There's a lot of discussion about the problematic Google manifesto, and one of the issues that's brought up is the science of sex differences. Part of the problem is that it's easy to cherry-pick evidence to support your own biases. Based on research on young chimpanzees, I could claim that humans have an evolutionary basis for females to be tool-oriented and males to be more socially-oriented. But it's actually quite harder to truly understand human sex differences because we still don't have the data to fully understand them.
If you only know a little bit about human biology, it might sound simple. XX or XY? Ovaries or testicles? Estrogen or testosterone? But in reality, there's a wider range of developmental possibilities. Development of sex-typical traits isn't just determined by the sex chromosomes, but is guided by multiple pathways that start in the womb, and continue through adulthood. For example, individuals with androgen insensitivy syndrome (AIS) may be XY, and produce androgens such as testosterone, but may develop female-appearing or intersex genitalia due to lack of functioning androgen receptors. Furthermore, the actions of any given hormone is interdependent on many other hormones, proteins that bind to them, and enzymes. For example, testosterone plays a role in male sex drive, but only because its converted to estradiol in the brain. Hormones can also have somewhat paradoxical effects. For example, administering oxytocin, a hormone associated with social bonding, in experimental settings can make people more trusting, but it can also make them more xenophobic. But most importantly, hormones are calibrated to developmental environment and daily experiences, which means which we can't attribute anything we measure in adult humans to "just" biology or genetics.
So, biology is complicated. BUT, measuring biology is equally as complicated, and very susceptible to biases. The impact of these biases in shaping scientific research was explained by Stephen Jay Gould in the Mismeasure of Man, and others (such as Marlene Zuk, Holly Dunsworth, and Ambika Kamath) have added to that. But we are still far from overcoming those biases. Biomedical research has only recently begun to require including female rodent models. Most of our understand of mammal biology and neuroscience comes from rodents, but its still largely biased toward understanding male rats. And most of understanding of human psychology and neuroscience is biased toward Western, Industrialized, Educated, Rich, Democratic (WEIRD) populations. That means a lot of the science of sex differences is based on the norms of a limited set of cultures. Furthermore, the research process often requires weeding out 'outliers.' Imagine that you are researcher studying sex differences in personality in a population of university students. Are you going to include students who are intersex or trans? Probably not. Are you going include students who are gay or bi? Probably not. To get the nice, neat data you are looking for, your selection criteria requires omitting a lot of the variation that could obscure your results. And even the variation you do see at that stage is going to be shaped by the common cultural enviroment your participants were raised in.
There's an extensive body of literature on sex differences in children, but many of these are attributed to socialization. It's impossible to separate the impact of biological sex from the influence of gender socialization that begins from birth. There's a reason baby clothes are so strongly gendered--it gives people cues that impact how people respond to "girl" or "boy" babies. This makes it very hard for us to separate out out the impact of biological sex versus the impact of cultural socialization in shaping sex-typical behavior.
But one area in which we have growing evidence is in the impact of perceptions of gender (and ethnicity) in evaluating students and potential job candidates. Research indicates that simply switching a name can influence how someone is perceived. For example, candidates for a lab manager position with a female name are rated as less competent, and offered lower salaries, despite equal resumes. And manipulating the gender or ethnictiy of a name can influence the likelihood of getting a response from prospective supervisors. There's also research showing that class indicators help men receive positions, but decrease opportunities for women.
Add to that hostile or unsafe environments and harassment, and we can see evidence of cultural barriers to women's participation and advancement in some areas of science and technology. So why are these discounted? Quite simply, it all comes back to those biases. We all have cognitive biases, and they are shaped by our cultural environment. In the study examining candidates for lab managers, female scientists were just as biased as male scientists in evaluating female-named candidates. We can take steps in critically examining our biases and take steps to account for them in our research and hiring practices. But nonetheless, part of "overcoming" those biases is realizing that we can never completely overcome biases, and instead recognize how they shape the science that we do and the we way we evaluate the competency of the people around us.
To examine your own biases, check out Harvard's Project Implicit. For further reading on the role of hormones in development, I recommend Randy Nelson's An Introduction to Behavioral Endocrinology. For more background in biases in studying human traits, I recommend Stephen Jay Gould's The Mismeasure of Man.