AIs and other opaque processes making decisions are incredibly frustrating. I'm still wondering why my tea-of-the-month business [0] got turned down by Stripe when I incorporated an LLC (technically a Godo Kaisha, as it is incorporated in Japan).
It came as a surprise as it was perfectly fine by them for as long as it was a sole proprietorship. Which is why I strongly suspect the rejection came from an AI. At the time of rejection we were still using the Stripe account of the sole proprietorship, and surely a human would have noticed and blocked that account if it was actually running against Stripe ToS!?!
It doesn't make any sense to me. In any case, I know that financial regulations are stringent and payment providers are often required to leave us in the dark. But is it frustrating...
Now I'm stuck with PayPal for the time being. Unless they, too, decide that selling green tea is against their ToS.
By the way, does anyone know if it's ok to re-attempt to open an account with Stripe? Our offering has been evolving, so we might very well be within their ToS now?
> AIs and other opaque processes making decisions are incredibly frustrating.
Simply better messaging would fix most of this.
One approach is to be upfront and say:
Our automated AI systems have decided that it probably isn't in our best interests to do business with you. We can't elaborate on why, because that would give clues to spammers to help evade our detection of them. However, if you are willing to pay us $15, we will have a human look at your case and make the decision instead. You will be able to send evidence to the human (for example evidence of legitimate sales, a registered company, etc.). The human will be able to ask for further information, but will not be able to tell you the reason for any rejection. Your $15 will not be refunded if your case is rejected. Before you do, double check that your use isn't borderline on our rules and T&C's - human review isn't an opportunity to push the boundaries of what is allowed.
then some business-school asshat will say "wait! I have the most disruptive idea! let's tweak the algo to get 10% more rejections because analytics show 5% of rejectees result in a conversion!" As he proceeds to get a bonus for momentarily boosting revenue while permanently enshittifying the product.
My company can't use Stripe to collect payments for our Continuing Medical Education conferences because, as a psychiatry company, we offer in-office-only esketamine therapy to a small subset of patients. No amount of explaining that we do not SELL eskatamine or ANY pharmaceuticals seems to get through to the supposed humans I've talked to at stripe. Nowhere in anything we publish including the website says we sell anything, because we don't. We don't process (and won't/can't) process patient payment though Stripe, because we accept insurance as we are a psychiatry group. All we get back are two types of responses: "We've suspended your account because you sell pharma" and "We're investigating the information you have given us". Mind you all we've been ABLE to give them is email replies because our offers of meetings and documentation to prove what we're saying go ignored.
Stripe is great, but something is going terribly wrong inside the company when it comes to customer vetting.
I can see where Stripe is coming from. At best this lands in some kind of gray area, but really it seems pretty clear why this is getting a no go.
On an extremely fundamental level, people give you money and you give them drugs. The fact you only administer the drugs in your office doesn't change the basic premise.
I don't know your specifics, such as if the patient's money goes straight to a pharmacy to pay for the drugs which are then shipped to you for in-office administration (which would probably impossible to explain anyway). However, even if that's the case people are still giving you money (for appointments, treatment, etc.) and ending up with drugs in their body.
You'd likely have to split the medical CE conferences off from your practice and run it as a seperate entity before Stripe would be willing to take payment.
We tried, to no avail, just got the same unhelpful answer each time. Although we didn’t push too much, as we were still relying on the sole proprietorship account, and didn’t want to jeopardize that!
Now that we’re 90% done with the migration to PayPal, I think we can try again…
The so called AI is very infuriating indeed. I was once banned by such dumb script because I was trying to call my cell operator via a regular telephone number and there were some technical issues either at my phone or at the base station, and calls were dropping. After about 5 tries I got a robot answer that I'm now banned from ever using support line for "abuse". And to appeal the ban, I would have to use a support line. Ok, I thought, I'll try robochat at the website. No luck - banned there too.
So I have resorted to searching the Facebook page and hoping that a human moderates it. Thankfully I got to a human support and they revoked this ban.
I've stopped posting anything on the Facebook a few years ago and keep the account alive just to have a backup contact option with dumb corpos.
What's wrong with your existing payment processors? This type of existential thing seems to come up over and over again with stripe, I'm surprised people still want to use them.
People want to use it because its very easy to sign up and very easy to use. Even compared to Paypal, its much easier in my experience. And the vast majority of users don't have any issues at all.
What I find most frustrating about this is that they straight up lie to you. They don't actually review your account and the ban isn't permanent. The minute people in this situation get through to a human (one with any autonomy, anyway) the ban gets reversed. So it's all about the cost and efficiency and convenience for Facebook/Meta.
