Several folks have asked, if Amazon's deranking of GLBT, disabled, and other books was the fault of an improper mapping, then why did it affect some titles as far back as February?
The Ja(y)nes at Dear Author have pointed out that Amazon started pulling in category data for titles.
If you look at the category data in the Dear Author post, you'll see that Heather Has Two Mommies has the Gay & Lesbian stamp, while A Parent's Guide to Preventing Homosexuality doesn't.
This is consistent with Amazon's official explanation: a content filter scheme incorrectly mapped anything categorized as GLBT and other categories into the Adult filter and deranked it.
So how come Mark Probst's book got deranked back in February?
Probst's book is new, and the newer titles will likely have more complete category metadata. If the mapping error had been in Amazon's system back then, Probst's book would had been an early causality.
Amazon has a huge catalog, so the mappings may not had been applied across all ISBNs, or they may had not had sufficient category data. In the February time frame, derankings may have been isolated.
Roll forward to April, and you've had more time to run mapping jobs and update the resulting data stores, as well as update the collection with category data.
Older editions and or paperback printings of titles that either don't have category data, or incomplete category data, are spared from the deranking.
More and more titles have been deranked and removed from search, books that don't have the blacklisted categories percolate up in the search results, and Amazon's caught out pantless.
Another thing, pointed out in a post by Soft Skull Press Publisher Richard Nash, when you're doing content filtering, it's easy for what he calls the non-normative stuff--GLBT, sex-positive books for people with disabilities, sexuality and race--to be vanished, because we have a cultural norm that codes those things as dangerous, and it's too easy for someone not being mindful to add those categories to a 'smut' filter.