If we're going to blame the employers, can someone show me that there's a disproportionately large number of unemployed minority engineers that are seeking work but not getting it?
Makes sense, but I'd say it should be a disproportionately large per-capita fraction of minority engineers with similar degrees, experience, geographic location, etc, relative to white engineers with the same characteristics.
For example, if minority engineers with Stanford degrees with 7-10 years experience and living in zip codes [A, B, C, ...] are 10% unemployed and their white classmates have a 5% unemployment rate, that could be evidence of deliberate discrimination. It's important to compare like to like, otherwise you can wind up with all sorts of weird conclusions.
The arguments for diversity usually say that organizations improve when people from different backgrounds are part of them. This is the argument for increasing diversity along gender, orientation, and ethnic lines. If different points of view help then we should be trying for ideological diversity directly.
The fact that Twitter enables diversity doesn't mean that they have or understand diversity. It may be that they simply haven't accidentally stepped on it.
If they don't know why what they're doing is working then they can't successfully improve it further, or avoid squashing it unintentionally and irreparably.
Honestly your comment reads very racist to me. "why do we need a non-Asian engineer?", "suppose I'm wrong, and black people actually do think differently. [if twitter is doing good] why do we think that "different" is actually better or even useful?"
So Twitter is doing great, and you've basically reduced it to either 'race/perspective never matters' or 'non white/asians are inferior'. I'm sure you'll say it's the former.
Wow.
(Now that I've re-read your comment, I'm happy with my down vote).
But yes, if different races do behave differently, it's valid question to ask whether you actually want that different behavior. As a silly hypothetical to illustrate the point, humans and leopards behave differently. Turns out one of them is a lot worse for the office environment than the other.
If you want to argue that A != B, you are explicitly allowing for the possibility that A < B. So some argument is necessary why that isn't the case.
But again - I think A == B, which I guess makes me racist.
Different experiences yield different points of view yield different ideas
Lets be concrete here, rather than appealing to vague platitudes.
As a Bayesian, I can certainly tell you how my perspective differs from folks like Leonid Pekelis or Evan Miller. As a person who leans towards parametric statistics and modelling, I can tell you how my perspective differs from the machine learning types.
Why is it so difficult to provide the (alleged) black perspective?
But software development is not just about CS theory. You're selling a product to people, and thus you introduce the human factor. If say, Facebook one day realizes they need to appeal to female users more to promote growth, does it not make sense that having female team members would be useful? That if Apple sees China as a growth market having Chinese team members will help them better target that userbase?
Obviously, there is no binary tree that is more friendly to the Chinese market, or black, or gay markets etc. But you can definitely change the UI, or messaging, or features that better speak to a specific culture. When architecting a feature in Facebook, a hispanic engineer could suggest a family-focused feature since she knows that family is very important in her culture. That's not to say a white male team of devs couldn't do a great job of satisfying a Hispanic user, but that a more diverse team might do an even better job.
That's a very business-centric take on the issue. Another, more noble side to it is that there's very likely a lot of people from underrepresented minorities who could be high quality software devs but because of their socio-economic status they were discouraged from pursing that career. A more diverse workforce, won't lead to quick results, but it helps.
But supposing these affinity effects are real, and extend beyond consumer products, then if your customer base doesn't contain these minorities then their presence on your product team might be harmful. I.e., MongoDB or Washington Square Tech should NOT hire a black guy, since a black guy's experience is significantly different from the (white/Asian) rockstar ninjas and banksters making up the customer base. Is this also a conclusion you would endorse?
If this director has a more human and compassionate perspective, she might help push back on all feature requests to give her team ample time to fix the fail whale during the day because she sees how unfair and stressful it is that they are getting paged at night.
This is hyperbole, but another director with a more analytical bent he might push his staff to work more hours, work harder and smarter, to rewrite the problematic systems.
Perspective of leadership is extremely important.
But this applies to all teams that work collaboratively on solving problems.
"Twitter as a platform has empowered underserved and underrepresented people. It has fomented social movements and..."
Sounds like twitter is killing it for people of all ethnic groups. It's as if a bunch of White/Asian dudes can actually design algorithms that work for everyone. So, um, why do we need a non-Asian engineer?
It's also worth questioning how a black person would think differently from the (apparently highly effective) white/Asian workforce. I've seen very few meaningful examples of this and I've never experienced it (I'm usually the only person of my race). My current job is mostly Punjabi's, no techies of my ethnic group, and my unique perspective is "lets all be Bayesian cause Frequentism is ass backwards" and "stop the multiple fucking comparisons!"
A naughty question: suppose I'm wrong, and black people actually do think differently. Given that whites/Asians seem to be doing such a great job, why do we think that "different" is actually better or even useful? In statistics terms, given two different functions f and g which are estimators for some truth t, it's unlikely that |f(x)-t(x)| = |g(x)-t(x)|. One of them is probably better.