We talk about how many women were raped last year, not about how many men raped women. We talk about how many girls in a school district were harassed last year, not about how many boys harassed girls. We talk about how many teenaged girls got pregnant in the state of Vermont last year, rather than how many men and teenaged boys got girls pregnant. So you can see how the use of this passive voice has a political effect. It shifts the focus off men and boys and onto girls and women. Even the term violence against women is problematic. It’s a passive construction. There’s no active agent in the sentence. It’s a bad thing that happens to women, but when you look at that term violence against women, nobody is doing it to them. It just happens. Men aren’t even a part of it!Jackson Katz, Pd.D.
Sit with that for a moment. Then read it again.
We can call the above a biased representation of the data or we can call it oppression. Depending on our choice, we locate the problem in different places.
Talking about bias in data and technology only takes us so far. Calling it bias subtly or not so subtly locates the source of the problem in us. There is a movement of people beginning to talk about ‘oppression’ rather than ‘bias’ because oppression is a concept that challenges power structures. Yes, bias in datasets needs to be addressed. But addressing bias in the datasets alone will not address or acknowledge the root causes of structural oppression. We need to call out the oppression and the way the data bias supports this.
When we leave people out in our design choices, our technology choices do we subtly blame them? Blame them for not having a job, a smartphone, a thumb, or being the wrong gender, colour, religion, and so on?