Hacker News Reader

Comment by fmitchell0 | Hacker News Reader

I understand the tendency to conclude 'that is great, but are they qualified b/c that is what matters' logic. It's something you (especially if you were educated in America, like me) have to consciously unlearn.

For those that want the science and math behind diverse teams (as I've given this talk numerous times), there are bunch of links on the last slide of my deck: http://slides.com/fmitchell/how-math#/10

If you want to see the actual talk I gave at O'Reilly, it's here: https://www.youtube.com/watch?v=02HS9wjgfQY

View on HN

Hi, I tried to read the two scientific articles present in that list of sources (the rest is opinion pieces and discussion related to the articles); unfortunately, the link to the first one (the study) is broken (as well as some of the others).

As for the second article, please correct me if I am wrong, but from skimming it I believe that first of all, it speaks of diversity in knowledge and problem solving approaches, rather than race, gender, age and other identifying features. In fact it remarks the following:

"The claim that perspectives and heuristics may be influenced by race, geography, gender, or age has much to recommend it, as does the claim that perspectives and tools are shaped by experiences, training, and preferences. However, even when applying our result to those cases when identity diversity has been shown to correlate with functional diversity, we need to be acutely aware that identity-diverse groups often have more conflict, more problems with communication, and less mutual respect and trust among members"

The other problem I see is that this model (from what I gathered) translates into software development terms as something like: 10 top C# developers are not necessarily better at writing a C# backend than, say a group of, 5 very good C# developers, 2 top C developers, 1 very good Python developer and 2 top Haskell developers. Rather than mix and matching experienced and junior devs because the juniors are more "diverse" in one aspect; as is encouraged and practiced by diversity evangelists.

This makes me further cement my position about the diversity question, and that is to just let people find their vocation on their own; the Internet has solved information access, we live in mostly free societies and we can learn on our own. Introduce people to CS? Absolutely. Artificially incentivize career choice by changing hiring practices? Bad idea.

I didn't realize the links had expired.

Here is a page linked to the research and modeling I ascribe in my talk: http://sites.lsa.umich.edu/scottepage/research-2/diversity-r...

Specifically:

http://sites.lsa.umich.edu/scottepage/wp-content/uploads/sit...

> it speaks of diversity in knowledge and problem solving approaches, rather than race, gender, age and other identifying features

The key piece is how you connect that statement to your next statement of what defines 'top' and 'very good'. Your heuristic is coding output. Code and programming has context. In most cases, it's a business context.

The point is value can be derived from experiences uniquely tied to race and gender, especially given how we're taught race and gender. It often doesn't coincide with reality.

That intrinsic value, of lived experience, shapes thought. It shapes how people approach problems. That value has to be tapped. Those evaluating talent (and its potential) often cannot 'see' why it should be a factor.

This is a limitation of the evaluators' local optima (referenced in the research).

"The point is value can be derived from experiences uniquely tied to race and gender, especially given how we're taught race and gender. It often doesn't coincide with reality."

I can barely understand what you are saying (what does "coincide with reality" mean? What use is it for a company if it doesn't coincide with reality?), but perhaps you could clarify it with a couple of examples?

Also, isn't that argument a bit treacherous in that it implies that "diverse" people are different after all, meaning if they don't get ahead in some environments it might actually be because of their different ways of thinking, and not because of bias they face?

Edit: I was asked once in a tech interview to prove the theorem of Pythagoras. Yesterday I was in a museum for kids that also featured that theorem, so I feel inspired to use this classical problem:

Could you provide an example on how a woman, a woman of color, a white man, a black man, a latino, a gay person (or any other "diverse" person you can think of) would prove that theorem in different ways, unique to their race and gender?

Of course having the qualifications on paper isn't all that matters. But limiting yourself to a palette of people fitting the "protected groups" isn't what you need to flesh out a good team, either.

Wouldn't it be strange if, despite all the claims that race and gender, etc., don't matter, it really does matter as EEO regs keep telling us?

But people differ in ways other than skin color, gender, religion, age, physical ability and sexual preference. Members of our team from the city versus the country; short versus tall; early-risers versus nightowls; family size; headbangers versus country music fans; vegans or steak-eaters; and so on.

Why should we think that it's critical to have a pastiche of races, genders, etc.; while simply assuming that all the other interesting ways that people experience life differently is irrelevant?

> while simply assuming that all the other interesting ways that people experience life differently is irrelevant?

I don't know of anyone who values diversity in gender/religion/sexuality/etc but draws the line there, and I know a lot of very unusual people from all walks of life. That would be a weird line to draw. A gay genderfluid punk rocker probably has lots of experiences a biromantic genderfluid cellist would be fascinated by.

Diversity is good. The more intersections, the better. Why do you think anyone would draw this line?

Because some things are a choice, and some things are not. Some things can be turned off and some things cannot.

Gender (and gender identity), along with race, are things that cannot be unshed.

The language you choose to speak, what music you listen to, and whether you choose to share your sexuality, while they are differences, they cannot be instantly detected.

The natural heuristic we use drawing upon our experiences in an American society have a unique relationship with gender and race. It is especially because those things are non-choices and instantaneously recognizable, that they are the most important to address first.

This is a better answer to the question in the post I replied to. I was responding to CWuestefeld's faulty belief that people who want diversity in involuntary attributes don't care about diversity in others.

Because some things are a choice, and some things are not. Some things can be turned off and some things cannot.

This heuristic doesn't accurately predict which attributes are protected (legislatively or otherwise) and which are not. There's a pretty much open-ended set of predetermined attributes that can be discriminated on. Start with height and physical attractiveness: these factors are more-or-less immutable, but are well known as drivers of discrimination.

Why do you think anyone would draw this line?

Because the line has been drawn. The government has told me that there's a short list of protected attributes. And my employer (and thus by extension my hiring and compensation decisions) is held accountable based on those and only those.

And in the discourse on the topic, we always hear "why aren't there enough blacks in tech" or "why do women get paid less", but not a peep about the achievement gap for punk rockers.

https://news.ycombinator.com/item?id=13563855

I think science might be overstating it a bit. There are some indicators for some things. For example the first article unfortunately doesn't include a link to the actual paper, so we don't know what tests they used on the groups. I am willing to bet that for specific problems, qualification trumps diversity. For example, to build a new nuclear reactor, putting ten nuclear physicists together would yield better results than 10 random diverse people.

Also, seeing as recently a major share of psychology papers has been found to have serious methodological flaws, I would take this one with a grain of salt, too.

The thought that qualification has a direct correlation to perceived binary problems, like building a nuclear reactor, is exactly the reason why this problem exists.

I reference it in my talk (not on the last slide), but I invite you to read the story of Valeria Aurora (it's been referenced on HN before): http://bit.ly/1WZiNYQ

Her unique experience perspective to a mathematical problem is quintessential. The same is true for the women featured in the Hidden Figures movie.

Concluding that 'hard science' and 'hard math' realms don't benefit from diversity and only knowledge is why tackling this problem is so hard.

Ugh - come on, you seriously suggest only a feminist would have listened to a fellow human being and have been able to come up with a solution? She (Valeria) framed it that way because she is a feminist. That is just an anecdote, not science. There have been thousands, perhaps even millions of bugs fixed in Unix over time. What makes this one so special that only a feminist could have fixed it?

Likewise, you can not prove things with movie references. I haven't seen "Hidden Figures", but it was obviously made to make a certain point, so they would insert stories "proving" their point.

Reply on news.ycombinator.com