Any group underrepresented in [valuable activity x]relative to its distribution of talent.
You can theoretically have a distribution in which natural aptitude is equal, the distribution is skewed (more men than women) and it's not horribly inefficient, but I think if you spell out the assumptions of such a distribution, it's not especially realistic.
Depends on why. If it really is because regardless of what social expectations there are, women are just less interested, that's harmless. But I think it's obvious that with a look at history and present attitudes, we can't assume an observed difference in interest is neutral that way.
You can theoretically have a distribution in which natural aptitude is equal, the distribution is skewed (more men than women) and it's not horribly inefficient, but I think if you spell out the assumptions of such a distribution, it's not especially realistic.