Because the "science" is shaky as hell. The gender disparity in CS and engineering is both enormous and resilient to attempts to dispel it, yet we're supposed to believe these "internalized gender stereotypes" are the main cause despite those same stereotypes somehow not affecting chemistry or advertising and only barely affecting mathematics?
I'm sorry, but why do you say the science is shaky or that it barely affects mathematics? Google "stereotype threat mathematics" and there are oodles of credible scientific studies documenting the effect. From one of the articles, such as this one from NYU's Department of Psychology (http://steinhardt.nyu.edu/news/2008/1/29/Stereotype_Threat_Affects_Women_in_Highlevel_Math_Courses_Aronson_Study_Finds):
"Considerable research over the past decade has shown that women's performances on math tests are compromised by stereotypes. In over 200 published experiments, females as young as first graders and as old as 22 have been found to perform worse on math tests whenever the testing environment cues them to think about their gender, a phenomenon named "stereotype threat" by the psychologists Claude Steele and Aronson in the mid 1990s. (Emphasis mine.)
Where's the shaky research here? The studies showing that women behave worse on difficult math tests when they are reminded they are women prior to taking the test is both reproducible and fascinating.
And your choice to cite mathematics as a counter-example is particularly interesting, given that 27% of math PhDs in 2012 (drawn from a pool of U.S. Citizens) were women. (http://www.ams.org/profession/data/annual-survey/2013Survey-NewDoctorates-Report.pdf). If we look at faculty positions, the data are show that women are incredibly underrepresented in professorships (the numbers are equally striking across most STEM fields (http://www.human.cornell.edu/hd/ciws/upload/SexDifferencesMathIntensiveFields.pdf):
In the top 100 U.S. universities, only 9% to 16% of tenure-track positions in math intensive fields are occupied by women (Nelson & Brammer,2010). Among full professors, women number around or fewer
than 10%: computer science, 10.3%; chemistry, 9.7%; economics, 8.7%; chemical engineering, 7.3%; mathematics, 7.1%; civil engineering, 7.1%; electrical engineering, 5.7%; physics, 6.1%;and mechanical engineering, 4.4%. In contrast, women are much better represented in the rest of the sciences and humanities, often making up one third or more of professorial posts.
One of the things that bothers me about the responses I see here are that posters never grapple with the actual data. You cannot claim to be a nerd if your response to empirical studies showing that gender cues drive significant differences in testing is to simply ignore the data. I used to think like you, and then I started reading the research. When you do that, you discover that the human brain is amazing at internalizing external social cues. If scientists ignored inconvenient results, we'd never have discovered that the universe is actually accelerating. (https://en.wikipedia.org/wiki/Accelerating_universe)