The computer did it…

It seems that ‘the computer did it’ is becoming as much a cliché in the early twenty-first century as ‘the butler did it’ was 100 years ago. There’s an interesting link by Cory Doctorow on bOING bOING to a blog post by one Pete Ashton about the already infamous ‘Keep Calm and Rape A Lot’ T-Shirts being sold through Amazon’s marketplace.

Computer-generated 'Rape' T-shirts sold via Amazon
Computer-generated ‘Rape’ T-shirts sold via Amazon

Only the explanation given is incomplete in important ways. This is not to encourage people to attack Pete who, as his post explains, is not in any way connected to or responsible for the T-shirts or the company that produces them. However the explanation that ‘it was an algorithm that did this and the company didn’t know what was being produced until it was ordered’ is inadequate as an explanation. Here’s why.

1. This was not simply a product of computer generation nor do algorithms just spring fully formed from nature. All algorithms are written by humans (or by other programs, which are in turn produced by humans) and the use of an algorithm does not remove the need to check what the algorithm is (capable of) generating.

2. There was a specific number of verbs included in the algorithm for generating these T-shirt slogans (621 verbs in fact).  Even if they were generated by selecting all the 4 or 5 letter words in a dictionary of some sort, it’s not that hard to check a list of 621 verbs for words that will be offensive.

3. There words following the verb were not even as random as this. In fact, they are specifically A LOT, HER, IN, IT, ME, NOT, OFF, ON, OUT and US. Several people have checked this. There are some very interesting words missing, notably HIM. This list is clearly a human selection and its choices reflect, if not deliberately misogynistic choices, at the very least a patriarchal culture.

Algorithms, as cultural products, are always political. They are never neutral even when they appear to be doing entirely unremarkable things. The politics of algorithms may be entirely banal in these cases, but in some, as in this case, the politics of algorithms is accidentally visible. T-shirts may be a minor issue, but what’s much more important is not just to accept the idea that ‘the computer did it’ as an infallible explanation when it comes to rather more consequential things: all the way from insurance and credit rating through police stop-and-search and no-fly lists to assassination by drone. Otherwise, before we know it, the opportunity to question the politics is buried in code and cabling.