I definitely think there need to be laws made about this. Because it's really clear that the market has no power to correct this behaviour - yet it can be utterly devastating to individuals when they have life long bans put in place on what are (whether you like it or not) pretty much essential online services.
There exist laws about this, which provide a market solution. It's just that everybody has their fingers in their ears yelling about how big tech aren't monopolies so the existing laws aren't being applied.
In an actual competetive environment, you would go to the next door buisness that didn't arbritrarily ban you/has actual people in their customer service...
But why is it not in Meta's interest to have a human take a quick look at this? Or ask the user to pay for a human review? How much can it cost, in human time? $20?
They don't have enough human moderators to be able to handle it, and they don't want to hire more. You can't hire a moderator just for an hour - you'd have to recruit them, train them, continuously employ them, manage them, provide them benefits, etc. - it'd make an amortized per-review cost likely way higher than $20.
* The Scunthorpe problem, AOL didn't let users set their hometown to Scunthorpe because of the "cunt" substring. This is a classic.
* Amazon banning sales of Guns N' Roses merch (because guns).
* People named Miranda having issues with bank transfers (because of the substring "Iran").
* Some games displaying the nickname "Nasser" as "N*er", suggesting that it is the n word.
Alexa not being able to say "pussycat" properly, which is an issue to this day.
* Parental control software that filtered on "anal" in URLs, which also affected "analysis".
* Another parental control software that removed any file with "sex" in the name. This also included "sysext". I know of a school whose entire computer room got bricked because of that bug, the IT person decided it would be a good idea to update all the computers to Windows 10 at once, without checking for software incompatibilities first.
* The British politician Dominic Cummings not being able to set up a Twitter account.
My read of it is that's exactly what they were saying. The software replaced "ass" with "*", which made it even worse because now it read like the software was correcting the n-word (and in turn suggesting the person was using the n word).
This reminds me of middle school while I was temporarily in a country with heavily censored internet, where, I couldn't even search for info about sexual reproduction in plants for an assignment without a VPN because they filtered on "sex".
So while I couldn't find info on actually innocent topics, it had been trivial to play around with search terms to find porn that the filter couldn't catch.
> temporarily in a country with heavily censored internet
So, the USA huh?
For "children" (read: under 18), schools are some of the most despotic, censorious, and anti-democratic institution we have, and we send young humans from 6-18 through this meatgrinder.
During health class, our teacher was unable to see anything about "breast cancer", even though various cancers were in the discussion topic. Didn't matter. "Breast" is a evil horrible sexualized word that we must never allow anybody ever see... Even though you're legal to consent to sex at 16 in most states, and 17 in the rest.
Even the community college had the same blocks on it as the local high school. By definition, people who attend are 18+ , and yet treated like 3rd graders.
It may not be as censored on your own connection (cell, cable, fiber, etc), but this country absolutely does heavily censor, and they censor humans when most vulnerable and under state compulsion to attend.
I still remember when Google, in a fit of one of the many moral panics, decided to ban everything related to guns from their shopping site. Which, of course, lead to banning of all the Burgundy wines, because they have a "gun" inside! Poor fellas were so hasty to implement it they couldn't even do the filter right.
I think that's why an online card game I play won't let me name my tarot-themed deck "Tarot", it changes it to "*ot". Not sure what it means but I guess their censoring mechanism taught me a new slur?
Pretty sure I ran into this problem trying to use my American credit card to pay a bill from the University of Sussex in the UK. (Note the substring of "sex".) No matter how many humans we got involved, every attempt would immediately lock down our credit card beyond the usual fraud alert locks that I can personally revoke by responding to a text message, and would require a manual unlocking by bank support staff. Finally we just tried the PayPal option and it worked fine.
In other "we can absolutely totally build ai, no really guys, this shit works" news: A friend is losing some weight via diet and exercise changes. We have a messages convo wherein I congratulated him and we discuss both our exercise routines.
One of Google's autosuggested response emojis was a kiss. This has happened repeatedly over the 6 months-long conversation.
In the early 2000s I couldn't download Winamp 3 at school because the URL contained the string 'mp3'. I'm sure the school did not care about the distinction.
I've multiple times tried to open a seller account on Ebay. Two minutes after the account was created and confirmed it got blocked due to me having violated their terms of services. At this point I had not done ANYTHING with the account yet, not even putting in a profile picture.
And the rest went the same as with you: Appeal process is done by a bot who only knows one answer: "We are right, you violated our ToS, your account is blocked, this decision is final".
Oh how I miss the times when an organization for which you potentially are making a lot of money was still willing to hire a human for customer support...
