Who to hire is a very important decision for most companies and a very difficult one to make. Given that we have data science as a potential tool that can help in so many different aspects of society. It's not a surprise that companies, particularly high tech companies, turn to data science to help make hiring decisions. The general idea is you train an algorithm to learn what traits, what characteristics of applicants are correlated with success on the job. So you can take anybody who applies or anybody who's a prospect and you code up whatever you can know about them. Their qualifications, their experience, their age, their demographics, perhaps their performance on certain test questions that you give them at the interview. You can fold all of that into some prediction about how well they will do their job. This sounds like a very sensible, nice, rational thing to do, but let's consider a possible scenario. So let's say that you have a company which, like so many other high tech companies, is mostly male. And let's say that there is a work culture that actually makes it very difficult for women to succeed. This is a certain clubiness To the way that things operate. An algorithm might, in a completely neutral way and a completely accurate way deduce, based on current data from the few women who were working at the company, that the average woman performs less well than the average man. >> She's a good one, too. Just look at that record. >> Okay, now Walter, you've had your little joke. Give her to somebody else, I asked for a man. >> And therefore, in terms of its rating, when it's rating applicants, may rate men higher than women where on other characteristics they're equal. The net consequence of this is that this company is going to hire even fewer women and this problem that they have in their workplace is going to continue and perpetuate with a supposedly neutral algorithm actually being a tool in this perpetuation. >> We are alive think we ought to quit before we had a chance to fire us. There's plenty of other good jobs around here. I don't know what you're going to do Sally, but I'm leaving early today to find me another job. >> If this company wishes to change their work culture and wishes to employ more women, they will actually have to modify their algorithm to take that into account. And have that algorithm become quote, not neutral just to be able to break out of the vicious cycle that they would otherwise be trapped in.