Sorry, haters.

Graph showing the percent of women in each of 4 majors that were previously dominated by men. All four lines increase in lockstep until 1984, at which point Computer Science starts losing women relative to the others, to the extent that Computer Science in 2010 is comparable to Computer Science in 1970.
Source: “When Women Stopped Coding”, NPR (Planet Money), 2014

This article is Part 2 of a series about the Google Manifesto. [Part 1] [Part 3]

I hate to say it, but the Google Manifesto has revealed a truly staggering degree of science illiteracy. People read the memo, see that the author [1] can put together ‘expert’-sounding sentences, and assume that the author has actually thought his argument through. It is clear that he has not.

One of the manifesto author’s key points (heading “Personality differences”) is that research finds that women and men exhibit statistical differences in the Big Five personality traits. He places a particular emphasis on Openness:

Women, on average, have more:

  • Openness directed towards feelings and aesthetics rather than ideas. Women generally also have a stronger interest in people rather than things, relative to men (also interpreted as empathizing vs. systemizing).

  • These two differences in part explain why women relatively prefer jobs in social or artistic areas. More men may like coding because it requires systemizing and even within SWEs, comparatively more women work on front end, which deals with both people and aesthetics.

This is a testable hypothesis, so we’re going to do science on it.

Before we begin, though, we need to understand a few things. “Science” is a complicated beast: each field has its own criteria for evidence, its own methodologies, its own process for establishing new ideas and replacing old ones. But for many fields, the process can be summarized as: observation, hypothesis, prediction, experiment, conclusion.

  • First, you observe the world and find something about the world that doesn’t fit into your field’s existing body of knowledge.

  • Next, you construct a hypothesis that, if true, would help to explain that observation.

  • After that, you make predictions by studying your hypothesis and looking for something that it predicts that no one has realized yet.

  • At last, you experiment: you manipulate the world in a way that (you think) will make the prediction come true.

  • Finally, you reach conclusions: if your prediction came true, your hypothesis survives… for now.

If you are lucky, the hypothesis fits snugly into the existing body of knowledge and becomes part of a scientific theory. A theory is an explanation of why; where a hypothesis says “if I do X, then Y will happen”, a theory says “in situations like X, the result will necessarily be Y because Z”. A hypothesis can be confirmed or disproven, but a theory can only be replaced by a deeper theory.

(Not all sciences work this way. In history, there’s no such thing as an “experiment”… at least, if we make some reasonable assumptions about the non-existence of time travel. In psychology, medicine, and other fields involving the study of humans, it’s often unethical or prohibitively expensive to actually manipulate people, so instead one often does an observational study. A rule of thumb is, it’s best if you formulate your hypothesis before seeing the data that could invalidate the hypothesis.)

Okay. Hopefully that was all review for you, as that was supposed to be a basic high-school level review of how science works. (If you didn’t learn all of that in high school… well, the U.S. education system needs some work.)

Now we’re ready to drop some science on this memo.

  • Observation: Women are less likely than men to major in Computer Science. I think we can all agree on this.

  • Hypothesis: Manifesto Guy argues that this observation is due to biological differences between the average man’s brain and the average woman’s brain: namely, that women are more oriented toward people and prefer more-social occupations, and men are more oriented toward things and prefer less-social occupations.

  • Prediction: From this hypothesis, we would predict two things:
    (A) We expect that people-oriented majors would have men:women ratios that trend together with other people-oriented majors, and similarly that thing-oriented majors would have ratios that trend together with other thing-oriented majors. (The people-oriented majors and the thing-oriented majors may trend separately.)
    (B) We expect that the men:women ratios of people-oriented majors would be consistently and measurably more biased in favor of women when compared to the same ratios of thing-oriented majors.

Let’s pick, oh, Medical School and Law School as people-oriented control majors, Computer Science is our experimental major and is obviously thing-oriented… and let’s aggregate the Physical Sciences into one thing-oriented control pseudo-major so we have enough data to work with.

  • Experiment: Reveal the data by looking at this graph. Whoops, wrong link. Looking at this graph.
    (A) The ratio for Computer Science has a stationary point at 1984, a 1984–2008 downward trend, and a 1966–1984 upward trend, all of which persist when relativized against the more-consistent upward trends of BOTH the people-oriented control majors AND the thing-oriented control major. (We predicted that the Computer Science trend would follow the Physical Sciences trend. Oops.)
    (B) The ratio of Computer Science crosses the ratio of the people-oriented Medical School major several times (1973, 1977, 1987), crosses the ratio of the people-oriented Law School once (1973), and indeed in 1965 Computer Science is more biased toward women than any of the other majors. (We predicted that Computer Science would be less biased toward women than both Medical School and Law School, and similarly biased toward women compared to Physical Sciences. Oops.)

  • Conclusion: Both predictions are falsified. If any bias favoring women truly exists in the people-oriented majors, it is clearly overwhelmed by other factors not controlled in this thought experiment, so we are forced to reject the idea that our initial observation (more men are majoring in Computer Science than women) was caused by our hypothesis. Manifesto Guy’s hypothesis is ruled out fairly conclusively.

This is just a small taste of the poor scholarship in the Google Manifesto: two minutes with a search engine invalidates one of his core premises. In a future article, I’ll pick apart the places where the memo logically contradicts itself.

The next article in this series discusses the memo’s bad statistical reasoning.

Footnote 1

No, James Damore is not a Ph.D. He lied about that on his LinkedIn profile, and has since removed it. His specialty of Computational Biology is also irrelevant to Neurobiology, Psychology (“Evolutionary” or otherwise), Sociology, or any other field relevant to the citations he makes in his memo.

However, he should be well enough educated to know better. The conclusion that springs to mind is that he did know better, but he was being sloppy. And if getting fired for writing a memo is bad, getting fired for writing a sloppy memo is just plain embarrassing.

Addendum

It’s also worth noting that this statement in the Google Manifesto is blatantly wrong when it comes to the arts. In the visual arts, men and women earn MFA degrees at roughly the same rate, but art made by women makes up only 3%-5% of the major permanent collections in the US. Likewise in music, major orchestras have men:women ratios that vary from 3:2 all the way to 13:1. And in literature, books that get reviewed are written by men at a ratio that varies from 11:6 (about 2:1) to 37:13 (about 3:1). Men earn much more respect (and money) in the arts industries compared to women, even though the Google Manifesto claims that arts and aesthetics are the areas where women are stronger than men.