And of course in these days, where your whole business may depend on being able to advertise or sell on the small number of platforms, a bot may decide to kill my business and company at any time. It's scary.
I'm banned for life from all Facebook properties (except for WhatsApp) because of my surname. They believe it's fake (or too controversial) and a copy of my passport hasn't changed their minds.
I used to have a Facebook account ~13 years ago before their real name policy came into force, but I closed it because I found it awkward to refuse or accept so many contact requests by people from my distant past.
Now unfortunately I can't access the Facebook group that is the only source of information for important things going on in the building I live in. My wife does have access though, so it's not a big deal.
It seems the only way for me to get around Facebook's real name policy is to use a fake name. But I guess that wouldn't allow me to advertise on Facebook or use any of their developer related stuff.
In my experience, once you close your facebook account, good luck re-opening one. The only way I was able to do it was getting hired by Meta (then facebook) because they required me to have one in order to on-board.
That is particularly egregious because there are plenty of people running around FB with obviously fake names. In a couple of the picture-oriented groups I'm in, I'd say more than half of the posts are from people with fake and sometimes even offensive names, but nothing ever gets done despite what I know to be multiple reports. Even a bad policy can usually be made worse by enforcing it inconsistently.
Assuming that they don't dox themselves some examples that would trigger a filter:
- Phillip K. Dick, etc., So like a body part or cuss word, or even an ethnic slur, for instance in Papua New Guinea a hundred or so people have the surname “Wop” which might trigger an algorithm. “Fokker” is a Dutch example of this, refers to someone who was an expert in getting animals to breed—possibly some other culture has a variant like Fucker that's the same.
- The name could be identical to a celebrity, for instance a Malcolm X born to Arthur and Jenny X having nothing to do with the more famous person of that name. Then you would still say “it’s because of my surname.” Alternatively the celebrity could be notorious e.g. if your name were “Jack Hittler.”
- The James Randi Educational Foundation once dismissed an application for a prize that they administer, because the person did not appear to be using their real name. This had to be reopened when notarized copies of a court order changing the name were submitted, confirming that he had indeed changed his first name to Prophet and his surname to Yahweh. So if that guy has kids with more normal first names...
> Meta won’t be seeing any of my money, whereas companies like Google, who seem to employ at least some humans in their advertising department — will
That is a convoluted way to say you'll do business with companies that accept your business. Google is known for fully automated bans on your full account, not just the advertising part.
My immediate thought when speed reading the blog post was "just create an advertisement company, have yourself as a client and call it a day". Sometimes we gotta fight stupid with stupid, you know?
Interesting point. Automated bans with no recourse will mostly deter legitimate businesses. And there is a finite supply of legitimate businesses, but infinite fake ones.
I know that Google has all sorts of automated things in place, and that people have gotten frustrated with them in the past. That said, my friends who have been stymied by Google's automated systems have always eventually managed to find a human who could fix things. Meta doesn't seem to have any such offer for customers. And even three employees, who agreed that the situation was absurd and could vouch for me, didn't manage to reverse the decision.
I once had a Google adwords campaign shut becuase I was advertising software for the Mac. The software only worked on a mac but I wasn't allowed to use Apple, Mac etc. as Google said they were trademarks. To be clear here, these words were used in context, eg. "Widget counting software for the Mac" not "Genuine Apple Software".
ISTR I appealed, and was turned down with no recourse. There was basically nothing I could do to advertise software just for the Mac.
I've had similar problems with Fabebook/Meta before and eventually gave up on buying ad services from them.
One solution to this problem would be a law that 1. mandates to give users the information when they are dealing with an algorithm only, and 2. forces companies to make it possible to contact humans within a specific time frame. I very much hope at least the EU will enact such a law soon.
Apart from monumental wastes of time with AI interactions for which no one is compensated, the thing that worries me most is that either AI "customer support" cannot do anything of substance anyway or the AI is given the power to make substantial changes to user settings, billing, etc. The first option means the company is bullshitting their customers, the second option can have very undesirable consequences.
> forces companies to make it possible to contact humans within a specific time frame.
Some EU countries do require this and it works a charm. However it doesnt seem to affect meta. That company, along with google and amazon and any other company that only relies on ai for customer support should fined into bankruptcy.
It's the kind of thing I thought had been handled way back in the 90s when AOL was rightly ridiculed from stopping people from entering the town Scunthorpe as their location because of a certain substring it contained.
Extrapolate this type of "unfortunate incident" to medicine, finance, insurance, policing, education, corporate hiring etc. where automation and "AI" is supposed to slash costs and improve efficiency and ponder what kind of society we are heading towards and why.
