Diversity quotas suck. Here’s why

That is actually a helpful reminder - the context of the conversation is a blog entry that struggles over the objectives of quotas and the utility of using race and sex as a proxy for diversity quality beyond... race and sex.

> Presenting a solution, or an attempt at a solution, is an expression of one's thoughts.

Yes... are you trying to defend your prior conflation of behavior and thought?

> Example 1 ... do you hire for more general diversity of thought and expect more overhead to train people in COBOL?

That depends entirely upon your organization's capabilities and priorities. Hypothetically lets say that the bank's long term objectives don't include a migration from COBOL, the IT department doesn't have a long history of successful inhouse training, and being a bank - is generally risk averse. First, filter for candidates that demonstrate an acceptable level of competency in COBOL. Second, filter for candidates that have skills and interests that are not organic to your team - but could feasibly be useful (in your mind, that is all we've got). Third, ask the candidates for examples of times that they've come up with novel solutions to difficult problems. Here is a personal example: I once accidentally landed a contract when I was having lunch with a friend and his boss, I was later told the clincher was my long exposition on fault tree analysis in ballistic missiles. The contract involved the integration of time management and security systems.

You seem to be struggling with the prioritization of diversity of thought over skills that are actually need to perform a job. Here is a hint: the first order of business is getting somebody who you imagine can do the job (skills, job history, etc). If you have more options that positions, of those people, select the one who demonstrates the ability to reason in a way unique to your team.

> Example 2.

You hire the guy who points out that you did a poor job of framing the problem. Not only did you describe it in a way that could be interpreted to demand an implementation that spits in the face of POSIX utility conventions (newlines as argument delimiters), but you also failed to establish a success metric (time, maintainability, performance, etc).

> Surely that's even more diverse thinking - and completely within the test protocol as given.

You described a bunch of potential implementations, not different ways of thinking. The protocol you gave wouldn't be useful to measuring diversity of thought, unless you modified it to include the possibility for interviewer-interviewee interaction, where you might get some clues about their thought process. "What is the success metric?", "Are the sorted values bounded?", "Is the source untrusted?", "How does this fit into the larger process flow?"

> That is, should I do something which I know is less maintainable simply because I know it's more obscure and thus shows my diversity of thought?

No, even in the cartoon funhouse of an example you provided - they might already have a Python weirdo running amuck, you'd add nothing. You don't know that ahead of time.

Where do I do that?

You wrote: "You seem to be struggling with the prioritization of diversity of thought over skills that are actually need to perform a job."

First, I am struggling over how to identify 'diversity of thought' during the interview process. Is it something different than "can come up with innovative solutions" or "out of the box thinking" or "creative problem solver"?

Eg, freedomben writes that 'diversity of thought' can include "different socio-economic status ... different political views, different learning styles"

While your focus is only on "the ability to reason in a way unique to your team".

These seem like two different interpretations of that phrase, and I lean towards ubernostrum's down-voted comment that "diversity of thought" seems often used as a euphemism for "put up with assholes" (my interpretation).

You wrote: "You hire the guy who points out that you did a poor job of framing the problem."

Which means you end up biased towards rules lawyers. Which may be what you want, but bear in mind that you are presenting one performance goal while you have withheld a secret goal that you are actually looking for.

Here's a less secret goal: the test is meant to see if you know what modern C++ is like (and not simply the C-ish subset), and if you have a good idea of what the POSIX mindset is like (so you don't end up asking pointless rules-lawyer questions).

Which of these possible secret goals should the interviewee try to optimize?

You commented "unless you modified it to include the possibility for interviewer-interviewee interaction".

Well, yes. But no matter what scenario you come up with, the essential problem remains - does "diversity of thought" differ from "highly competent and creative problem solver"?

You wrote "newlines as argument delimiters". In my description, the first argument is a filename. The contents of the file are a set of lines, terminated by a \n. The goal is to sort those lines by byte value. Otherwise the desire to sort the contents makes no sense.

"I assume 'diversity of thought' refers to expressed thoughts..."

> First, I am struggling over how to identify 'diversity of thought' during the interview process.

