Everyday bias in everyday algorithms

With its apparent distaste for any names not traditionally ‘British’-looking, the computer spellcheck illustrates a wider issue: that our supposedly neutral technology is skewed by the bias of those who built it, says Hannah Ellis

When I think about mistakes, the written kind, there’s a certain nostalgia to how I imagine them. Grey-ish clouds of rubbings out, complete with the occasional rolled fleck of eraser embedded into the page. Liquid-paper, never quite convincing, that dries like Artex and is less written over afterwards than carved into. Crossings out, from the neat single-line strikethrough to the tangles of circular doodles ensuring that the words beneath can never be read again.

But of course, I’m thinking of pre .doc mistakes. The kind made before word processors and the present general ambition of society towards complete technological integration, towards total efficiency and paperless-ness. These errors, unlike my nostalgic ones, are backspaced away as if they never happened to begin with. Around them forms a unique semiotic language, a vocabulary of red squiggly lines and dashes that trill ‘Here’s a mistake!’; words underlined in bright and unmissable RGB red, picking out ‘unmissible’ and its clumsy second ‘i’ that needs correcting to an ‘a’.

The genius isn’t particularly in the code itself, which is relatively simple apparently, but in the sheer untidiness of the leftover results. Pages crawl with red lines, infested with errors – or, at least, mine do – and the perfectionist in me worries over them. They’re as difficult to ignore as they are to miss.