RSS LJ

April 13, 2009

Some thoughts on #amazonfail ()

by fluffy at 12:33 AM
So, this Easter weekend, a bit of a storm hit the Internet when word got out that Amazon was removing GLBT materials from search results and sales ranks. The Internet (well, Twitter and dozens of weblogs) assumed this was some sort of puritanical hokum from Amazon, Amazon's official response was that it was a glitch, and the meta-response was that Amazon has purposefully done this and is just trying to "cover its ass" and that there's no way such a thing could happen as a glitch.

As a former Amazon employee (who left the company voluntarily on very good terms two years ago), I should share some of my thoughts.

4:15 PM Read this. Mike Daisey's explanation makes perfect sense. Of course my main point is that you should put the damn pitchforks down because the actual situation is much more complicated than what the Internet masses want to believe. Whatever the root cause was, I am positive it wasn't the result of directed malice from anyone at Amazon.

First off, I never had to deal with the catalog system directly, but my understanding is that it's actually a fairly byzantine system which has a lot of fiddly bits. It doesn't update instantaneously (there are several layers of caching which go on — it's a big database, so this is necessary), and its category/tag/etc. structure is pretty ad-hoc. Pushing out code and content updates takes a while. You-the-public see it as a single site but every single page request touches hundreds of servers on the backend.

I don't know how the "safe" search is implemented, and I don't know what criteria are used to exclude an item from search or salesrank. Here is what I do know about the situation:

  • Amazon has an extremely active internal GLBT community (called GLAmazon) which is very vocal, large, and would never, ever allow this action to knowingly occur, purposeful or not.
  • The Amazon catalog is huge. I can't emphasize how big it is. There are literally millions of books. There is absolutely no way they could be individually tagged and excluded from search results.
  • Defining even something as simple as "adult" content is extremely difficult. There are false positives. There are false negatives. Things have to happen based on keywords and data which comes from the publishers and from user-submitted tags. Anywhere along this chain it is quite conceivable that something that correlates with "GLBT material" ended up getting on the adult filter list.
  • Amazon is based in Seattle, which has one of the highest GLBT per capita in the US, if not the world. Many of its core engineers are openly gay. There is absolutely no reason to believe that this would be a conscious decision made by anyone high up in the organization. If it was a conscious decision, it was probably by a single person whose contribution wasn't being appropriately audited.
  • The firestorm hit on a day when most people are busy with their families, the flames being fanned by people who see Amazon as idiots and/or bigots. Several people are trying to capitalize on this by opening CafePress stores protesting Amazon which is, frankly, sickening.
  • It is completely unreasonable to think that an error of this scale could be fixed or even diagnosed during a few normal work hours, much less on a holiday, but because an hour feels like a year on the Internet, a lack of instantaneous Twitter-style responses and apologies is seen as arrogance, ignorance, or malice. But believe it or not, Amazon attempts to be a responsible company, and part of that means not shooting from the hip and spamming the Internet with off-the-cuff responses to every little uninformed gripe.
  • I mean, seriously, what are they supposed to do, personally reply to every single twitter post with an instantaneous, detailed apology?
The sad reality is that complicated issues take more time than it takes for a bunch of babbling idiots to spout off their theories about what's going on, and completely judge/jury/execute a company based on a complete lack of understanding. How long did it take for people to start calling for a boycott, or to start calling in to yell at already-overworked CSRs, on a day when most people are actually, you know, not at work?

Obviously, if there is a problem with a company's attitude, it must be addressed, but holy cow, try to find out if there's actually a problem first. Give them a chance to at least fix it.

Comments

#11921 04/13/2009 05:41 am
The sad reality is that I saw "#amazonfail" and thought it was an IRC channel.
#11922 04/13/2009 07:50 am
Twitter tags are the new pointless Internet babble.
#11923 04/13/2009 01:58 pm
And I imagine you've already seen this, but I'm linking it anyway:

http://community.livejournal.com/brutal_honesty/3168992.html
#11924 04/13/2009 02:02 pm
Yes, it seems plausible. A lot of people claim to have "debunked" it based on that functionality no longer being available, but that's rather like debunking that George W. Bush was President based on the fact he no longer is.
#11925 04/13/2009 02:17 pm
Moreover, the whole tell-all opens with
Amazon removed its customer-based reporting of adult books yesterday. I guess my game is up!

so yeah.

