{key: value} : algorithmic debiasing in practice
Algorithmic bias is familiar to many of us by now. Like, probably, you, I’ve read papers and watched talks on it for years now. And yet, when I sat down to code an interface that relied on machine learning, I realized I had plenty of questions – but no answers. As much as I’d learned about the big picture, I had no idea how to translate my values into the next line of code. When I looked for literature, I found lots of context, but nothing like a code sample. And yet, I still had decisions to make about interaction elements like autocomplete, or image display, or data visualization. How do I get from context to code? How do I implement these values? In this talk, I’ll describe specific situations I faced, and the concrete questions of interface design they raised for me. I’ll ask a lot of questions. Perhaps you will have some of the answers.
