At Data For Good, we've used all the AI resources at our disposal to implement the Bechdel test: “are there at least two women talking to each other about something other than a man?” Did you know that only 50% of films pass this simple test? We've gone even further, using audio, text and video processing algorithms to automatically detect broader indicators of sexist bias in audiovisuals (age of actors whose characters are “in a relationship”, distribution of speaking time, representation of bodies…). An exciting project that provides a wealth of sociological and technical insights :)