And anyway, even if this guy's lying about what happened at Amazon, there's a real issue: any system which automatically flags content based on complaints is broken.
#11926 04/13/2009 04:06 pm
This is the best collection of explanations I've seen:

http://www.lilithsaintcrow.com/journal/2009/04/idosyncratic-code-amazonfail/

Of course the Twitterstorm on #amazonfail continues unabated.
#11927 Anonymous 04/13/2009 04:14 pm
Excellent writeup. You would have been more persuasive if you had managed to explain Amazon's correspondance with authors indicating that this change was new policy. Correspondance which dates to February of this year. Despite the fact that this issue blew up over the weekend, the "censorship" began in February.

I don't think that Amazon targeted all of these books (perhaps they hadn't targeted Heather Has Two Mommies for example) but to say that this is entirely the act of rogue code leaves several questions unanswered.
#11928 04/13/2009 04:16 pm
The adult-content policy has been in place since February (at least), yes, but it seems that the CSR who originally responded to the author that started all this misunderstood what the author was asking. Amazon's first-level CSR stuff all happens via boilerplate "blurbs" which are essentially copy-pasted by trained monkeys.

This becomes especially apparent on HOLIDAY WEEKENDS when most people are AT HOME.

I also never said it was the result of "rogue code," I just said that it's perfectly conceivable that it would be, seeing as how Amazon's catalog system is extremely ad-hoc and flaky. Of course, that troll was taking credit for it which only helped to confuse the issue.

All I was trying to persuade people of is that the people at Amazon are fundamentally good folks who aren't out to discriminate, and that the issue is certainly much more complex than the people who are calling for blood realize. The knee-jerk torches-and-pitchforks response that the Internet community has had to this unfortunate situation is extremely frustrating.

Mike Daisey's explanation seems completely legitimate and believable, and that should be the end of the story (but of course it won't be).
#11929 Anonymous 04/13/2009 05:12 pm
>> Amazon has an extremely active internal GLBT community (called GLAmazon) which is very vocal, large, and would never, ever allow this action to knowingly occur, purposeful or not.

Perhaps some of these people will speak out in defense of Amazon then. That would probably be much more effective at damage-control within the small glbt community than Mr. Daisey's explanation.
#11930 04/13/2009 05:16 pm
Well, Amazon has a policy that employees aren't to talk about company-related stuff in public without clearance from the PR department etc. It's pretty corporate-standard, and generally good business sense. I'm sure lots of GLAmazonians are just burning up inside wanting to post to the web about this issue.

For all I know, several already have, but the links just haven't made the rounds yet. It's a lot easier for major celebrity bloggers' vile spewing to circulate than for much-later good-news sorts of things. It's just the way the Internet works.
#11935 Anonymous 04/16/2009 11:53 am
Me again...

I wonder what the GLAmazons think about searches for homosexuality STILL turning up anti-gay literature as first link. It's Thursday. I'm not a conspiracy theorist but its getting harder and harder to give amazon the benefit of the doubt on this one.
#11936 04/16/2009 12:02 pm
Yeah, I wonder that too.

Unfortunately, the search engine is at least as hacky as the salesrank stuff, and it takes a while for things to get updated. I think it takes a week or more for things to settle out, so it'll probably take at least that long for salesrank-weighted search results to return back to how they were before.
#11999 Anonymous 05/04/2009 04:55 pm
Me again. Just wondering if you had any thoughts on the subject of amazon site-wide searches for "homosexuality" still returning "pray yourself straight" books as top results.

I don't want to rob Amazon of any benefit of the doubt. But at what point can we agree that their "new" search is legitimately offensive?
#12000 05/04/2009 04:56 pm
I think their search engine has always been kind of crappy and fragile, and it's pretty likely that these results have ALWAYS been the case, it's just that nobody really noticed it until there was a reason to have it under the magnifying glass.