A quick overview and dialogue on gender bias in AI

For Worldwide Girls’s Day, I wished to jot down a brief article about gender bias in AI.

AI fashions mirror, and sometimes exaggerate, present gender biases from the actual world. It is very important quantify such biases current in fashions with the intention to correctly deal with and mitigate them.

On this article, I showcase a small choice of vital work executed (and at present being executed) to uncover, consider, and measure completely different features of gender bias in AI fashions. I additionally talk about the implications of this work and spotlight just a few gaps I’ve observed.

All of those phrases (”gender”, “bias”, and “AI”) may be considerably overused and ambiguous.

“Gender”, inside the context of AI analysis, usually encompasses binary man/girl (as a result of it’s simpler for laptop scientists to measure) with the occasional “impartial” class. “AI” refers to machine studying methods skilled on human-created information and encompasses each statistical fashions like phrase embeddings and fashionable Transformer-based fashions like ChatGPT.

Throughout the context of this text, I consult with “bias” as broadly referring to unequal, unfavorable, and unfair therapy of 1 group over one other.

There are lots of alternative ways to categorize, outline, and quantify bias, stereotypes, and harms, which is exterior the scope of this text. I embody a studying record on the finish of the article, which I encourage you to dive into should you’re curious.

Right here, I cowl a very small pattern of papers I’ve discovered influential learning gender bias in AI. This record is just not meant to be complete by any means, however moderately to showcase the range of analysis learning gender bias (and other forms of social biases) in AI.