They found that contrary to their initial hypothesis, women with voices modulated to sound like men were still not getting hired at the same rate as the men masked to sound like women. Not only that, but they found women modulated to sound like men did worse than unmodulated women, and men modulated to sound like women did better than unmodulated men:
http://blog.interviewing.io/we-built-voice-modulation-to-mas...
It seemed to me that the natural conclusion was that woman have lower confidence and thus performed worse at interviews.
And this is a deep realization for me, because it pushes me in the direction of current diversity policies. You see, if there is perceived bias against you, then you lose confidence. So we must, for a while, try our best to remove that perfection. That might mean hiring more woman even if they don't seem to perform as well as men, because, once we have done that for a while, woman will feel more confident and the good candidates will appear.
In any case - whether my last paragraph makes sense or not - the conclusion of the experiment interviewing.io did is that only removing gender perception during the interview process isn't enough, for woman may already have been affected by the bias and thus will perform worse than men during the interview.
So we come back to the conclusion that we must invest to bring woman to tech early - during college or even high school - and fight biases there.
It's a long road anyway, isn't it?
Use a voice scrambler.