There is nothing wrong with using technology to empower humans (in this case Meta employees) to be more efficient in processing information. But removing the human from the loop in matters that affect other humans (whether in small or major ways) is a disastrous direction.
Coincidentally on the frontpage of the Financial Times today: "Meta’s Yann LeCun argues against premature AI regulation".
Readit News
It came as a surprise as it was perfectly fine by them for as long as it was a sole proprietorship. Which is why I strongly suspect the rejection came from an AI. At the time of rejection we were still using the Stripe account of the sole proprietorship, and surely a human would have noticed and blocked that account if it was actually running against Stripe ToS!?!
It doesn't make any sense to me. In any case, I know that financial regulations are stringent and payment providers are often required to leave us in the dark. But is it frustrating...
Now I'm stuck with PayPal for the time being. Unless they, too, decide that selling green tea is against their ToS.
By the way, does anyone know if it's ok to re-attempt to open an account with Stripe? Our offering has been evolving, so we might very well be within their ToS now?
[0] https://tomotcha.com
EDIT: wording
Simply better messaging would fix most of this.
One approach is to be upfront and say:
Our automated AI systems have decided that it probably isn't in our best interests to do business with you. We can't elaborate on why, because that would give clues to spammers to help evade our detection of them. However, if you are willing to pay us $15, we will have a human look at your case and make the decision instead. You will be able to send evidence to the human (for example evidence of legitimate sales, a registered company, etc.). The human will be able to ask for further information, but will not be able to tell you the reason for any rejection. Your $15 will not be refunded if your case is rejected. Before you do, double check that your use isn't borderline on our rules and T&C's - human review isn't an opportunity to push the boundaries of what is allowed.
then some business-school asshat will say "wait! I have the most disruptive idea! let's tweak the algo to get 10% more rejections because analytics show 5% of rejectees result in a conversion!" As he proceeds to get a bonus for momentarily boosting revenue while permanently enshittifying the product.
Stripe is great, but something is going terribly wrong inside the company when it comes to customer vetting.
On an extremely fundamental level, people give you money and you give them drugs. The fact you only administer the drugs in your office doesn't change the basic premise.
I don't know your specifics, such as if the patient's money goes straight to a pharmacy to pay for the drugs which are then shipped to you for in-office administration (which would probably impossible to explain anyway). However, even if that's the case people are still giving you money (for appointments, treatment, etc.) and ending up with drugs in their body.
You'd likely have to split the medical CE conferences off from your practice and run it as a seperate entity before Stripe would be willing to take payment.
Apparently Stripe has great difficulty with non-Japanese people, living in Japan, owning a company overseas.
Now that we’re 90% done with the migration to PayPal, I think we can try again…
I've stopped posting anything on the Facebook a few years ago and keep the account alive just to have a backup contact option with dumb corpos.
An LLC can’t be incorporated.
If you say you had an Inc when you had an LLC is a good cause of rejection.
Although in this case it’s neither an Inc. nor an LLC: it’s a 合同会社. That’s what I wrote on the form, obviously.
I definitely think there need to be laws made about this. Because it's really clear that the market has no power to correct this behaviour - yet it can be utterly devastating to individuals when they have life long bans put in place on what are (whether you like it or not) pretty much essential online services.
In an actual competetive environment, you would go to the next door buisness that didn't arbritrarily ban you/has actual people in their customer service...
* The Scunthorpe problem, AOL didn't let users set their hometown to Scunthorpe because of the "cunt" substring. This is a classic.
* Amazon banning sales of Guns N' Roses merch (because guns).
* People named Miranda having issues with bank transfers (because of the substring "Iran").
* Some games displaying the nickname "Nasser" as "N*er", suggesting that it is the n word.
Alexa not being able to say "pussycat" properly, which is an issue to this day.* Parental control software that filtered on "anal" in URLs, which also affected "analysis".
* Another parental control software that removed any file with "sex" in the name. This also included "sysext". I know of a school whose entire computer room got bricked because of that bug, the IT person decided it would be a good idea to update all the computers to Windows 10 at once, without checking for software incompatibilities first.
* The British politician Dominic Cummings not being able to set up a Twitter account.
Clbuttic!
This actually looks like it removes "ass" and not related to n-word? Nevertheless, rather assumptious!
Edit: Oh I misread the post. The new name is even worse... That is rather hilarious and sad at the same time :/
So while I couldn't find info on actually innocent topics, it had been trivial to play around with search terms to find porn that the filter couldn't catch.
So, the USA huh?
For "children" (read: under 18), schools are some of the most despotic, censorious, and anti-democratic institution we have, and we send young humans from 6-18 through this meatgrinder.
