The case for an "Escalation" label for political threads
This is a follow up to the thread from a few days ago, and specifically my comment in that thread regarding the use of a "Escalation" label.
As many users identified in that topic, political discussion on Tildes has the potential to become very heated, very quickly, and often the standards of discussion on these topics is below what we expect elsewhere on Tildes. In that thread, many suggestions were offered in order to remedy the situation, including banning overt political content entirely, more liberal moderation by @Deimos, more liberal usage by the community of labels, addition of new labels, and more. All of these solutions have their advantages and disadvantages, but I want to talk about the one I believe would be the most effective and least disruptive to the site as is: addition of new labels.
Right now, there are two main tags that might be used on a comment that is seen as falling short of Tildes's standards: noise and malice. Users seem to have some variation in how they interpret how each tag should be used, but it seems like there is at least some agreement on the 'noise' tag being used for comments that are clearly low effort. Users seem to have more hesitation to use the 'Malice' tag, however. While it is sometimes clear when a comment is hostile or malicious, this is not always the case. Argumentative is not always hostile, and sometimes topics are naturally contentious. One takeaway from that thread (for me) is that labeling something as malice confers a judgement on intent, and users are not always comfortable doing this as it can be difficult to tell if someone truly meant to be malicious. But in political threads, the intent matters less than the effect a comment has in a discussion. Someone can not be acting maliciously, but still be clearly making the situation worse. This is the point of an 'Escalation' label.
An "Escalation" label should be applied to comments that have made the situation worse.
Furthermore, an "Escalation" label would not only affect the sorting of a comment or thread, but has the potential to halt the discussion if there is too much escalation in a short amount of time. Here is what I envision:
Define the heat of a comment (as in, "ohhh this conversation is getting heated") as follows:
H = k*n ∑ Ni / di
where k is a tuning constant, n is the number of escalation tags given to the comment in question, and the sum ranges over the comment's direct ancestors and descendants in the thread with Ni being the number of "Escalation" labels given to the other comment and di is the distance from the current comment to that other comment. Here is an example thread:
.
├── A
├── C0
│ └── C1 (N=1)
│ └── C2 (N=0)
│ └── C3 (N=2)
│ └── C4 (N=1)
└── B0
└── B1
The heat of comment C3 would then be
H = k*2 (1/2 + 1) = 3k
Finally, define the heat H(T) of a thread T to be the sum of the heats of its comments. My proposal is that if the heat of a given thread surpasses some threshold value Hc, replies are locked in that thread only. This essentially shuts down extremely heated conversations before they get out of control and cause an entire topic to be locked.
The above definition can obviously be modified, but it has a few good properties that I think should be retained.
- It takes into account the relative positions of comments. A thread that is 20 comments long that has a comment with 1 "Escalation" at the beginning, midpoint, and end is probably a better and more controlled situation than a thread with 3 "Escalation" labels in a row.
- One extremely heated comment (n is large) that generates many okay or slightly heated replies (n~1) is oftentimes just as bad as many comments that each escalate a bit (a long chain of comments, each with n~1).
- It considers a the whole thread as opposed to on a comment by comment basis. If there is only one person in a thread posting heated comments, even if the replies are measured and reasonable, there is a good chance that thread is not producing a worthwhile discussion. If that one problem user stays problematic too long, eventually the heat of the thread will surpass the threshold and the chain will be locked.
I am sure there are disadvantages that I am not thinking of right now, but I truly think a system like this could be beneficial if implemented and used by Tildes. Furthermore, if two people are genuinely interested in the discussion and want it to continue, it is in their interest to avoid posting comments that get generate a high heat score so that the thread doesn't become locked. If they are not interested and keep escalating anyway, that conversation probably shouldn't continue.
I am interested in your thoughts on this idea. However, I don't intend for this topic to become a repeat of many of the suggestions and comments in the thread linked at the beginning - I don't mean to reignite that discussion.
