Think about it like the movie Moneyball. By rejecting the flawed collective wisdom of baseball insiders, they were able to create a winning team on a budget by using overlooked metrics. Likewise, job interviews are notoriously unreliable and inconsistent at finding good employees. The uneven demographics in tech jobs is likely partly the result of flawed wisdom in the hiring process. If you sample more heavily from an irrationally underrepresented group, you would get on average higher value candidates than if you sampled evenly from the whole pool.

Selecting solely based on merit is only effective if your selection process works properly. In practice, we see that both conscious and unconscious biases pervade the selection process, so I think this method is worthwhile.