During health class, our teacher was unable to see anything about "breast cancer", even though various cancers were in the discussion topic. Didn't matter. "Breast" is a evil horrible sexualized word that we must never allow anybody ever see... Even though you're legal to consent to sex at 16 in most states, and 17 in the rest.
Even the community college had the same blocks on it as the local high school. By definition, people who attend are 18+ , and yet treated like 3rd graders.
It may not be as censored on your own connection (cell, cable, fiber, etc), but this country absolutely does heavily censor, and they censor humans when most vulnerable and under state compulsion to attend.
Also I utterly hate all those online services that require at least 6 letters in nickname...
If you insist on searching for "kinky kids" (using quotes) you will find stuff that is closer to what the filter warns you about, still SFW though.
One of Google's autosuggested response emojis was a kiss. This has happened repeatedly over the 6 months-long conversation.
Guy Carbagiale Fuck wasn't allowed to have his name on his jersey.
Deleted Comment
There's a lot of shitakemushrooms going on.
And the rest went the same as with you: Appeal process is done by a bot who only knows one answer: "We are right, you violated our ToS, your account is blocked, this decision is final".
Oh how I miss the times when an organization for which you potentially are making a lot of money was still willing to hire a human for customer support...
And of course in these days, where your whole business may depend on being able to advertise or sell on the small number of platforms, a bot may decide to kill my business and company at any time. It's scary.
I used to have a Facebook account ~13 years ago before their real name policy came into force, but I closed it because I found it awkward to refuse or accept so many contact requests by people from my distant past.
Now unfortunately I can't access the Facebook group that is the only source of information for important things going on in the building I live in. My wife does have access though, so it's not a big deal.
It seems the only way for me to get around Facebook's real name policy is to use a fake name. But I guess that wouldn't allow me to advertise on Facebook or use any of their developer related stuff.
Let me go see what jobs they're hiring for...
- Phillip K. Dick, etc., So like a body part or cuss word, or even an ethnic slur, for instance in Papua New Guinea a hundred or so people have the surname “Wop” which might trigger an algorithm. “Fokker” is a Dutch example of this, refers to someone who was an expert in getting animals to breed—possibly some other culture has a variant like Fucker that's the same.
- The name could be identical to a celebrity, for instance a Malcolm X born to Arthur and Jenny X having nothing to do with the more famous person of that name. Then you would still say “it’s because of my surname.” Alternatively the celebrity could be notorious e.g. if your name were “Jack Hittler.”
- The James Randi Educational Foundation once dismissed an application for a prize that they administer, because the person did not appear to be using their real name. This had to be reopened when notarized copies of a court order changing the name were submitted, confirming that he had indeed changed his first name to Prophet and his surname to Yahweh. So if that guy has kids with more normal first names...
That is a convoluted way to say you'll do business with companies that accept your business. Google is known for fully automated bans on your full account, not just the advertising part.
I haven't thought about this before, but it could explain some of the degeneration of ads over time I have been noticing.
I know that Google has all sorts of automated things in place, and that people have gotten frustrated with them in the past. That said, my friends who have been stymied by Google's automated systems have always eventually managed to find a human who could fix things. Meta doesn't seem to have any such offer for customers. And even three employees, who agreed that the situation was absurd and could vouch for me, didn't manage to reverse the decision.
ISTR I appealed, and was turned down with no recourse. There was basically nothing I could do to advertise software just for the Mac.
One solution to this problem would be a law that 1. mandates to give users the information when they are dealing with an algorithm only, and 2. forces companies to make it possible to contact humans within a specific time frame. I very much hope at least the EU will enact such a law soon.
Apart from monumental wastes of time with AI interactions for which no one is compensated, the thing that worries me most is that either AI "customer support" cannot do anything of substance anyway or the AI is given the power to make substantial changes to user settings, billing, etc. The first option means the company is bullshitting their customers, the second option can have very undesirable consequences.
Some EU countries do require this and it works a charm. However it doesnt seem to affect meta. That company, along with google and amazon and any other company that only relies on ai for customer support should fined into bankruptcy.
More specifically, it'd need to make it possible to contact humans with the authority to overturn the algorithmic decision.
There is nothing wrong with using technology to empower humans (in this case Meta employees) to be more efficient in processing information. But removing the human from the loop in matters that affect other humans (whether in small or major ways) is a disastrous direction.
Coincidentally on the frontpage of the Financial Times today: "Meta’s Yann LeCun argues against premature AI regulation".
Let AI moderation commit genocide once, shame on Facebook. Let AI moderation continue to go unregulated, and...
https://erinkissane.com/meta-in-myanmar-part-i-the-setup