Hacker News new | past | comments | ask | show | jobs | submit login

Not an uncommon practice, Jeff Atwood calls it "hellbanning": https://blog.codinghorror.com/suspension-ban-or-hellban/





No, he calls it slowbanning. Hellbanning is when you let them keep posting but other people can’t see the posts. Hackernews hellbans as well and you can see the comments from the deplorables if you turn on “showdead” in your profile.

I know that as shadowbanning. This is one of my favourite reddit posts on the subject:

https://www.reddit.com/r/tifu/comments/351buo/tifu_by_postin...


IIRC hellbanning is a variant of shadowbanning that prevents easy discovery of the banned status by putting all suspect accounts into the same "hell" invisible to normal users.

This works well if you have a network of fake accounts from a single "persona" or ring of personas - by all their indications they can't see their own posts are being ignored.

Notice: it's almost the exact response to the persona management software problem [1] (aka bots).

[1] https://www.theguardian.com/technology/2011/mar/17/us-spy-op...


This also works incredibly well for cheaters in videogames.

Give them their own queue with their own games with other cheaters to play against, and as long as nobody is cheating in a way that breaks the servers, they can play their own version of the game if they want without ruining the game for those who don't cheat.


This reminds me of someone telling me that they are using cheat sites for an online Scrabble game because they suspected their opponent (a “friend”) was cheating. It’s hilarious to think that two humans are watching two instances of a likely optimal bot play against each other and rooting for their instance of the bot.

Also used in the excellent TV series Mythic Quest :)

It doesn't work for logged-out users. If I can just look at e.g. the Internet Archive's copy of reddit and see if my accounts are in there, it defeats the purpose.

Neither reddit nor HN make any attempts to make it hard for sophisticated users to figure out they're shadowbanned.


Forcing you to query IA at least reduces the frequency of feedback, as they're taking snapshots instead of giving a live feed. You could also shadowban IA. You can also do things like guess based on IP address or browser fingerprinting, or require a login from various IP ranges.

Of course your main point - that this is all terribly imperfect and won't stop a determined, sophisticated user, who has realized what's happening - is spot on. That, however, is perhaps a rare combination, rare enough to simply continue dealing with manually.


The feedback doesn't have to be very fast. If I'm botting correctly, my accounts will almost never be banned. Even once every 24 hours will be more than enough.

IA was just an example, and Tor would be easier. But anyway, I think it shows the flaw in doing so:

> You could also shadowban IA.

If the spammer manages to get all the IPs hellbanned just by looking at things, he gets more eyeballs on his spam.

My point is, you can't get much better than normal shadowbans, which are trivially detectable for moderately sophisticated users (just log out and try to check your profile) but not anyone else. "Hellbanning" is a stupid extension of this concept which only works in video games.

Also, shadowbanning is a spineless and deeply unethical move. If I get banned, I know what I did wrong and can reflect on that. If I get shadowbanned, I'm just screaming into the ether. That is not a Good Thing™, it is atrocious.


> Even once every 24 hours will be more than enough.

Depends on the use case. Once every 24 hours is a lot easier to moderate than a minute by minute spam wave.

> IA was just an example, and Tor would be easier.

TOR would indeed be easier... assuming it's not already blocked through other means, as it frequently is. There's a whole ecosystem around blocking TOR and other proxy mechanisms - imperfect and permeable though they may be.

> My point is, you can't get much better than normal shadowbans

Not sure I agree or that you've supported your point - however, even shadowbans are often unnecessary. The goal is never perfect moderation, merely to stack the deck in favor of the moderators for blocking problematic content in terms of time effectiveness until either the moderation effort available can handle it, or until the spammer moves onto easier, more cost effective targets (which even basic shadowbanning can achieve, mooting the need for better tools even if they're available.)

> Also, shadowbanning is a spineless and deeply unethical move.

As a first line of defense against mere rules breakers, I might agree. As a second, third, or nth line of defense against particularly problematic ban evaders and spambots, I will gladly resort to such tools - or worse - and sleep soundly at night.


I guess we know who gets banned from places!

We only do that when an account doesn't have much history on HN and there's evidence of spamming or trolling. For established accounts, we tell people we've banned them and why.

https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...


Oh wow! Thank you! The setting was on for me. (By default? Or maybe I turned it on.) I just turned it off. Thanks for this suggestion.

Even if it was accidentally set on, one sees [dead] against shadow-banned, greyed-out comments.

I think of myself as hard-nosed, but I don't think I could be this cruel to someone.

Reading said comments may convince you otherwise.

I've always thought people might catch on, since no one is engaging with their takes. I'm curious if letting bots do some markov responses might keep them in the dark a little longer.

