AmazonFail Followup

AmazonFail seems to have been squared away reasonably well, with, as I understand it, most of the blame going to someone toggling something in a database somewhere and 50,000+ books losing their sales ranking all of a sudden. The company’s fixing that, which is nice, and I have to say I’m not entirely surprised to discover that it had more to do with a technical screwup than an otherwise generally tolerant corporation experiencing an atavistic twitch of homophobia or whatever.

As for people getting worked up and demanding explanations and threatening boycotts on an Easter Sunday, well, you know, look. People really do need to accept that while outrage happens at the speed of someone banging out 140 characters on Twitter, fixing corporate-level problems is going to be slower, especially on a weekend, especially on a holiday weekend, and especially if the people trying to respond to outrage want to actually find out what the hell happened, so as to possibly a) give a coherent explanation to people demanding the same, b) avoid having the same or similar screw-up from happening again. All of which is to say that people leaping straight to a boycott because a large corporation has not answered to their satisfaction questions on a complex and confusing issue late on the most holy day in Christendom may possibly have unrealistic expectations.

That said, I don’t think there’s any question that Twitter having a fit on the matter certainly gave Amazon a goose to address the issue, so I wouldn’t say the AmazonFail uproar was all bad. It wasn’t. If only there were a way to have to have thousands of people on Twitter go “Hey this looks bad; you might want to explain that” instead of FOAMY FOAMY FAIL FAIL BOYCOTT GAAAH. But people rarely freak out in a moderate sort of way.

This also brings up a point which I think it worth airing, which is that I and at least a couple of authors I know got e-mails about AmazonFail — not the “did you hear about this?” e-mails, which are fine, but the “you need to speak out about this now” ones, which are pretty much not. First, of course, I don’t need to do anything about anything, other than what I decide I need to do. Random people e-mailing me about what I need to do have a grave misunderstanding about their powers of persuasion regarding me. Second, even when I’m inclined to do something, at this point, what I’m inclined to do first is make sure I have an understanding of what’s actually going on, and to use my own judgment regarding whether I need to know more before making a substantive comment.

In the case of the Amazon thing, for example, I recognized that something was going down on a holiday weekend, and that I wanted to hear what the Amazon brass had to say about it. Which is essentially what I said in an earlier post on the matter. As it turns out, the situation seems largely what I expected it to be, i.e., technical foul-up, and since Amazon seems to be busily rectifying it, the various shifting excuses laid out for the sake of PR are largely immaterial. In short, I’m happy to have been a moderate on this one.

114 thoughts on “AmazonFail Followup

  1. I understand and accept that this is the result of a technical glitch. What I think most people are still worked up about is that there was some sort of categorization going on that seems badly skewed, ideologically.

    Somewhere in the Amazon database a number of, otherwise innocuous, books got labeled as adult because they dealt with gender and sexuality in a non-puritanical fashion. The database error that snatched away all their sales rankings simply brought this to light.

    The question of why so many works were thus classified remains.

  2. One thing this really points out, though, is how managerial idiocy can get in the way of doing anything.

    Didn’t any manager look at a calendar before giving the alleged (mis)coder a task potentially involving a change to customer interfaces? There were three disqualifying reasons that the whole database should have been locked since not later than midday on Wednesday of last week:

    * Passover, starting on Thursday evening
    * The upcoming Good Friday/Easter weekend, which for many is a three- or four-day period away from work… particularly in Europe where the error allegedly occurred
    * April 15, given the tendency of Americans to simultaneously panic looking for tax advice and go on a comfort-spending spree now that the taxes are done

    A competent managerial structure would have taken each of these into consideration. All three are overwhelming.

    So, even though this is going to be treated — eventually — as a “computer failure”, it’s more a managerial failure than anything else… leaving aside the dubious ethics of screening information from a search function in the first place. Mr Bezos et al. are answerable for this one; it’s not the fault of the drone who pushed the button that did not have a big “Do Not Push This Button” sign on it (even though he/she is the one who’s going to be blamed and fired).

  3. Well, as your President pointed out recently, it’s good to wait to speak until you actually know what the hell you’re talking about.

  4. As a proud, fairly activist lesbian, I just could not get worked up about this. Perhaps it’s due to the fact I’m laid up in bed being all sickish and apathetic, I don’t know. What I do know is that I heard about it, read some frothy LJ posts, decided upon my own mini-boycott-of-the-moment by refusing to hit the Amazon site, and figured it would either be explained and die down, or would grow into a legitimate cause to get behind.

    There’s been too much fail in 2009 already, in all arenas. Blah.

  5. I might have bought the glitch excuse if I didn’t actually know several people who received their boilerplate letter discussing their new corporate policy toward adult material. It doesn’t wash. Authors and publishers deserve the truth on this one. And Amazon.com really needs a new corporate communications director, becaue this whole thing was extreme #PRfail.

  6. Moral of the story (A.K.A. Hanlon’s Razor): Never attribute to malice what can be explained by stupidity.

  7. From what I’ve been reading the glitch was not the categorization but the breadth of it which led to people actually finding out about it. That categorization is still going on and still skewing in favor of heterocentric adult material (Playboy) but against the homocentric (Ellen Degeneres biography). It still smells strongly fishy to me.

  8. Re: “Didn’t any manager look at a calendar before giving the alleged (mis)coder a task potentially involving a change to customer interfaces? There were three disqualifying reasons that the whole database should have been locked since not later than midday on Wednesday of last week…”

    Actually, the timing of the problem might have been a feature, not a bug (not the problem itself, of course.)

    I have a friend who works in IT at a retailer with a major world-wide 365×24 web presence (not Amazon) and she’s frequently called upon to do installs at 2AM Sunday morning because although the web never sleeps, some times are more sluggish than others. I wouldn’t be at all surprised to find out that Amazon tries to time upgrades to weekends and / or holidays.

  9. Well, I think it was intentional. I believe Amazon did this on purpose, hoping to slip it under the radar and got busted. That said, they are now doing the right thing and hopefully they won’t try something like this in the future. I like Amazon and I like shopping there.

  10. I’ve been of the same mind as you, Scalzi, on most of this. But I would like to see a couple more things from Amazon — an explanation of how things are categorized as ‘adult’, e.g., are heterosexual books with fairly graphic sex (and this would include a hell of a lot of romance novels, usually considered ‘safe’) included too? And also, it would be nice to see an apology of sorts.

  11. I agree with 1, 6, 10–this was not a technical glitch. The costumer service rep responses, the letters to authors, and the subset of entries affected all point to a change that was thought-through and deliberately made (I’ve worked in systems for twenty years). Bezos has conservative streak, and it wouldn’t shock me if that has extended to company policy. I’m not gay, and it annoys me; if I were gay, I think I’d be getting a little frothy.

  12. And when was the last time anybody, let alone the largest online retailer, was successful slipping something of this magnitude “under the radar.” There are a lot of organizations *cough RIAA / Republican Party / Associated Press cough* that I could see trying something this stupid and being surprised when they were caught. Amazon is not that stupid.

  13. Good people of the Internet, one thing to keep in mind is that large sites deal with lots of data. Lots of it. And if a change is made to 60,000 individual items, it might very well not show up on a site for several days. Unfortunately, that means that reverting those very same changes takes time. Even as the villagers gather pitchforks, engineers are likely working on getting the catalogue data from server A replicated to the database on server B so that the sales ranking algorithm (which takes six hours to run, don’t you know) can populate the info that gets stored in a giant cache so that shoppers can load it, etc. In other words, this may not be a lesson in the power of Twitter so much as it is a lesson in how hard it is to change data on a site that’s designed to avoid reloading data unless absolutely necessary.

    Also, when you categorize books as dealing with sexuality (something you probably want to do so you can link them together for customers and/or provide reasonable search hierarchies), and when you categorize books as dealing with gender issues (see above), you are also risking shutting off access to, or changing the characteristics of, all of these items at once. I don’t think the solution is getting rid of the “sexuality” tag on Natalie Angiers’ work, for example.

  14. I think the biggest reason people got all rabid about it is that with the economy in the tank, people are looking for a corporate entity to take their rage out on. Any company that stumbles right now is risking a good ol’ lynching.

  15. Right now, if you search “homosexuality” on Amazon, the top result is “A Parent’s Guide to Preventing Homosexuality.” However, if you search “gay,” it returns all kinds of gay erotica. Probably because people writing and looking for gay-friendly literature tend to prefer the term “gay,” while homophobes tend to prefer the much more clinical-sounding “homosexuality.” In other words, the current search algorithm seems to work correctly to find the right stuff, depending on what the searcher is looking for.

    Now, maybe Amazon ought to include the more gay-friendly literature higher up the search results for “homosexuality,” but maybe they should also rank some of the “homosexuality” results higher up in the results for “gay,” for readers who might be trying to combat homophobic literature. Or maybe Amazon should go even further and bury the homophobic literature. But I’m not sure that I want a corporation telling me what I “should” read at all; I prefer the way Amazon currently works, suggesting “other people who read this also read that.”

    Thousands of people piled on Amazon, and they’ve responded. Sure, keep a watchful eye on Amazon in the future, but let’s move on.

  16. ADM@11, I read at one place or another that the romance section was hit pretty hard with this as well. Not being familiar with the genre I cannot confirm this, but I know that it was reported it happened.

    For all you people who still think this was an intentional stab at the gay community – have none of you ever had jobs in medium to large corporations? This clearly had the earmarks from the beginning as just a screwup, from two different groups not communicating. It’s almost certainly a combination of something like this:

    Group A – “We get complaints because things like sex toys show up on fairly innocuous searches. Let’s find a way to fix that” resulting in the creation of an ‘adult’ tag that filters stuff from searches.

    Group (or even person) B – “Your task is to improve our searches by applying tags to items. Here’s the list of tags we support, exported from the database, and here’s a bunch of lists from various places on items.”

    In a situation like this, Group B is going to have no idea that a tag has side effects, and testing is extremely unlikely to catch the fact that one tag behaves differently.

    Now, I have no inside info that this is, in fact, what happened, but it’s plausible.

  17. Collen, et al:

    The policy itself wasn’t the error. What got categorized under the policy was a error.

    The CSRs appear to have heard the description of the issue / looked the book up internally, matched it up to the policy and replied with that. The wrong books were being included in the policy, but there was no way for the CSRs to know that.

    If former Amazon employee Mike Daisy’s “internal sources” are accurate, it’s because a french staffer misunderstood the categories (due partially to translation issues); see here. As I can’t see how delisting so many gay/lesbian books could possibly have a positive effect, I can’t see any business reason for doing this on purpose.

  18. The problem with the alternately overinclusive/underinclusive flagging of “adult” materials sounds a lot like the problem with website filtering services like NetNanny. We’ve all heard of websites on topics like breast cancer being filtered out for the kids in public libraries and such.

  19. I had to chuckle at John’s comment re emails that tell him he has to do something.

    I also think his first reaction to anything or anyone trying to force him to do anything (except for his wife and daughter) would be along the lines of “Yeah? You and what army?”

  20. John, I appreciate your desire to gather data to come an informed position but given the information I’ve looked at I’m not convinced that this was a one-off or even a glitch.

    I noticed the de-ranking on certain titles long before this blow-up happened. Just not enough for it to raise a red flag.

    If not for the fact that individuals began digging deeper, tracking and speaking out, I have to wonder whether Amazon would’ve reinstated the rankings and main page search functionality.

    Yes, I know glitch. Sorry. Unlike you, at the moment I’m very on-board with the argument Jane at DearAuthor makes. I believe we need to dig deeper and ask a lot more questions. And not just of Amazon.

    —-

    Thomas #1, based on a sampling that DearAuthor.com was able to document prior to Amazon’s reversal, any books in their small sampling that had a subject category listing** beginning with

    Books > Subjects > Literature & Fiction > Erotica
    and
    Books > Subjects > Gay & Lesbian

    were filtered.

    The categories were not title specific they were ISBN specific leading me to believe that the filter ultimately used might have been originally provided by either the publisher or a third-party meta-data provider. Which begs the question of whether we’d ever want something like subjective category designation to be used to make filter policy because subject cateogization isn’t consistent. Period.

    Technological foul-up or not. We’ve put a lot of power in Amazon’s hands and we need to be asking a lot more questions (not just of Amazon) beyond just whether or not the rankings and search functionality have been restored.

    ** There were two exceptions in the small sampling and those ISBNs had Kindle categories as part of their designations. Those exceptions might be explained if the Kindle designation had a higher designation than any subsequent filters. Another subject category that might have also been involved was Books > Subjects > Health, Mind & Body > Sex but the sample did not contain enough books with that designation to provide a conclusion.

  21. What bothers me more than anything is that Amazon had this de-ranking code in place at all. Is it right that they even consider such a thing?

    Sure, they have the right to do so, but I think we have the right to be furious with them for doing it. This isn’t a coding mistake, this isn’t a middle-management mistake, this is a philosophical mistake.

  22. I think a lot of the vitrol would have calmed had they said “hey, we’ve got a corruption someplace in the database, we’re working on it!” rather than “screw you you stoopid unwashed internet masses, it’s a glitch and that’s all we’re gonna tell you”

  23. This isn’t going to make me the most popular kid in class, but I think a lot more people should feel like they have egg on their faces.

    Amazon is a big company that, by all appearances is run by moderately smart people without a history of anti-LGBT action or agenda. They’re smart enough to understand that prejudice carries its own punishment and that they’re not in a position to proclaim that henceforth all the gay things are icky. There wasn’t a secret attempt to marginalize anyone.

    Anyone with half a brain at Amazon would have realized that putting all LGBT books on the adult content list would have been noticed and objected to, had that always been the intent of the new policy. If Amazon were doing that and sticking to its guns there would have been an announcement of some kind, basically a PR-speak “up yours, gay people” in the most polite language possible, released before the change. That’s how corporate America works.

    I think that people are clinging to the notion that Mr. Bezos is a closet homophobe, or that jumping the gun on a boycott is “beside the point” and that there’s somehow “still an issue here” because righteous indignation feels good, and embarrassment doesn’t, and if they ignore all that yolk maybe it’ll disappear.

  24. Twitter is a great example of mob mentality as well as more proof of the lack of critical thinking skills among US citizens.

  25. I just want to know when they intend to apologize. Even if it was a technical error, it scared/hurt/insulted a large number of people. At least acknowledge this and make amends.

    Until then, I will take my money elsewhere.

  26. This all reminds me of the Live Journal fiasco a year or two ago, which garnered the exact same response, plus a laughable one day “strike” against blogging on LJ.

  27. I have to call bullshit on the “oh hey someone at amazon.co.fr got all confused, WHOOPSY DOODLE”.

    Adult isn’t a synonym for porn in French.

    The French words for porn, erotic, and sexuality are all English cognates, so there’s no reason to get them confused.

  28. @27,Mike

    Which a lot of people thumbed our noses at and posted anyway because realized that such a strike was ridiculous.

  29. It could easily be both a new corporate policy and a glitch (for values of glitch which include “human error”). De-ranking or otherwise hiding “adult” items, which I personally think is not useful, is still somewhat defensible, especially if it was part of the runup to a new safe-search/unsafe-search option. Once this policy was in place, a technical glitch or human error in the *categorizing* of those items would trigger the “fail” of the weekend, without discrimination against GLBT or other literature ever being part of the actual policy.

    Having done some (limited) web development in the past, I’m wondering if maybe we’re seeing an unexpected consequence of some background metadata they’re playing with. I could imagine Amazon thinking it would be a good idea to give the user more control over what kinds of items would appear in search–highlight/ignore GLBT-related items, for example–and that the data they’re putting together for this project got conflated with the adult-content flagging policy alluded to in e-mails to authors recently.

    In other words: a typical corporate right hand/left hand miscommunication screwup.

  30. I am in complete agreement with Nobilis.

    The “glitch” that people are saying they’re content with the explanation of does not explain why the code is in place to de-rank (and thus hurt the sales of) “adult” material.

    I’m an adult. Whether or not this de-ranking was unintentionally wide-spread over the weekend (which, sure, I buy even if I am in total agreement with the community uproar that demanded it stop targeting LGBT/feminist/etc material), I don’t want it happening at all.

    If you can’t stop porn from showing up when people search Harry Potter without hurting, say, the erotica sales accidentally when erotica books stop showing up on the “people who bought this also bought…”, I’ll go to a seller that has more competent coders and a less skewed corporate outlook on the acceptability of denying me information because of its moral judgment about the ‘okay-ness’ of adult material… which all of its customers are de facto presumed to be adult.

  31. @22 What bothers me more than anything is that Amazon had this de-ranking code in place at all. Is it right that they even consider such a thing?

    Yes THIS. This is what I’d like to see discussed. (Er… In general. In places. Not trying to force Scalzi to discuss anything. Uh. So yeah.)

  32. I don’t know if the “epic troll” theory is still viable or not, but all I do know is that if I were a troll, I would definitely want the credit for something like this.

  33. “Herdener did not respond to requests to clarify the cause of the error, nor about why works such as the “Milk” pictorial — which did not appear to be listed in any of the categories mentioned by Amazon — may have been removed from the search listings.”

    This is why people are still angry. The computer didn’t just randomly marginalize GLBT books on it’s own. It would seem obvious that someone made this choice and told the computer to do so — but the organizations unwillingness to come clean on the issue keeps the questions in the forefront.

  34. (I think if people want to filter out “Adult” content from their searches, it should be a personal choice, something they select in their own profiles, NOT a default and not imposed systemwide. Surely Amazon could swing that?)

  35. I have two issues with this, really:

    1) Would Amazon have responded so quickly and so forthrightly without “#amazonfail”? Given the way they seem to have responded to initial complaints in February, I am not sure this would have gotten past the CSR level without #amazonfail.

    2) This sort of search restriction code should be opt out.

    But yeah, this is certainly an issue of mistake, not homophobia.

  36. Sheesh – somebody messed up. It got fixed. If you hate Amazon for this, then take your business elsewhere.

    Nothing to see here, move along…

  37. Or maybe Amazon should go even further and bury the homophobic literature.

    No. I disagree. (Keep in mind, I AM gay.) Don’t bury it out of some sense of political correctness, that just leads to ridiculous kerfuffles. Keep that nonsensical crap out there, until it is finally, brutally mocked by the majority and thusly laid to rest forever. And how we get the majority to see that it’s stupid is NOT by burying it. Not in this day and age.

  38. This is a great post. It is much easier to join a mob and start parroting phrases and ideas. Did we not all learn after 8 years of George Bush that things are often more complex than can be summed up in a catchy slogan or phrase?

    Is Amazon either with us or against us? Should we invade them?

    I agree that amazon could have done a better job communicating what was going on, how long it would take to fix things. No doubt there that this was a massive pr fuck up. The canned email response with the phrase “ham-fisted” whoever came up with that should not work in PR. The communication should have been more detailed and serious. ham-fisted is more, aw shucks, we made a boo boo.

    People are assuming that there is some tag that is common between both porn and glbt material. That is not true. The person who made the database changes added the glbt material AND added the adult tagged material. They were two separate criteria, not one tag in common.

    Time to leave #amazonfail to the trolls.

  39. I have trouble buying the “adult material” excuse because the books delisted were all over the place and do not cleanly fit into one specific tag or category on Amazon’s site.

    I am with Colleen Lindsay on this one. This delisting had occurred earlier and Amazon sent out an explanation, at least one time, saying the book was delisted for being adult. (See Craig Seymour’s timeline here http://craigspoplife.blogspot.com/2009/04/my-amazonfail-timeline.html .) It would make more sense if a massive “attack” by fundamentalists tagging books as a category that gets delisted is what occurred…although I doubt that theory as well both because of examples like Mr. Seymour’s experiences with Amazon and because it would make no sense for fundamentalists with an agenda to tag and thus delist a number of the delisted books.

  40. Mac #37

    (I think if people want to filter out “Adult” content from their searches, it should be a personal choice, something they select in their own profiles, NOT a default and not imposed systemwide. Surely Amazon could swing that?)

    I agree with your premise, but what exactly is ‘adult’ content?

    Does that definition change when we are comparing books to movies to products like dildos or tasers? Where is the line in the sand when it comes to adult? Violence, sexuality, profanity, smoking? If you say all of them, who is going to read all the books to determine whether or not they should be included in the new ‘adult’ filter? How does a rated R movie compare to a book by Stephen King, George R. R. Martin, Anne Rice or J. K. Rowling? What about…

    I don’t want to be protected like that.

  41. @29: It sounds like “adult” not being a synonym for “porn” in French is exactly why this happened – someone applied the “adult” label where it did not belong precisely because they did not realize what it meant. You would not confuse them in French, but you would confuse them if you spoke French as your first language and were unaware of the other connotations we English speakers sometimes place on the word “adult”.

  42. My phrase of the week shall be “FOAMY FOAMY FAIL FAIL BOYCOTT GAAAH”. When my wife announces dinner I shall cry “FOAMY FOAMY FAIL FAIL BOYCOTT GAAAH”, when my children complain about their home work load I shall reply “FOAMY FOAMY FAIL FAIL BOYCOTT GAAAH.” As I write my tax checks I shall wimper “FOAMY FOAMY FAIL FAIL BOYCOTT GAAAH”.

  43. There were two exceptions in the small sampling and those ISBNs had Kindle categories as part of their designations.

    I can’t claim any insider knowledge, but I’ve worked with large, distributed datasets in the past and I’ve built tools that make heavy use of Amazon’s own data via its external APIs.

    When the buzz first started circulating, two innocuous facts made me think this was programmer error: first, books with words like ‘Gay’ and ‘Homosexuality’ and ‘Lesbian’ and ‘Sex’ in the titles were still showing up in searches and with rankings as long as they didn’t have publisher-provided metadata categorizing them as such. Even gay-positive childrens books were appearing on site-wide searches as long as they didn’t belong to specific metadata categories. Second, Kindle, audiobook, and specialty versions of the de-listed books were still showing up in searches.

    If you’ve spent time trying to clean up the results of a giant database snafu, those bits of information should be making your debugger-sense tingle. It smelled like an accidental bulk operation on a set of ambiguously named metadata tags, tags used in the ‘books’ section but not in the Kindle section, and unrelated to any other data like title, description, etc.

    Turns out, that’s exactly what had happened and only Amazon’s massive PR faceplant obscured the fact.

    The details of this problem were confused by the fact that their front-line customer support script just pointed people to the ‘Adult materials’ policy without recognizing that lots of products had been incorrectly categorized. One-by-one delisting of products using the Adult tag is still an issue that deserves discussion, but as our esteemed host points out, freaking out about the Easter incident is a distraction.

  44. Native speakers of French on other sites have said that ‘adulte’ means just the same thing in French as in English.

    Amazon’s explanation doesn’t quite cover all the facts, does it? And it contained no apology. They need to apologize for this fiasco, even if it was just a mistake. “We’re sorry for the implied insult to our gay and Lesbian customers and their friends,” just as a first stab at it.

    Alternatively, as someone suggested over on Making Light, they could make all the affected books 20% off for the next two weeks. Probably be a revenue-positive move in the long run.

  45. Jeff Eaton: Given that there was no discussion about it at all before #amazonfail, it seems to me that it was not a distraction but rather the cause of the discussion.

  46. Well said, Sir. I emailed them saying in essence “Hey, this looks bad, what gives?” and sat back to wait. I agree with commenter one, though, about the tagging system requiring investigation. But generally, I try to hold with Sofia Kovalevskaya – “Say what you know, do what you must, come what may.”

  47. So yeah, a screw-up.

    That said, Amazon’s practice of removing ranks of so-called “adult” items to remove them from search results doesn’t serve their customers or their suppliers well. It creates an invisible filter that can’t be opted out of. It prevents customers from seeing the current popularity of an item. It may prevent authors from seeing the current popularity of an item (and while agonizing over one’s sales rank is neurotic, seeing that spike after appearing in a major interview or receiving an award is cool).

    Amazon needs to update their search algorithms and database to use the adult flag differently, and not depend on stripping ranks to filter invisibly.

    They also need to actually apologize. They’ve said “we screwed up.” Now they need to say “We’re sorry.” A “we reacted badly, and were dismissive of authors reports of this problem” wouldn’t hurt either.

  48. Jeff Eaton: Given that there was no discussion about it at all before #amazonfail, it seems to me that it was not a distraction but rather the cause of the discussion.

    There was discussion, though. People passionate about the issue of Amazon delisting had been talking about it since February, and their friends and associates pointed this out many times over the weekend. I agree that the policy of hiding “adult” content from casual browsers will always result in problematic classifications at the edges — both with content that’s hidden AND content that’s left out in the open.

    At the end of the day, a couple things stand out:

    1) Amazon has a policy of preventing “adult content” of particular types — driven by complaints from their own customers — from appearing in search results or “You might also be interested in” displays. This has the unfortunate side effect of causing certain titles to effectively vanish from Amazon unless you’re very very deliberate when hunting for them.

    2) Some books receive that classification and are subsequently disputed. For example, some of the authors who complained in February.

    3) Over Easter weekend, a developer screwed up (either via code or incorrect pseudo-manual management of product metadata) and incorrectly classified ~50,000 books as ‘adult’. This included GLBT themed books regardless of sexual content, some feminist literature, some books about sexuality for people with disabilities, and so on.

    4) Amazon’s notoriously crappy customer communications resulted in #3 being conflated with #2, and a 36 hour orgy of outrage ensued with no corrections from Amazon. It was a colossal, monumental, cataclysmic PR screwup that could have been prevented had they just established lines of communication beforehand.

    It’s important to note that one can object to #1 and #2 without regarding #3 as a conspiracy or malicious act.

  49. I’d imagine delisting books over a holiday weekend would have a measurable impact on the sales of the books in question. I don’t think it’s fair to excuse Amazon because it was Easter. They should have called people in.

    I guarantee they would have if the site had gone down.

  50. Amazon’s response does not explain, in a way that I can understand, how _Heather Has Two Mommies_ got deranked. Surely even a deranged, barely-verbal, incompetent Frenchman could tell that that was a children’s book? I imagine there’s more to the story that we may or may not hear eventually.

    However, what I do understand is that there’s no Homophobic Conspiracy at Amazon, and that they are working on tweaking and debugging their search algorithms so that people who want to find books about rare orchids can find books about rare orchids and people who want to find erotica can find erotica. I can get behind that and since it is to their benefit as well I think it’s best handled internally by them with a minimum of harrassment from angry people with pitchforks, shouting in 140 character bursts.

  51. @50 Xopher:

    This is Amazon we’re talking about. They’re nothing if not multilingual. Their category terms are translated for every market they have a presence in. None of the credible reports actually suggest that this was a translation error.

  52. Jeff Eaton: I completely agree that it wasn’t a malicious act. My point was that prior to “#amazonfail”, *Amazon* didn’t seem to be part of the discussion.

    This got their attention in a way that the earlier complaints to customer service didn’t.

  53. Amazon’s response does not explain, in a way that I can understand, how _Heather Has Two Mommies_ got deranked. Surely even a deranged, barely-verbal, incompetent Frenchman could tell that that was a children’s book?

    Most likely because the Frenchman in question wasn’t going in and manually updating each and every book in the 57,000 item list — he was bulk-updating based on metadata. Heather Has Two Mommies was explicitly tagged as being sexuality-related by publisher metadata, while other LGBT-positive childrens books weren’t — and they continued to show up on site search and maintained their search rankings all weekend.

  54. Jeff Eaton: I completely agree that it wasn’t a malicious act. My point was that prior to “#amazonfail”, *Amazon* didn’t seem to be part of the discussion.

    This got their attention in a way that the earlier complaints to customer service didn’t.

    This is true. On the other hand, I’m not sure that the character of the #amazonfail discussion will help change anything — once the erroneous bulk-update is rolled back, will Amazon change their ‘adult’ policy? At this point, it’s easy for them to group legitimate complaints with the conspiracy theories.

    That’s my biggest concern; it’s something that matters whenever calls-to-outrage circulate on the Internet.

  55. Jeff 60: Wait, I thought the problem was that it was tagged ‘Gay/Lesbian’ and the update made all books so tagged “Adult.” Do I have this wrong?

  56. I think that keeping a book out of search results by setting its sales ranking to zero instead of setting a “don’t include in safe searches” flag is just plain stupid. If they couldn’t set up safe searches without hacking their own system by deleting accumulated data, they weren’t ready to roll out safe searches. (I’d bet that people inside Amazon who rely on those ranking numbers were pretty pissed off to discover that someone had cooked the stats, too.)

    Implementing safe searching without notice is also pretty dumb.

  57. Xopher: That’s correct. However, Amazon has several dozen *types* of internal metadata, including machine-generated automatic tags based on internal word usage patterns, publisher-provided metadata, and a host of others.

    A blog post from the day of the controversy includes a couple of pointers towards the metadata in question.

    Some of the books that were marked as ‘Adult’ in the Easterfail didn’t have and LGBT metadata, but did have ‘erotica’ as a publisher-provided categorization. Others that weren’t marked as Adult had ‘Homosexuality’ in the title but only included ‘Family’ and ‘Psychology’ in their publisher metadata.

    There’s a danger in using broad terms like ‘Tagging’ when describing what Amazon does to categorize their products. It’s a really, really big and messy tangle of data and the vast majority of it is handled solely by algorithms and bulk operations.

  58. My understanding is that is that someone did the equivalent of ‘update table books set adult=”true” where tags contains “sexuality”;’ under the mistaken assumption that “sexuality” meant “pornography”. I can certainly see how that would happen.

    What makes this confusing is that like every system that relies on human entry of metadata tags, the tags in Amazon’s library are most likely massively inconsistent.

  59. “Sheesh – somebody messed up. It got fixed. If you hate Amazon for this, then take your business elsewhere.

    Nothing to see here, move along”

    Agreed – stop buying from them, who cares. There are plenty of people who still will buy from them.

  60. Wait, people weren’t already boycotting Amazon for their anti-competition practices?

    Just me, then?

    Good to know.

    Ah, well, Powell’s loves me.

  61. @46 AQ:
    I agree with your premise, but what exactly is ‘adult’ content?

    Yes. This is a difficult question, true, but not any more difficult than defining porn/not-porn OFFline.

    If people take responsibility for what they personally do and do not wish to see, and don’t infringe on my right to see what things (nonexploitative!) I want, I don’t see a huge problem. If someone doesn’t want to see anything sexual at all and cheats themselves of some great literature, I’m hard-pressed to feel sorry for them at this juncture. (I have a feeling I’m in a particularly uncharitable mood at the moment and am not being entirely fair to the ostriches.) Hell, if the sight of a vibrator is terrifying and upsetting to some, then by all means they could tick whatever ticky boxes they like to stop this happening.

    I don’t want to be protected like that.

    Nor do I. Which I why I wouldn’t flip the
    theoretical switch, and why I would fight to keep such a switch under MY control. But Amazon wasn’t even providing a switch, just a blanket.

  62. Oh, I’m sure this wasn’t malicious, but it’s still stupid, and still shit PR/customer service. If they delist ANY books classified as adult, it may make it harder for me to find books I want. Bad PR/customer service and bad librarianship aren’t things I want to reward too much, and I really should have started giving more of my money to my local indies long ago. This is a good excuse to start.

    Also, their statement ISN’T an explanation. It’s a statement. Like “No officer, I didn’t kill him, he slipped and fell on that kitchen knife.” It might be true, but a jury isn’t going to buy it without some more evidence. I’d like to hear what the glitch was and how it happened.

  63. I may end up losing some friends over my position on this, but oh, well.

    More than anything, I’m really damned annoyed at the technological illiteracy on display in much of this protest.

    The idea that any one person–or even a team–could change categorization on 57,000 INDIVIDUAL books over a weekend is just mindblowingly preposterous.

    The ONLY way something like that happens is if something alters how those things are categorized EN MASSE, based on keywords.

    The explanation of how this happened–that a switch got flipped and suddenly anything with publisher-provided metadata containing “erotica” or “sexuality” got lumped in with their “adult” (porn) filter–makes perfect sense given the technical constraints of how large databases work.

    Anyone still arguing that this was somehow deliberate is only proving their utter ignorance of basic concepts of technology. It’s just as ridiculous as people who didn’t understand that the HTML rendering engine in IE couldn’t be removed from Windows without disabling its entire file navigation system. Or creationists who think that because evolution is called a theory, that means it’s not a fact in evidence.

    When something in science or tech goes wonky, if you don’t understand even the basics of what’s going on, then do the rest of us a favor and STFU and get out of the way of the people who do, so they can fix it.

    Whether the adult filter (which has been in place since at least August) is a problem in itself is still up for debate. But the fact that a bunch of non-porn stuff got accidentally thrown into that bucket, and it was NOT intentional company policy is not.

    Sometimes, my fellow progressives really piss me off with how quick they are to assume that anything that goes wrong at a large company is clearly company policy, and evidence that they’re run by Satan. Amazon is not Monsanto, folks, and merely being small doesn’t insulate a company from being jerks. The failures of two people– a lowly CSR and a code monkey–does not make Jeff Bezos and everyone else who works at Amazon suddenly evil incarnate.

  64. The ONLY way something like that happens is if something alters how those things are categorized EN MASSE, based on keywords.

    Dude, chill out. That the above was the only way it could happen is surely true. Whether it was deliberate or not remains up in the air.

  65. Sigh. Tal (#70) expresses frustration over the enormous Easter hissy-fit and points out that technical errors like this are just one of those things that happens sometimes with large automated systems, and silbey (#71) tells Tal to “Dude, chill out.”

  66. I’d like to hear what the glitch was and how it happened.

    So would I. And I’m willing to wait for them to actually track down and fix the full problem, first, instead of issuing a statement that might not be accurate.

    Heck, chances are good that the people they had to pull in to track this down were probably even out of town visiting family over the holiday weekend. Expecting an immediate response under the circumstances is ridiculous.

  67. Whether it was deliberate or not remains up in the air.

    Um, no, it’s not.

    If it were deliberate, it

    1. Would have been part of the original version of the filter which has been around for months.

    2. Would have included ALL copies of the affected books, and not just the ones with those keywords.

    3. Would not have included non-GLBT content like books on sex for people with disabilities.

    4. Would have had an available canned PR response at the ready when complaints happened. The response the CSR gave? Wasn’t PR at all, but clearly something designed for the filter itself, and not something for these particular titles.

  68. Mac #68
    Hell, if the sight of a vibrator is terrifying and upsetting to some, then by all means they could tick whatever ticky boxes they like to stop this happening.

    LOL. It is funny that I could find 4 pages of dildo products during the umm snafu.

    So I guess no switch, no blanket. Maybe just a tiny eye-patch for the ‘right’ eye. [yes, very groan worthy.]

  69. People should click on the link for Mike Daisey. Stand up progressive guy I’ve worked with before he got broader success. And he’s got a good handle on what happened–more human stupidity than anything conspiratorial or malicious.

  70. Whether the adult filter (which has been in place since at least August) is a problem in itself is still up for debate. But the fact that a bunch of non-porn stuff got accidentally thrown into that bucket, and it was NOT intentional company policy is not.

    Considering that Amazon itself said in an email, “In consideration of our entire customer base, we exclude “adult” material from appearing in some searches and best seller lists. Since these lists are generated using sales ranks, adult materials must also be excluded from that feature.”, I don’t think it’s out of the realm of possibility to consider something deliberate.

    Especially since that was in response to an inquiry by author Mark Probst about his young adult novel, “The Filly”.

    So either the person who responded to his email didn’t check to see whether there had actually been a mistake and just sent out the standard email or this was Amazon’s stance until they realised the backlash wasn’t worth it.

  71. Ana Stoica: Keep in mind that the person who told him this was likely a minimum-wage employee who had a script that told her exactly what to say when people ask about adult material and is specifically required not to deviate from the script. This CSR likely had no clue what “The Filly” was and probably spent all of 30 seconds on the issue.

  72. So either the person who responded to his email didn’t check to see whether there had actually been a mistake and just sent out the standard email or this was Amazon’s stance until they realised the backlash wasn’t worth it.

    Amazon’s front-line customer service is scripted. You hit the right keywords, you get the canned answers. It takes an act of God to push through the crap, and a number of books that received that level of dedicated push-back had the ‘Adult’ classification removed again.

    Easterfail was an intersection of many bad practices by Amazon (an un-toggle-able adult filter, canned customer service, zero communications channels, over-reliance on inconsistent publisher metadata) and developer error.

    The individual bad practices can be discussed — even attacked! — but this particular event needs to be seen as a consequences of multiple bad infrastructure decisions, not a deliberate censorship attempt.

  73. Jeff:

    I agree. But it still makes me badly disposed towards Amazon, particularly given that they refuse to explain what happened. Their statement was not an explanation, let alone an apology to customers/affected authors/publishers for the error. Someone in their PR department should be getting a very stern talking to. This is the sort of thing that you get called back from vacation for if you’re high up in PR, and I’m astounded they didn’t get anyone. It ended up on the websites of the bloody Guardian AND the New York Times. And even in Foreign Policy magazine! It turned into more than just a twitter shitstorm, and Amazon’s response has been slow, unco-ordinated, and just generally very poor.

  74. Sifting through everything, I am inclined to blame stupidity rather than conspiracy, but the damage is still there and Amazon’s reaction has been less than ideal. I still don’t see any indication of anything resembling an apology for the screw-up. At least after the Strikethrough debacle on LJ, pretty much the first thing they said when normal business hours resumed was something to the effect of “Wow, we really screwed up, didn’t we?” It would be nice if Amazon would make the same effort.

  75. Nobilis wrote: “What bothers me more than anything is that Amazon had this de-ranking code in place at all. Is it right that they even consider such a thing?”

    I suspect the normal mode of operation is not meant to be a ‘de-ranking’, but ‘never ranked from the time a new item enters the database’. But since the flag was toggled on existing items, it became a deranking.

  76. Am I the only one who thinks this was an outstandingly crappy hack for keeping the porn out of searches? If they’re willing to sell “adult” materials, why shouldn’t those items have a sales ranking? Isn’t it just as important to the publishers and authors as it is for “family” fare?

    This strikes me as a) lazy and b) puritanical on the part of their development staff.

  77. “I don’t think it’s out of the realm of possibility to consider something deliberate.
    Especially since that was in response to an inquiry by author Mark Probst about his young adult novel, “The Filly”.
    So either the person who responded to his email didn’t check to see whether there had actually been a mistake and just sent out the standard email or this was Amazon’s stance until they realised the backlash wasn’t worth it.”

    How would the CSR person know anything about the content of the novel? Do you seriously expect them to be familiar with the text of every item in the database?

    The CSR person looked up the book in the database, saw the ‘adult’ flag checked, and then explained how adult items are handled. The CSR person doesn’t know how it came to be marked as adult, and doesn’t know if that is accurate. So the CSR person assumes “The Filly” was marked ‘adult’ because it’s really ‘adult’, NOT because it’s a GLBT-interest book.

    When the CSR says ‘it’s adult material’, that isn’t based on the content of the item. It’s based on the flag in the database.

  78. “This strikes me as a) lazy and b) puritanical on the part of their development staff.”

    I can only guess that, perhaps, they’re working on a parallel Amazon for the puritanical Home School market, which would provide an alternate, filtered view of the catalog rather than the regular view. And they screwed it up. They’d probably exclude non-book product categories such as dildos, which would help explain why those didn’t lose their ranking.

  79. Jon H:

    That’s the thing – the book was NOT tagged adult. Not all of the GLBT/Romance/Feminist theory books had “adult” in their tags. Some were simply classed “gay”. The issue isn’t a person/scripting error/whatever adding “adult” tages. It’s books NOT tagged as adult being included in the same filtering category.

  80. Xopher, I’m pretty sick of the use of the words porn and adult. It’s like by using those words there’s some sort of justification.

    Someone above mentioned the filter itself (I know probably outside the scope of John’s post) but seriously would this conversation have the same feel if the filter had been applied to a subject category designation like political commentary?

  81. Eddie:”That’s the thing – the book was NOT tagged adult.”

    I don’t mean the user tags you see in the web page, I mean a database flag that is NOT visible to the customer on the web page, but would likely be visible to a CSR. A Yes/No boolean field in the record of an item in the catalog. An implementation detail.

  82. I think it’s fairly inexcusable to tag a book “Adult” in a way that’s not visible to the user, and without telling the author or publisher. It’s even less excusable to tag a YA romance with no explicit sexual content as “Adult” at all.

  83. Jon H:

    Fair enough. The problem, for me anyway, then becomes – why the hell are Amazon invisibly tagging books as adult, screwing with their search results and bestseller lists, and not really telling their customers or clients about it? (Which, from your post @86, you’d probably agree is pretty damn dumb?)

  84. “I think it’s fairly inexcusable to tag a book “Adult” in a way that’s not visible to the user, and without telling the author or publisher.”

    I suppose that’s reasonable, but it’s probably an implementation detail. I bet there are lots of database fields that aren’t represented on the page because they’re just optimizations or have no meaning to the customer.

    ” It’s even less excusable to tag a YA romance with no explicit sexual content as “Adult” at all.”

    I expect what was tagged Adult was not “a YA romance with no explicit sexual content”, but some internal id#, along with thousands of others in a batch operation. The actual content of the item was irrelevant.

    I don’t know how the Amazon database is structured, but I would bet you can easily navigate from a hierarchical category to all the books that are in that category.

    Probably the coding bug said something like “for all books in categories (category id numbers a, b, c, d, e…) set all member books to ‘adult'”

    and the list of category id numbers was wrong, resulting in _way_ too many books being flagged as adult.

  85. “why the hell are Amazon invisibly tagging books as adult, screwing with their search results and bestseller lists, and not really telling their customers or clients about it? ”

    One possibility is what I mention above, the home schooling store idea, where the definition of ‘adult’ might be a *lot* stricter than usual. Books that are ‘adult’ for home schoolers might not be ‘adult’ for the general market, so there’s no point putting ‘adult’ on the main Amazon.com listing page.

    Same thing if Amazon set up a Middle Eastern store.

    This is all conjecture of course, and probably completely wrong. It’s just a theory as to why you’d want a hidden adult flag that isn’t reflected in the item’s listing: to serve varying thresholds of moral offense with one catalog.

  86. Jon H 92: I expect what was tagged Adult was not “a YA romance with no explicit sexual content”, but some internal id#, along with thousands of others in a batch operation. The actual content of the item was irrelevant.

    I don’t quite buy that. There had to be something that led them to put that in a category that led to it’s being tagged Adult. No matter how far back you go, someone had to make that decision somewhere along the line. Now maybe it was the publisher’s submitted metadata, but I suspect it had more to do with the fact that the book in question was a YA romance with no explicit sexual content about two boys.

    If Amazon explained exactly what happened in detail, I might believe them. But they’ve so far shown no inclination to do that, and what anyone else says is conjecture.

  87. Jon H. #93

    One possibility is what I mention above, the home schooling store idea, where the definition of ‘adult’ might be a *lot* stricter than usual.

    I’d like to point out that ‘adult’ as defined by the filter used only applied to specific ISBNs that met the filter’s criteria. It had no noticeable impact on other ‘adult’ items such as dildos, vibrators, crotchless panties, Hustler, Playboy, etc. And the tagging you’re speaking of appears to be specific so although Playboy had the word erotic within its subject category listing. It did not have Books > Subjects > Literature & Fiction > Erotica as one of those listings. A subject listing which was shown in the DA sampling to be caught by the filter. That doesn’t mean it was the only criteria used for the filter but it is an interesting nonetheless.

    As for the ‘catalog’ in question it’s the default All Department filter from the main page. If you choose Books, the filtered items appeared as expected even though those books did not have rankings.

  88. “I’d like to point out that ‘adult’ as defined by the filter used only applied to specific ISBNs that met the filter’s criteria.”

    See my comment to Xopher above, where I mention that they probably went from the hierarchical categories to the books in those categories. A wrong, overly expansive list of categories would produce the result seen.

  89. Wow, what part of Hanson’s Razor don’t an amazing number of people NOT understand? I guess the definition of being paranoid really is believing that EVERYONE is out to get you: no exceptions, no acceptable explanations, it (whatever IT is) was all malice aforethought.

  90. Xopher: ” but I suspect it had more to do with the fact that the book in question was a YA romance with no explicit sexual content about two boys.”

    If it’s true that the glitch caught books about disability, it strongly suggests that effected items were not individually scrutinized at all for membership. The incident has all the hallmarks of an overly-broad qualifier in a batch operation.

    My guess is that ‘The Filly’ was caught because it is categorized under ‘Gay and Lesbian’ in the hierarchical map, not because of its actual content.

    The question is, what was the alleged French person *intending* to do? I mean, it’s pretty clear that the person screwed up bigtime; even if their intent *was* to censor glbt material, they failed badly on that score and included plenty of other things.

    I don’t think actual content was considered. I don’t think there was a master list of books to expunge, either by name or by ISBN.

  91. It might help a lot of people to read the Daily WTF for a bit to see how badly tech can get screwed up. If you are assuming that Amazon’s system is internally logical or consistent, then you assume too much!

    I’m not sure why people think they haven’t explained this enough. To me, ‘”non-native English speaking person dealing with tagging” + “poor and inconsistent tagging” + “large changes made without real oversight” == MASSIVE FAIL’ is an entirely believable explanation.

  92. “It might help a lot of people to read the Daily WTF for a bit to see how badly tech can get screwed up”

    A story I heard at Britannica: they implemented a profanity filter for the search field of the website. Unfortunately there was a bug after it went live. If I recall correctly, profanities from the testers were cached. So if you went to the live site and searched for, say, ‘horse’, you were likely to see a results page starting with “You searched for: horse f***er.” Only without the asterisks. So it was inserting profanities you hadn’t even typed, if your search term was part of a term used during filter testing.

  93. Sigh. Tal (#70) expresses frustration over the enormous Easter hissy-fit and points out that technical errors like this are just one of those things that happens sometimes with large automated systems, and silbey (#71) tells Tal to “Dude, chill out.”

    And Jim Milles (#72) gives a pointless play-by-play.

    If it were deliberate, it
    1. Would have been part of the original version of the filter which has been around for months.

    If it was a policy decided on later, no it wouldn’t necessarily.

    2. Would have included ALL copies of the affected books, and not just the ones with those keywords.

    Not if certain types of the books (aka the Kindle) work on a different system than the others do, which seems to be the case.

    3. Would not have included non-GLBT content like books on sex for people with disabilities.

    Not if it were a policy to censor all “adult” content and LGBT was included as “adult” deliberately.

    4. Would have had an available canned PR response at the ready when complaints happened. The response the CSR gave? Wasn’t PR at all, but clearly something designed for the filter itself, and not something for these particular titles.

    Not if it was part of a graduated rollout (“let’s see how this flies before we commit to it”).

    My point is not that I think it was deliberate; I, in fact, agree with you that it probably wasn’t. What I don’t agree with is your leaping up and down screaming abuse at people taking it as deliberate. Just like people should give Amazon some time to explain, you should give people some time to process that explanation.

  94. Steve Burnap:

    Do you have a link to the “frenchie did it” explanation? I only saw that as an unverified leak, not the company’s official response. Happy to be corrected if there is a statement on Amazon letterhead (so to speak) with that explanation. Currently the only official one I’ve seen is “Glitch. Fixing it.”

  95. Eddie: I have only seen the one link, but honestly it is the only explanation I’ve seen that makes any bit of sense to me given what I know of coding in general, and from what I know of Amazon’s corporate culture and technical infrastructure from friends who work there.

    I doubt you’ll ever say “this was caused by our crappy system design” coming from an official PR source. Best you’ll get is “glitch” which is, if the above is the explanation, true.

    I find it really hard to believe that a company with a good track record on gay issues would do this deliberately, and from what I have seen, this issue could not have been caused by external user tagging.

  96. Steve @65 wrote: “My understanding is that is that someone did the equivalent of ‘update table books set adult=”true” where tags contains “sexuality”;’ under the mistaken assumption that “sexuality” meant “pornography””

    I can certainly imagine someone doing a where clause with “WHERE TAGS LIKE ‘%SEX%'” thus getting “sex”, “sexuality”, “sexagenarian”, “Essex”, etc.

    Probably not what happened here, exactly, but you can do immense damage very quickly with a big database.

  97. Jon H. #96

    See my comment to Xopher above.

    —-
    I don’t know how the Amazon database is structured, but I would bet you can easily navigate from a hierarchical category to all the books that are in that category.

