Hacker News Reader

Comment by mjg59 | Hacker News Reader

Four scientists with no expertise in the field of diversity and, as a result, in no position to argue that the scientific argument made justifies the conclusions drawn (as one of them admits).

A scientific argument is more than "Here's some science, here are some facts and here's the conclusion" - you have to use the science to explain the observed facts, and doing so takes more than saying "And therefore". There was no attempt to do so whatsoever, and so (from a scientific perspective) this was junk.

View on HN

There has to be a more charitable explanation than your implied "four scientists don't know how to do science properly" thesis.

Is diversity honestly so narrow a field, that a social psychologist, personality psychologists, evolutionary psychology, and a sexual neuroscientist are in no position to say anything?

Gender diversity always seems to bring up two issues (1) are men and women different in mental reasoning, and if so where and how significant are these differences and (2) would differences in mental reasoning lead to different outcomes in choice of studies or employment?

Is there a specific scientific field for that?

> There has to be a more charitable explanation than your implied "four scientists don't know how to do science properly" thesis.

Indeed - one of them knows how to.

> Is diversity honestly so narrow a field, that a social psychologist, personality psychologists, evolutionary psychology, and a sexual neuroscientist are in no position to say anything?

Unless they've done any research in it, sure. Why should any of them have more insight into diversity than, say, a nuclear physicist?

The issue isn't whether there are gender-related differences that are relevant to the job. It's not even whether those are significant enough to result in a measurable difference in outcome. It's whether those factors are the primary reason for the observed disparity, and neither the document author or the quoted scientists present any evidence whatsoever that that's the case. Failing to do that means that the conclusions drawn (ie, that existing diversity initiatives discriminate against men while providing no benefit to the company) aren't scientific.

Those scientists specifically studied gender differences so they can at least confirm the Google author was actually correct in his factual claims.

Where anyone can disagree is how he interprets them, but I am glad to see four experts clarify a little bit about what is true.

They can confirm that he was correct in a subset of his factual claims, but not in the further factual claims that he builds on those. Implying that they can is bad science.

Perhaps I misread it, but it sounded like all the gender differences he claimed were backed by research. What were the factual claims he got wrong?

I didn't say that the factual claims were wrong, I said that the scientists quoted in the article were not in a position to confirm them. Someone with appropriate domain expertise would be able to do so, but none appear to have been asked.

Also, even if the gender differences were backed by research, what's not backed by research is the assumption that those differences have any significant impact on women's capacity to perform the same work as men. Remember--the scientific method is about making a falsifiable hypothesis and then gathering enough evidence to be able to reject it. In this case the falsifiable hypothesis would have to be that men and women have no significant difference in their ability to perform the same work.

Strawman. Nobody and especially not the memo made a claim about capacity. It was all about interest on average, which is absolutely backed by research :

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166361/

Abstract : "There is considerable interest in understanding women’s underrepresentation in science, technology, engineering, and mathematics careers. Career choices have been shown to be driven in part by interests, and gender differences in those interests have generally been considered to result from socialization. We explored the contribution of sex hormones to career-related interests, in particular studying whether prenatal androgens affect interests through psychological orientation to Things versus People. We examined this question in individuals with congenital adrenal hyperplasia (CAH), who have atypical exposure to androgens early in development, and their unaffected siblings (total N = 125 aged 9 to 26 years). Females with CAH had more interest in Things versus People than did unaffected females, and variations among females with CAH reflected variations in their degree of androgen exposure.

Results provide strong support for hormonal influences on interest in occupations characterized by working with Things versus People."

OK, in the paper you linked to presumably they've classified IT/CS jobs as 'thing-oriented' work. Now, let's classify them as 'people-oriented'. Whoops! Suddenly women are very interested.

You see the problem? The study is biased from the outset because it's designed to 'prove' that women aren't interested in IT.

> The study is biased from the outset because it's designed to 'prove' that women aren't interested in IT.

No, the study was designed to investigate a potential link between prenatal androgen exposure and a preference for Things vs People work, which it did find.

> let's classify [IT/CS] as 'people-oriented'. Whoops! Suddenly women are very interested.

But precisely, observation shows they're not, as a group ! That's the very thing we're here discussing, trying to find out why.

And yeah, let's just arbitrarily switch jobs classifications like that without any grounding in reality, just because that's the only way your hypothesis will not fall apart.

Can you see the extraordinary amount of unlikely premises it takes for your logic to stand, compared to the alternative, simpler explanation ? That is exactly the kind of situation Occam's razor is here for : "among competing hypotheses, the one with the fewest assumptions should be selected."

This is starting to feel like a flat-earth debate, honestly. I'm pretty sure many people here would engage with the available evidence much more candidly, if this were a less politically charged topic (and I'm a liberal. But science before ideology, always.)

'Arbitrarily' switch job classifications? Sorry, what is the reality grounding of classifying IT jobs as 'thing-oriented'? It seems about as scientific as the zodiac. Tell me, if you're a programmer, how much time do you spend communicating with people versus actually coding? How much time reading and writing documentation? Choosing variable and class names to communicate your code's intentions?

Can you see how classifying IT jobs as something women aren't interested in, leads to finding a 'result' that women aren't interested in IT jobs? That's literally a tautology.

the field of diversity

uh? What? What is that? Define 'diversity.'

The field of study of the causes of unrepresentative populations and the strategies for mitigation of those causes.

Is that a thing? or did you just make it up? Can you link a couple .edu departments.

Does this field just care about "unrepresentative populations" of people interested in coding, or do they also care about "unrepresentative populations" of olympic sprinters, fishing boat crews, and kindergarten teachers?

https://www.google.com/search?q=diversity+studies+site%3Aedu

Can you link to one that cares about the apparently highly important matter of "unrepresentative populations" of people interested in coding, and why that is an issue of greater interest than any of the other countless "unrepresentative populations"?

The one's I looked just seemed like a mish mash of meaningless postmodern babble.

What happened to the scholarly tradition of penetrating insight and clarity?

Research fields use field-specific jargon. Why is that surprising to you? If you look at a CS department page you'll fine plenty of content that makes no sense unless you have some understanding of the field, too.

Reply on news.ycombinator.com