Arguing with Software
I self-publish a number of books on Amazon using KDP, Amazon’s self-publishing subsidiary. A little more than two years ago (I quote from an old blog post):
They sent me an email saying that, because I had multiple accounts and my accounts were connected with one that had violated their guidelines, they were terminating my account, would not permit me to open another, would not pay me royalties accumulated between the last payment and the date when the account was terminated. All of the books of mine that I had self-published vanished from Amazon.
I wrote multiple replies, pointing out that so far as I knew I had done none of those things and asking what connection they thought I had to what account. The only response I got was to be told that they had reviewed the case and my account was still suspended. All their messages were form letters. When I received two identical form letters responding to two quite different emails of mine, one sent three minutes after the other, both purportedly by the same person, I concluded that I was interacting with software, not humans.
Today I received an email telling me that my account had been restored. There was no explanation and the tone of it — "you must review your catalog and remove any other titles currently available for sale on Amazon that do not comply with the KDP Content Guidelines" — implied that I had done something wrong, although they were willing to forgive me if I behaved myself in the future.
It happened again a few months later, but that time the account was restored fairly quickly. Restoration both times was apparently due to readers of my blog intervening with Amazon-KDP on my behalf.
I was reminded of that incident a few days ago by a less extreme example of the same pattern. I had ordered a car from Uber to take me from my home in San Jose to San Francisco. According to the Uber program on my phone, the car, with a named driver, was coming shortly. It didn’t come. The program now showed the car on its way to San Francisco — without me. I messaged the driver, got back a two or three word response that she was coming. I waited a while longer; the program showed the car continuing towards San Francisco. I phoned the driver, got no answer. So I cancelled the ride and ordered another, which showed up and took me to my destination.
I received an email from Uber billing me $23.69 for the cancelled ride.
After a little searching on their web page I found where problems were supposed to be reported, reported what happened, got back:
I clicked on “Go to the conversation.”
I logged in for personalized support, repeated my story, got back:
Hi David,
Thank you for reaching out. I'm Pooja from Uber Support, and I'm here to assist you.
I'm sorry to hear about the inconvenience you experienced during your trip. It seems there was a mix-up, and someone else may have taken the ride you requested. I understand your concern about being charged for a trip you didn't take.
To make things right, I've fully refunded your trip price, and you should see the refund appear within 3-5 business days. An updated receipt has been emailed to you.
My conjecture is that the software automatically rejects complaints at the first round in the hope that you will then drop the subject, shifts the matter to a human if you don’t.
Both experiences reflected the same pattern — decisions made by software with complaints responded to by software pretending to be a human, resolved only when and if a real human somehow gets involved. In the Uber case the human apparently comes in automatically at the second round. With KDP I went through multiple rounds of complaints responded to by form letters; my guess is that if I had not reached people inside Amazon I would never have gotten the account back, would have had to shift my books to one of their competitors, something I was starting to do before KDP restored my account. The experience was like talking to what I thought was a human and discovering that it was a brick wall.
The natural reaction to both incidents is to be angry. That would be the appropriate response to a human being who claimed to be responding to a message he had not actually read, but software is not a human being to be angry at.
KDP is also not a human being. It is a firm that deals with a very large number of authors, some of whom probably do things that violate their rules such as plagiarism or copyright violation. It would be nice if they dealt with suspected violations by having a competent staff member spend the time to carefully check them but competent staff members are expensive. It may make more sense, from their standpoint, to use software to look for possible violations of their rules, perhaps in accounts reported to them by readers, even if the software doesn't do a very good job. My interaction with them lowers my opinion of KDP and Amazon I am not sure that, if I were running things, I could do better, nor whether it makes sense to react to a firm as if it were a person.
One response by a commenter on my blog:
You could do better. You could write a more honest form letter.
What about Uber? Their policy, if I correctly interpret it, is an automated version of a familiar tactic: The first time someone complains, ignore him; if he keeps complaining have someone look into his complaint.
It’s the squeaky wheel that gets the grease.
At least, when they did look into it and discovered that the fault was theirs, they had the grace to apologize. Amazon didn’t.
I still have not figured out what went wrong with my Uber ride. It is hard to imagine a software bug or a trick by a driver who did not want to take me to San Francisco that would show the car on its way to San Francisco when it wasn’t or any way the driver could have picked up the wrong passenger.
The series of blog posts on the KDP problem:
Why You Can't Get Most of My Books
Amazon has reinstated my account
Should I Be Mad at Amazon/KDP?
Amazon KDP Cancels my Account Again
Has Someone Discovered a New Way of Censoring Books?
Past posts, sorted by topic
A search bar for past posts and much of my other writing
The Uber thing happened to me too. What happened was that there was another person nearby also hailing an Uber and he took my car by mistake. The Uber driver is supposed to check ID but that check consists of asking, “Are you so and so?” Not much of a fail safe. But the Uber had my destination not his, so that trip terminated a few blocks away. I got billed anyway.
I’ve had the opposite experience. When I interact with a robo–customer-service rep, I find it easy to push it to its limits and figure out what it can or cannot do. However, I get frustrated when I’m transferred to a foreign call center where the human agent follows a script yet seems to have little real knowledge of the product or service I’m calling about. I have spend many aggravating hours on such calls.
I do believe AI will greatly improve the customer service experience in the coming years. Unfortunately, that will likely come at the expense of customer service representatives themselves, as many of those jobs could disappear.