1yr suspension and all other mods fell in line so as not to undermine it. there was apparently chg from 2day to 1yr by ?...

I don't know why you'd think a subjective decision-making process cannot be an algorithm. It's not a deterministic algorithm, it's a stochastic one.

@BalarkaSen actually am in favor of making it more algorithmic if it comes out as giving users (not mods) more flexibility/ "wiggle room" but humans will always make the final decision. it reveals the nature of power/ authority/ sociology on SE, a topic that is not too much considered, but comes out occasionally on some exceptional events...

I know the reasons behind 0celo7's suspension, I have seen the mod message that was communicated to me outside of chat. DavidZ just explained the suspension procedure in quite a bit of detail.

@vzn Sorry, I don't follow any of what you said.

@BalarkaSen he explained it in apparent detail, and there is some appearance of transparency, but its indirect and hes not implying any of it specifically was carried out in the present situation, its generalities.

If you had seen the mod message it would have been clear to you that there was no conspiracy behind the suspension.

@BalarkaSen the mod msg presumably does not indicate any one particular mod or group of mods made the decision, those msgs have a element of vagueness just like everything the mods are stating in the chat room(s)/ meta on the subj. as an analogy, its not unlike college rejection letters in that way... as for "conspiracy" think it would be interesting to deconstruct the exact defns of the word wrt this situation... o_O

You haven't seen the mod message, so that's dangerously close to public speculation regarding 0celo7's suspension, which is not tolerated in SE.

So I am not going to continue this line of discussion.

("conversely") you might think, having seen the private mod msg to 0celo7, you have more special information than what is open/ public knowledge, but am dubious of that. the general outline is known and substantially discussed.

I don't think I have special information. I do think I have more definitive information.

In any case, the fact that publicly discussing speculative interpretations of individual suspension is not allowed applies to me as well, as a user of SE.

0celo7 shared the rough details/ outline of his suspension msg in his meta msg.

@vzn I do not see where the meta question was edited after the moderator message was sent; as far as I know that information is still private and you have not seen it. Until you see a direct quote, do not speculate further.

@vzn You know, stating what the mod message presumably does or doesn't indicate is just exactly the sort of speculation that we have asked you to avoid. Take a day off of chat and don't do that any more when you come back.