I see it far less frequently now (thanks @dang!), but a few years ago it wasn't uncommon to see shadow banned HN users continuing to post for years, talking into the void. Sometimes I'd look into their post history and so many of them had been banned for utterly trivial reasons, it was pretty sad.

Would you have a profile or two handy that I can take a look at? I left showdead on for a while but found it useless in terms of coming across interesting comments from such users. Thx.

I seem to remember being pretty upset when I found out Terry Davis of Temple OS was being shadowbanned. Maybe it was just temporary for a short period or a single comment, but I didn't like it.

I found this (related?) poll https://news.ycombinator.com/item?id=7818823


Terry was permanently shadow-banned, the subject came up quite often. He suffered from schizophrenia and would make racist and paranoid comments about 80% of the time, but his remaining posts were often pretty technical / insightful. As the victim of an illness it seemed unkind that he was banned, but at the same time a lot of his output was obviously very offensive.

Cases like that are why we introduced vouching, which allows the community to restore the good posts made by a banned account.

Incidentally, if a banned account is making only good posts, we're happy to unban it. I often look at the recent commenting history of banned accounts in the hope of finding such cases, and users sometimes email us about them (as mirmir mentioned elsewhere in this thread). That's super helpful!

One strange phenomenon is that there are banned accounts that post good comments, but revert to posting bad comments that break the site guidelines as soon as we unban them. Then we ban them again and their comments get good again...go figure. Any large-enough population sample includes a long tail of behaviors.


> One strange phenomenon that comes up occasionally is that there are banned accounts that post good comments, but revert to posting bad comments that break the site guidelines as soon as we unban them. Then we ban them again and their comments get good again...go figure. Any large-enough population sample includes a long tail of behaviors.

It would be somewhat ironic, if re-enabling interaction with the community is what's driving them to back to bad behaviour. You know, HN as a bad influence.


That's one model; I've come up with others over the years. But there's no easy way to test any of them. We can't simply ask, either, because asking a question like that is enough of a perturbation to significantly affect what one is asking about—and who knows if people are even aware that they're doing this to begin with.

I've seen people like this one message boards before and I always just assumed they handle disagreement poorly. Specifically, they can make an original comment that is fine, but if someone responds in a way they don't like, they kind of fly off the handle. So I guess if they never have people respond to them, they seem fine.

I only lurked on his comments, but I remember some to be very interesting and thought that his unique approach to OS design was food for thought

Somebody should make a shrine of his lucid and technical posts. And also a book.

Anybody know of such things?


If I see an account that's been shadowbanned for years, and has consistently posted appropriately for maybe several months, I report it via the contact address.

I think nowadays you might catch on because the culpit might not be identified to your site with both their phone and their PC or something among these lines.

> deplorables

Isn't this a pretty mean thing to say? Just making an account from tor is enough to get you shadowbanned in hn.


That's not true. Software filters sometimes kill such comments, but the accounts themselves are unaffected, and moderators review the killed comments and unkill the clearly good ones, and mark such accounts legit to immunize their future posts from those filters. Also, users often vouch for comments that have been killed in this way, which restores them.

My main issue was with the deplorables comment. I actually quite like the HN system of vouching.

> and mark such accounts legit to immunize their future posts from those filters

I remember seeing you restore a post from someone who made their account via tor and their comments were auto-deleted. Their next post was autodeleted in the same way, so I presume that this feature is buggy (or was, as this was quite a while ago).


There might be other explanations. For example, if an account shows signs of being connected to previous banned accounts, we might unkill a good post, but not immunize the account overall, until it establishes more of a track record.

I'm glad you like the vouching system! I still feel like it's the best single change we've made to HN since pg retired.


Yep.

And you can vouch for good comments from shadowbanned accounts, just as for manually flagged comments.

Being rather paranoid, I check this account occasionally via Tor.


Let me check it. If you see this pls vote or reply.

ack

And that's the reason why arts don't follow these rules.

Imagine Van Gogh on HN...


Unfortunately I can't upvote you

I agree with the sentiment, shadowbanning is a passive aggressive way to hide communities problems under the carpet by hiding them from the public

Banning is legit if done publicly with reasoning, so people can make a clear idea of what happened and why, like

> you broke rule number 3

not when it is done just to keep "the community clean" without any explanation whatsoever

I guess this is the reason why HN rules are so vague


I always referred to this practice as “tarpitting”, I was surprised to see nobody else here does!

“Tarpitting” is a specific type of network-layer defence which is not related to degradation of service. A tarpit will typically stretch out response time to network-layer and some application-layer communications in order to waste wallclock time of spammers.



Applications are open for YC Summer 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: