When you submit a paper to an academic journal, the reviewers do not know the name or origin of the authors, and make a decision based on the contents only. Conferences should select talks using a similar process. This would prevent potential bias against women, and also shield the conference from allegations of bias.
This has come up a few times for PyCon, but we get a very wide range of proposal qualities from a wide range of people that would make it hard to come up with the best schedule we can, which is the ultimate goal of the Program Committee. Some of the best speakers submit relatively poor proposals if you read them without their name and experience. Those things obviously aren't everything, but they do hold some weight in the review process.
Anyway, we ended up with 20% women on the schedule, up from 6% in 2012. We did it purely through outreach.
Conversely, the difference is that women are incentivised just as much as men to apply and be accepted as (in the UK at least) it's tied to research outcomes.
At least one other article has made the point that conference organisers should aim at getting more females to submit by a) finding female networks and making them aware of the CfP and b) even putting a quota on the submissions. That way, unless something is very wrong with the pool of people, some women's talks should just get through by sheer probability.