    If this the comment you’re referring to? I read it. I have a hard time buying the argument because the category listings that I’ve seen have books as the top hierarchy. Which makes sense because it’s one of the departments/search engine filters. Could you have another top tier hierarchy? Sure but it would be difficult to classify different product categories under a single hierarchy because the standards for what qualifies are very different for each product line.

    A wrong, overly expansive list of categories would produce the result seen.

    Let’s say for argument’s sake that I agree this was a glitch and nothing more. A filter that got triggered with unintended consequences. In order for that to happen the programming parameters must exist in the first place.

    So who created the parameters and what criteria was used for testing? We know it’s overly broad. Yet it’s not even close to wide enough for this filter to be considered an ‘adult’ filter? Not even if say it’s only for books because of so-called sexual content aka ‘the porn’.

    Some where along the line the criteria for this filter was approved. It made its way through the testing environment into the production environment. If what I’m hearing is true then all it took was an ‘accidental flip of the switch’ to make it happen.

    So please tell me, or better yet have Amazon tell me, how they accidentally created such an overly broad categorization that filters an entire grouping Books > Subjects > Gay & Lesbian with no thought to differentiate between non-fiction vs. fiction yet isn’t nearly broad enough to capture items that and yet doesn’t come close to capturing other ‘adult’ content.

    You can call it overly-broad and an honest mistake. I question the fact that a flag or programmed condition like that made it into the production environment at all. Unless it was already part of an official policy. Either that or the procedures in place for testing results against stated criteria are crap. Or the stated criteria was crap. If either of those things are true then what else about Amazon’s IT policies and procedures are crap?

    Or there was no pre-existing filter criteria and Amazon got caught implementing a policy to de-rank and de-list specific subject categories to service the needs of an identified portion of its customer base.

    Farfetched? Maybe but until we get more detailed answers from Amazon nothing should be off the table. Not even those questions that raise the possibility of censorship.

  98. I would STRONGLY suggest that anyone making sweeping statements about what is ‘possible’ and ‘not possible’ and ‘plausible’ take the time to check out The Amazon Ecommerce APIs. They provide a very tentative look at the mechanisms used to query, manage, and navigate Amazon’s internal data structures.

    It’s also important to note that the document linked above — all 570 pages of it — is a simple representation of the even more complex systems that they have internally.

    People here are throwing around words like ‘tag’ and ‘filter’ and ‘category’ and that’s fine for a cursory back-of-the-napkin explanation of a system you understand. But when you are inventing theories based on your personal feelings of how a ‘filter’ works, and what is possible for a ‘filter,’ it is no different than a creationist explaining that evolution violates the law of thermodynamics. If you aren’t willing to take the time to understand the system that’s being talked about, and you ignore people who have worked with it, do not invent theories about its internal operations.

  99. Jeff @106. Ah. Okay, so frex the ‘Gay and Lesbian’ book category is implemented as a “BrowseNode”, represented by an integer, and each node is part of a hierarchy.

    To get a list of all book records in that category using the API, you’d use an ItemSearch function call and provide the integer BrowseNode as an argument, and use the Books index. (Evidently different product types have separate indexes.)

    No mention of an ‘adult’ flag, so either that wasn’t exposed in the public API, or the docs are outdated, or they implemented it some other way.

  100. Just like people should give Amazon some time to explain, you should give people some time to process that explanation.

    True, the death of a pet theory deserves time to mourn.

  101. the death of a pet theory deserves time to mourn

    The death of any pet is worth morning, animal or theory.

  102. Jon @108 – yep, that’s the kind of selection criteria that would be used to assemble a list of products for any large batch operation. Amazon’s read-only public API lags behind their internal one, and I’d seriously doubt they would directly expose the ‘adult’ flag, but the underlying model of how they organize their data — into various hierarchical indexes — is important.

    Then, take into account the shfiting nature of Amazon’s indexes. Specific browsenodes can change and drift over time, publishers can change their own terminology, magically creating new browsenodes, and some books get misclassified into the wrong browsenodes if publishers or editors have odd notions of where a target market will look. Each new revision of the API spec comes with a detailed list of stuff that isn’t working this month, too. Heh.

    With those factors in mind, we can take a look at the mix of what books were affected — GLBT books, health and sexuality books regardless of orientation, disability sexuality related books, gay AND straight books that belonged to a general ‘erotica’ category… The fact that all of those were grabbed in one fell swoop indicate a bulk update that was too broad on many fronts — not just as it related to the GLBT community.

    I want to reiterate that I don’t think this negates concern over Amazon’s general policy of how to handle adult material. And I don’t think that it negates concerns over what does and doesn’t get classified. But a lot of the outrage also focused on disbelief that “dumb mistakes” in routine management of a large dataset could produce the results that everyone saw over Easter. In studying the product categorization system, and in reflecting on my own experiences managing large data warehouses, it’s totally, utterly believable that someone could make that kind of a screwup and not realize the implications until the data has propagated.

    There are lots of really important issues at play that this incident brought to the forefront, but they’re getting eclipsed by the “did they/didn’t they mean to” accusations. That’s worrisome to me because this incident has/had the potential to be a different kind of catalyst.

  103. You refer to “a” glitch, and “a” techincal failure.

    This was, apparently, the concatanation of three separate failures.

    First, somebody in France apparently changed the settings such that all books that had the meta-tag “gay and lesbian” were labeled “adult content”, possibly (or possibly not) with the awareness that this would remove them from the sales rankings and from the display by the search engines.

    Second– the “glitch”– because of bad coding, a setting done by one person in one place in one country in the world propagated across the world, to the Amazon in America (and elsewhere) not just in France.

    Third, and most notably, Amazon had set their “sales rankings” and their search results to censor the stock– the “sales rankings” are not actually the real sales rankings, but are the sales rankings minus the books that Amazon has decided you shouldn’t see. The search results are not actually the search results, but are the books the meet the criteria you search for minus the books that Amazon has decided you shouldn’t see. And they implemented this policy with no notice to their users.

    It is, in fact, the last of these failures that annoys me the most, and I’m not mollified by some idiot telling me that it’s just “conservation of moral outrage.” I am no less morally outraged because Amazon tells me “oh, we really hadn’t intended to be censoring just gay and lesbian stuff, we intended to censor ALL KINDS OF STUFF that we didn’t thing you should see.”

    Amazon, don’t tell me what I can read.


    Geoffrey A. Landis

    http://www.geoffreylandis.com

  104. All the holes had to line up for last weekend’s catastrophe to happen; but each of the holes was bad enough by itself. We just didn’t notice them until they lines up and let the light shine through.

  105. OK, if the facts presented here are true, this was definitely not an accident, it’s been true on the Kindle section for over a year, and Amazon was bullshitting everyone, and deleting any discussion of it from their boards, for a year.

    The publisher in the article pulled all its GLBT titles from the Kindle page and republished them as unmarked romance novels. They immediately started getting sales rankings. This was after spending months trying to get the situation resolved through channels and being stonewalled (¬π) at every turn. It was entirely intentional and the accusation of homophobia is supported by their experience.

    Time to do your shopping somewhere else, my friends.

Comments are closed.