I already created an "argument/bickering label" feature request on Gitlab quite a while ago, but it was not very detailed... so I have edited in a reference to this topic in that issue, since they seem so closely related.
https://gitlab.com/tildes/tildes/-/issues/544
The word “bickering” conveys the sentiment that the discussion is unimportant and may seem hostile for those involved, making the situation worse. I believe the term “escalation”, as proposed by OP, is preferable.
Thanks very much for putting so much thought into the idea and writing this up. I think I'd definitely like to experiment with something along these lines, and there are lots of good thoughts and things to keep in mind between your post and the comments.
I don't know if actually locking is what I'd want to do—I think I'd like to try more of a "cooling off" than "totally stop". That is, do something like only allow each user to reply once every few hours, which would hopefully help de-escalate things more naturally. A lot of the heated threads seem to happen when two (or more) people are replying back and forth fairly quickly.
Lock it long enough for everyone involved to have slept on it. If sleeping on it doesn't cool the parties involved off I have a hard time imagining what could. We're not wedded to the 24 hour turnover, may as well take advantage of the longer turnover period.
If we had a time-sensitive comment tree system of some kind, I might also suggest filing the comment root that generated the lock under some form of 'old thread' tab and letting the comments start fresh from zero again in a new discussion. Give people the opportunity to forget.
What might be most important is just giving users something to play with so that we can begin sculpting it into a modding tool. A modding tool that helps the group within the group break up fights, perhaps... or at least stops fights from getting out of control. Collect some data, maybe we can find a flashlight in there that points the way forward.
"Escalation" isn't bad, but I think "Heated" might be more direct? Also, it seems like it should be a useful signal but I'm not sure how best to use it or what you'd be doing by placing this label? I guess you think heated discussions should be locked?
"Escalation" covers three or four use cases where "heated" only covers one or two. You can escalate because you're just losing your cool (heated) and by overreacting or getting defensive ("heated" can also fit here, but less well). But you can also escalate by being glib or tactless (unintentional) or posting flamebait (intentional) that pisses people off.
"Escalation" covers all four since it ignores the question what emotional state you're in. It just addresses the perception of your conduct, regardless of how you're feeling or what you intended to say.
It's the difference between saying "Hey, you're overreacting" and saying "Hey, it might be best to take a break from this."
One point I'd like to address is that "Heated" would be more straightforward and indicative, whereas "Escalation" is vague. Maybe the latter covers more ground, but the former is clearer.
“Heated” is emotional and has the potential to make things worse. “Escalaton” is a common word with a well known meaning. I believe it’s a sound choice.
How about "Confrontational"?
As an individual that is admittedly, unintentionally, tactless at times (working on it), I agree that perhaps "Heated" isn't enough, but think that the label should differentiate between something that is intentional and something that isn't. Especially since it can be the person placing the label misinterpreting what was said due to their own frame of mind at the time.
I'm sure everyone here has misinterpreted the intentions of a comment at some point.
I think this is precisely the issue with the Malice label currently. Well, not really an issue, but I think people are somewhat hesitant to use it because there is a big area between Malice and non-Malice where it can be difficult to discern if something was truly meant maliciously or just came off wrong. Ideally, an "Escalation" label should be agnostic to the intent of the comment. Sometimes, comments make things worse even if they were meant well. This could be because someone worded something poorly, made an offhand remark that was offensive to someone, joked about something that maybe shouldn't have been, etc. None of these are necessarily condemnable in a vacuum, but in the context of a comment chain it can be very clear which comments had the effect of lowering the quality of discussion or making an already tense situation worse.
Very ideally, in most cases escalation tags (in political threads) shouldn't even really reflect negatively on a user. Most of the time it takes two to argue, and unpleasant shouting matches are often more a product of many people throwing clumsily worded arguments out into the wild without much thought for consequences. Such a tag's main purpose would be to address problematic threads as opposed to problematic comments, although obviously comments make up threads so the distinction is a bit blurred.
As you note, everyone can be tactless at times. An escalation label (or rather, the 'threat' of one) would cause people to slow down a bit and consider what they've written. And even if something comes off poorly after such consideration and gets labeled as escalating, nothing much should happen unless it is surrounded by other escalating comments in a thread. One-off poorly worded comments are mistakes every person makes, but a thread full of them is a mistake made by people, plural.
