Feminists have the uncomfortable position

[Lucky Luciano]

70. The programmers don't make the inputs.

The programmers are writing statistical procedures with large datasets. Natural language processing (NLP) is based on statistical analysis of words in documents to try to make a computer grasp the meaning of the content.

Here is a paper that used thes techniques 25 years ago. English and French versions of Canadian parliament proceedings were pushed through their algos to see if they can teach the computer how to translate English and French. It is a foundational paper. Google Translate is based on this ultimately (all the authors are now at Renaissance Technologies - it is that Robert Mercer yes)

http://dl.acm.org/citation.cfm?id=972474

NLP has evolved a lot since then and can be used to work on the problems you suggest- more diversity is needed too, but analytics will solve the problem. We almost need to make the computers overcompensate for society's words that contain gender biases. This is a nontrivial problem.