Comment by anthony_r
5 years ago
You're not seeing the other side of the coin - the huge amount of spam and abuse that such systems correctly identify and remove. If every abuser requests those explanations (which they will) there will be far more spam going around the Internet.
Just think about the army of "Facebook content moderators" who were a popular topic on HN recently due to the concerns over their mental health.
(I am offering no solutions here, for I know none)
I think this is a convenient narrative for an abusive pattern of behavior by Google. The company is infamous for having non-existent customer service. It's not a matter of their AI having too many false positives, it's that when there is a false positive you have literally no recourse even if you're a well known business partner.
Are we really going to believe that Google, one of the highest grossing companies in the world, doesn't have the money to provide even basic level customer service? If it were really a matter of not being able to afford it, certainly they could offer it for a fee. No, they're stubbornly refusing to address the issues, relying on this lie, and using their market dominance to avoid having to answer for it.
Technically they do offer customer service if you pay them with their Google One product. I have phone numbers and human access very quickly, because I pay for it.
Although obviously if they banned me, I wouldn't have access to my direct support line anymore.
> Although obviously if they banned me, I wouldn't have access to my direct support line anymore.
Which they will do literally on a whim. Who are you going to call then?
9 replies →
> If every abuser requests those explanations (which they will)
It's not a request, it's a requirement. If your account is suspended, you deserve an explanation. You should get one without having to request it.
I'm not saying that companies shouldn't be able to suspend accounts temporarily. I'm simply saying that there needs to be a way to get your account unsuspended if you're innocent. The way it "works" now is that innocent consumers are without any recourse whatsoever.
I heard on a podcast recently that a trading system needs to keep logs of why a particular trade was executed for several years just in case the authority wants it. So it isn't too much effort to build a similar report or log of behaviour to explain why someone was banned.
Obviously this will also help the spammers who will use this information to get around the filters.
Complete speculation because I don't actually know how this works, but I wonder if the explanation would be something like this:
"You've been banned because our black box ML algorithm says your usage patterns share similar traits to those of known spammers."
2 replies →
Podcast link: https://www.twoscomplement.org/, I think it was the latest episode.
For the record, they don't give away these explanations because such explanation would hint the spammer to what they should _not_ do next time, to avoid getting caught. Same as with anticheat software.
> they don't give away these explanations because such explanation would hint the spammer to what they should _not_ do next time
We've heard this excuse countless times, but it's simply not acceptable. The foundation of our legal system is that it's better to let a criminal go than to punish an innocent person. How many innocents have to get caught in the crossfire before we start protecting them?
23 replies →
When someone is in court on charges of child abuse, maybe we don't want them to know in case they (After serving their sentence) or their friends go for reprisals. Maybe the next child abuser might know their likely avenue of getting caught. Yet still we tell them the charges and evidence and give them a chance to defend themselves. Often in my country, given the damage such allegations could cause to both the victim and alleged (but not yet proven) perpetrator, we don't even reveal the identities of culprits until there's a guilty verdict.
If we can extend that courtesy to people accused of child abuse, surely we should extend it to people accused of internet spam?
I imagine if that happen in real courts. And You got jail without any info on why on how to evade - or You will behave properly on not go in jail
1 reply →
You don't have to tell them how you detected them but you can tell them what they did wrong. A lot of times when these cases come up there is nothing in the reason you got banned that would help you avoid the ban. It's purely to avoid any kind of accountability (if they say you got banned for a reason that is plainly not true because their algorithms suck)
It would also give non-spammers a better understanding of why they were banned and teach them to be better humans. It’s this lack of empathy that’s leading to more and more anger online.
Relevant xkcd: https://xkcd.com/810/
I think we give up on that when we agree to the rule
Google has the right to suspend, remove your account without prior notice
I'm sure there should be a clause like that in their TOS
This is an age old problem in the criminal justice system. A solved problem.
After a lot of trials with various approaches, we settled on letting some criminals go free over convicting someone on weak evidence. Second we decided that trials should be open and evidence viewable by default.
Finally you generally have the option to give some security to stay out of jail during trial.
Closing a google account is a punishment worse than many criminal convictions. And will only get more important as we progress to an all digital existence.
> Just think about the army of "Facebook content moderators" who were a popular topic on HN recently due to the concerns over their mental health.
Hire them directly instead of via labor farms, pay them an actual living wage, give them full health benefits, and hire enough of them to prevent overload.
> the huge amount of spam and abuse that such systems correctly identify and remove.
Maybe allowing single service providers to capture several billions of users is the problem here.
Ding Ding Ding!!!
Perhaps the process should cost $100 or $500, so that actual spammers can’t use it
Maybe they really just need to offer a paid account option with real support, since that has much better incentives
Yes, refundable if the company ban proved in the wrong. Sounds like a great solution IMO
No need in charges. Strong person identification via Passport or Bank. Limit those request per identified person or throttle them.
There is a paid option: for $6/month you can use gmail with your own domain name. It's targeted at businesses but you can use it as an individual.
https://workspace.google.com/pricing.html
It includes support, but I'm not sure if that helps in cases where google thinks you have abused the service. I just use it because I like having my own domain, and so that I don't lose access to my email if google locks me out. The idea is that I can update my domain's MTX records and use another email service.
Support does not include if your account gets suspended or if you lose access.
We had a paid Google App account. One of our workers would only login from their computer. It died, and she tried to login from the new computer. It gave a unrecognized machine error, and we had to hire someone to resuscitate the old computer for her.
I know of a company that had the entire companies' accounts suspended without warning because one user did something that violated their terms, but they could not figure out what. The company lost three months of revenue from it and I am not sure if it caused bankruptcy. No help at all from G.
2 replies →
Meta: my comment above is being downvoted significantly. I'm not sure what I did to offend. There was a remark about the need for a paid option, and I pointed out that it already exists. I have no agenda here and was just sharing what I know.
1 reply →
But surely it’s possible to use methods other than what currently seems to be the first and only solution: “your account has been banned, bye”.
For example, if an automated system thinks an account is sending spam, enforcing a (very low) outgoing email rate limit would be a much more reasonable first step.
So just start charging for service, and keep a non-refundable deposit for spam/abuse.
Let every abuser requests those explanations, if the decision doesn't change, the money is still kept, which funds that service.
no if the AI can be used to automate the banning it can be used to provide the electronic news email of the rule violated.
Well their main argument against it is that if you don't tell scammers which rule exactly they are breaking they can't improve until the app is approved. But of course that hits normal customers too. It's the equivalent of arresting random people on the street and not telling them why - surely, innocent people will just get their lawyer to free them.
aka due process
So what is the proper Blackstone's ratio for you in these situation?
Is 1000 innocents ok to punish as long as 1 spam message is stopped?