“Confrontational” implies inference of intention. It’s also emotional and accusatorial. I still prefer “escalation”.
Locking is certainly one of the more extreme outcomes that could be implemented - it was my first thought because it seemed that the consensus in that other thread was that low-quality political discussions are not really worth having on Tildes. Other options include a temporary lock (again, just for the thread and not the entire topic), or they can be auto-minimized or outright hidden (as in, you don't even know they're there) based on individual user preferences. I am more of the opinion that shouting matches in the comments lower the quality of a topic as a whole, and leaving them there for users who like to get into heated arguments isn't really helping any. There are plenty of platforms where you can go and shout at people you disagree with, and I'm not sure Tildes should be one of them.
A lot of tech minded people look for technical solutions for social problems. Sometimes that can work, most of the time the underlying social root cause isn't treated/corrected/changed.
Human moderation is a social solution to the social problems of online conversation. Sometimes crude, sometimes adequate, but fundamentally a sosial solution to issues technology will not solve.
Tildes strives to be more than lowest common denominator fluffy conversation. We need to give each other social cues for when people cross a line.
That's sometimes uncomfortable to do, but it's often much more effective at deescalating the situation than an assigned arbiter stepping in: One other normal person spoke up.
We need to create the conversations on these platforms by individually saying what's okay, what's not and trying to lead by example ourselves when we contribute in conversations that can be divisive.
Politics is often the topic to give early warning that there's a social problem that needs solving. I think it'd be a huge mistake not to listen to that canary in the mine shaft: This is about how we talk to each other on all topics, and a warning we should pay close attention to.
I've actually expressed similar feelings in the past, and I am often wary of overly-engineered solutions to social problems. I do believe though that users online should be empowered with tools to help them "speak up", precisely because it can be so uncomfortable to do. This tools can also mitigate the fact that in-person reprimands from people that someone knows, trusts, and understand will always have more weight than a digital one in the form of a sternly worded comment over the web.
Ultimately Tildes' culture and standards will be what the users make it, as is the case with any site. I actually think the average Tildes user is pretty good at speaking up and saying when a comment is below what they think the standard should be. Nonetheless, we have more tools at our disposal with online conversations than we have with in person ones, and I don't think we should shy away from using them. If we can implement a system that let's the community easily say what is okay and what isn't, in a way that is fair and reasonable, I think that would be a very good thing.
Ultimately it will have to be a combination of factors - social and technological, given that the web is inherently both social and technological.
This is very clever, and I appreciate all the thought you put into it. It's situation-agnostic so it scales well to escalations of any type or location, and it prioritizes isolating and addressing the escalation without impacting other threads, with the option to scale it up to the topic level should things get too out of hand across the board.
Thanks for thinking of a creative solution!
I really like this idea, including the choice for the word “escalation”. It’s neutral, unemotional, and doesn’t make assumptions about people’s intentions. It has the potential of cooling things down without aggravating those involved. My “vote” is to keep it.
I cannot judge math I don’t understand, so I don’t know for sure how loose or stringent this proposal actually is. Doctors usually start prescribing a low dosage that they increase if necessary. Maybe we should do the same.
Finally, it would be interesting to know how much labels are currently being used, and which ones are more popular. I don’t see many labeling on Tildes, maybe we should remove the ones nobody uses.
Just as a note, the only label visible to users explicitly is the "Exemplary" label that you sometimes see. Other labels, like noise, manifest themselves mostly through minimized comments at the bottom of a page. Others just affect sorting, so it can be tough to tell when they're at play.
I see. Are other labels being extensively used?
Just based on deduction from noticing their effects, noise (which collapses comments) and offtopic/joke (which both artificially lower comments in sort) seem to get used decent amount, but it's not a crazy amount. Maybe once or twice every 100 comments?, if I had to guess. @Deimos is the only person that would know for sure the exact figures though.
I don't have any insight on that point, unfortunately.
