←back to thread

Stripe banned us for payment disputes but we never had a single dispute

(justuseapp.com)
693 points hienyimba | 9 comments | 14 Sep 21 10:51 UTC | HN request time: 0.636s | source | bottom
Show context
jsiepkes ◴[14 Sep 21 11:18 UTC] No.28522995[source]▶
I guess stripe wasn't kidding when they said they would disrupt online payments.

On a more serious note; How much further is society going to allow this kind of thing? Hiding behind templated e-mails without any explanation. Disrupting people's lives who become collateral damage with no way out.

replies(5): >>28523050 #>>28523128 #>>28523145 #>>28523324 #>>28524929 #
1. afarrell ◴[14 Sep 21 11:24 UTC] No.28523050[source]▶
For as long as it permits companies to hire fallible humans and to write machine learning models with false positive rates.
replies(1): >>28523139 #
2. nicoburns ◴[14 Sep 21 11:35 UTC] No.28523139[source]▶
The machine learning models with false positives aren't the problem. The lack of a timely appeals process that involves a human is.
replies(2): >>28523237 #>>28525092 #
3. jtbayly ◴[14 Sep 21 11:45 UTC] No.28523237[source]▶
But that’s why the prices are so low.

Society will put up with it as long as it works most of the time, because algorithms without humans are cheap.

replies(2): >>28523285 #>>28523296 #
4. syshum ◴[14 Sep 21 11:51 UTC] No.28523285{3}[source]▶
stripe's prices / fees are not really that much lower, they do have great profit margins though
5. nicoburns ◴[14 Sep 21 11:52 UTC] No.28523296{3}[source]▶
I'm not sure about that. The companies employing these kind of techniques are typically making huge profits. I suspect supply and demand would dictate that a bunch of the cost came out of those
6. naasking ◴[14 Sep 21 14:11 UTC] No.28525092[source]▶
I'm curious how many human reviews are triggered after ML flags a problem. If it's nearly 100%, why have the ML step at all?
replies(1): >>28525658 #
7. colinmhayes ◴[14 Sep 21 14:47 UTC] No.28525658{3}[source]▶
Because the algorithm only flags like less than 1% of users?
replies(2): >>28525685 #>>28525950 #
8. ◴[14 Sep 21 14:48 UTC] No.28525685{4}[source]▶
9. naasking ◴[14 Sep 21 15:04 UTC] No.28525950{4}[source]▶
Maybe I wasn't clear. I meant, why have the ML algorithm disable the account automatically if human review happens nearly 100% of the time, rather than simply have ML flag the account for human review, and let them decide whether to disable the account.