The advent of a label that can be used by group of people to silence the conversation being had by others is a dangerous one. Tildes, as we all know, thankfully doesn't have a downvote button, because again as we all know upvoting and downvoting has long since lost the initial meaning of the comment adding/subtracting from the conversation and just became an agree/disagree button.
The noise label can already minimize a conversation so that if you want to read it as a 3rd party an extra step must be taken to do so. The addition of a escalation label that can lock a thread can and will be used as a "I don't like you talking about this, so I'm going to vote to censor you" button.
You certainly bring up valid concerns regarding the potential abuse of a feature such as this, however I think this is largely addressed by something in the docs:
https://docs.tildes.net/philosophy/people#trust-people-but-punish-abusers
In order to do so will require strict, near 100%, moderation of the label. This may be feasible at our current size, but as so many here push for greater numbers of users it will quickly become unsustainable.
What? No it wouldn't, because 100% of the label uses are unlikely to be abuse, and if the label is misused others are likely to notice and report it... just like how it works with all the other labels. And if someone does get caught abusing the label, they can be banned... although ideally a more granular punishment system should be put in place where instead they are either temporarily banned, permanently have their ability to use labels revoked, or temporarily have their ability to use labels revoked, depending on the severity of the abuse.
I encourage you to view any political subreddit and go check the most downvoted comments in a thread. While there will be some flamebaiting and trolls, you'll also find perfectly sound and well thought out counterarguments that didn't fit with the groupthink of that subreddit. Give the masses a downvote button that also silences those counterarguments entirely and watch the echo chamber grow. Sure we don't have the factions that subreddits have created here (yet?), but to act as if Tilderinos would never do such is ignoring the obvious demographic leanings this place already has.
That said I think I may have already thought up a way around this.
Labels are not downvotes, nor do they function remotely the same. Downvotes are equally weighted to votes on reddit. 1 upvote - 1 downvote = 0 total. Whereas labels require multiple people to apply them before they have any effect whatsoever, and even then the effect is not 1 to 1 with votes. And while right now every user has 0.5 label weight (IIRC), the system is in place to allow certain users to be given more (or less), which could help mitigate a lot of abuse, by for example making all new users only have 0.1 label weight until X time or X trust (if/when that gets implemented).
Another thing that differentiates labels from downvotes is that on reddit moderators cannot see who applied downvotes, and the admins don't care about it being misused, so there is no way to punish those who misuse them in a subreddit. In fact even defining what misuse of a downvote is, is hard enough due to the ambiguity of their purpose. However that is not true for labels, which have a more explicitly spelled out purpose and so misuse of them is more apparent. And while only Deimos can currently see who applied labels, that may not always be the case, e.g. if more admins join Tildes, or moderators are given the ability to see which label have been used in their groups so they can flag them for admin review if they think they have been abused.
"Noise" could be used to do basically the exact same thing, and yet the entire time I have been using Tildes I have only seen that misused a handful of time. And I suspect that is largely because of one of the interesting things about the noise label that you may not be aware of... its effects can be cancelled out with enough votes. See: https://gitlab.com/tildes/tildes/-/blob/master/tildes/tildes/models/comment/comment.py#L336-338
Working half as well as a standard reddit downvote is very nearly the same. Discussing the potential weighting a user's votes/labels by age of account or any other factor is just another recipe to get Digg-esque superusers.
Here as on reddit the admin(s) can see who votes and if we're discussing hypotheticals then reddit can just as easily allow moderators to see which users vote in which direction.
Reddit's own reddiquette page outlines them quite clearly and your belief that they are ambiguous in spite of this lends to my point that given a censor button the masses will use it for such even if that wasn't the intent.
Noise minimizes and downvotes, but doesn't lock. It turns the volume down on a commenter, an escalation lock puts duct tape on the commenter's mouth.
I'm kind of inclined to agree. This seems like something people should hash out by talking about it rather than anonymous tagging. It's easy to attribute an anonymous tag to people being haters, which mostly has the effect of getting people even more heated rather than less. If the intent is to remind people to chill out or tell them they're not coming across well, that doesn't really have weight behind it unless it comes from someone you respect (either because they have authority or because you actually respect their opinion). What's probably needed is a cultural norm where it's okay to say "Hey, I think you could have said that better" or "You're being kind of a jerk right now."
The challenge, of course, is that comment threads are 1 to 1, but you can always just tag someone to say "You and @[user] need to calm down."
Edit: I suppose part of the challenge, though, is that some demographics could be predisposed to being abrasive. So this can still come across as tone policing as a silencing tactic. But at least it's more in the open.
Yeah, this is probably the most glaring issue with a system like the one I have described. It's difficult to change cultural norms on a site, and even then I'm not sure a comment telling me I'm being a jerk will have much weight from any pseudo-anonymous user online, especially if I'm in the middle of a conversation I feel passionately about.
There is always a risk that any site mechanic will be abused by users. I don't think it's necessarily good to design a site around that (or necessarily avoid adding certain features), though. As @cfabbro pointed out, the large majority of users are good, and the proportion of people who would abuse an Escalation label is certainly smaller than the proportion who would use it as intended. The effects of abuse are quite severe in this case though, because as @AugustusFerdinand points out it could essentially function as a censor button.
I will point out however: how many people are effectively 'censored' as is because they don't feel inclined to participate in threads that are already a clusterfuck? How many times have any of us typed out a comment only to not post it because you don't want to get involved in an ongoing shouting match? Low quality discussions of politics already suppress participation, even though these topics are extremely important to discuss. It's a sorry state of affairs when there's doubt that political discussions online can even be good in principle.
Overall I definitely agree that the potential for abuse is something that should be seriously addressed, but I think low quality political discussions probably have more of a chilling effect across the site on these topics than coordinated abuse of such a mechanic could feasibly have.
I just had a thought and not really having a place to put it I'll add it here:
What if the escalation/heated/confrontational label was only allowed to be selected by 3rd parties after one of the individuals involved in the discussion elected to use it? Requires them to be a bit more self aware, but leaves out the easiest route to censorship.
Example: You and I enter into a low quality confrontational argument about some issue. You and I have access to the escalation label for the comments between the two of us while others do not. At some point you decide that I'm being confrontational and label one of my comments as such. This opens the label to all users for our entire string of replies to one another and so the mob can decide if we need to be separated by labeling such. By labeling my comment as heated you open the gates for the opinion of others to decide who (be it one or both of us) is at fault for the argument. Those that are labeled the confrontational individual are the "abusers" in the trust people philosophy.
Thoughts?
This is certainly an interesting idea, I would have to think more about potential implications but I definitely agree that it seems like a step in the right direction in combating potential abuse. I like this and think people should think about it.
Another idea is to allow the escalation label to limit response time as steps up to a lock. You label my comment as an escalation and it opens others to do the same, another user does so as well and a response timer begins where you and I cannot reply to each other for 15 minutes. Giving a cooling off period and perhaps a time to reflect if the comment timer indicates that the thread was labeled as escalating. The timer could grow with each labeling up to the overall thread threshold that locks the string of comments.
I think the idea of an escalation tag is a good one!
However, I wonder, what's to stop someone from replying to another comment somewhere in the thread and restarting the process of escalation once a particular thread gets locked?
One good thing about the system I described above is that it takes at least two comments (it could even be modified to be two comments from unique users, but let's not get ahead of ourselves) to get the ball rolling on a heat score for a thread. So if one user gets a thread locked and just replies elsewhere in the topic with another escalating comment, as long as no one takes the bait, the thread should be okay. If there is one problem user who is the only one causing threads to get locked by repeatedly escalating, there's a good case to be made that that user is acting maliciously and that label should be used instead. Nonetheless your point is a good one and should be considered.
Maybe a solution would be to prevent users who took part in the thread from tagging each other in the comments for a little while? In theory that's ok but I'm not sure how it would work in practice, just locking a thread due to escalation could be attributed to everybody / everyone else whereas specifically preventing someone from doing something, even if it's an automatic process, could be seen as more direct or targeted.