I think we've covered this pretty well, so your struggle isn't over the "how" - but the "why".

> Is it something different than "can come up with innovative solutions" or "out of the box thinking" or "creative problem solver"?

It is no different, with one exception: it is measured in relation to your existing organization. If all your programmers are proponents of the functional programming paradigm, hiring another Haskell programmer, while relatively novel to the rest of the industry - likely does little to increase your organizations diversity in thinking (without additional screening parameters).

> These seem like two different interpretations of that phrase...

They are: one is selecting for proxies, presumably as a shortcut. The other is directly addressing what is desired. I'm always amazed at how proponents for such selection mechanisms are totally oblivious to how ridiculously prejudice it is.

> ...I lean towards ubernostrum's down-voted comment that "diversity of thought" seems often used as a euphemism for "put up with assholes" (my interpretation).

Clearly.

> Which means you end up biased towards rules lawyers. Which may be what you want, but bear in mind that you are presenting one performance goal while you have withheld a secret goal that you are actually looking for.

Boom, point proven. You have at that point learned something about that candidate's way of thinking, select on it or don't. If that was a hidden goal then it worked, if it wasn't then disregard.

> Here's a less secret goal: the test is meant to see if you know what modern C++ is like...

Not a diversity of thought test.

> ...and if you have a good idea of what the POSIX mindset is like (so you don't end up asking pointless rules-lawyer questions).

That is a diversity of thought test. Is the candidate willing to, in the face of ambiguity, insert his own opinion instead of speaking up?

> Which of these possible secret goals should the interviewee try to optimize?

As I said earlier: the one that, in your experience, indicates an ability to do the task that the candidate is hired for. Should there be a tie, the one that arrived at a solution that you did not anticipate. And no, that doesn't mean you should hire a guy who insists on sorting files using Node.js, just because he was the only one... you can't determine rationale based only on language selection. If he does it in shell script, ask why. You may learn that binary compatibility concerns are higher on his priority list due to some past experience that you wouldn't have considered.

> ...the essential problem remains - does "diversity of thought" differ from "highly competent and creative problem solver"?

Yes, but first I'll point out that "highly competent" is an unrelated concept. Because effectively nobody has unbounded useful creativity over an entire problem domain, you want to select individuals who's constraints overlap as little as possible - thereby covering more of the problem domain. Personal example: I took over a database from the engineering department because the guy maintaining it retired. I was horrified when I looked inside, the architect was obviously a PLC programmer - using triggers and views to form a hellish logic ladder. I was trying to figure out how I'd be able to untangle everything into a more tradition normalized database when I got a request from one of the engineers to insert a new trigger to account for some upcoming process change, I told him how much I didn't want to do that and asked if I could spend a day with him in order to get a better grasp on their problem domain - he was annoyed, but agreed. Well it turns out the old engineer wasn't totally insane, the problem domain was pretty much unbounded and constantly changing: new metrics, new datatypes, new requirements - totally normal for their department. Whereas I would have used a Domain-key normal form in that situation, he just created a new table and added a trigger. Both styles work, each has different weaknesses and strengths. Neither of us would have arrived at the other's solution. Diversity of thought.

> In my description, the first argument is a filename. The contents of the file are a set of lines, terminated by a \n.

Lol, even your attempt at clarification adds confusion. The ambiguity this time: is the set of lines terminated by \n, or is each line in the set terminated by \n, or is the file terminated by \n? Yes, I'm pretty confident that I know what you mean - but I have seen 0x1f used for stuff like this in production, bad things would have happened had I just made an assumption.

> Otherwise the desire to sort the contents makes no sense.

Sure it does: if instead of interpreting the explicit mention of '\n' to mean line termination within the file, one interpreted it as the termination of the first parameter. Again, one can guess the intent due to convention - but it is ambiguously phrased and immediately led to two competing possibilities of intent in my mind. Maybe that is because I think differently from you...

... these are all examples of places where bias is rampant. Banks, courts, government assistance, insurance premiums, and even medical care are not delivered to all individuals equally. Not at all. Not remotely.

Through activism and lawsuits that these areas have made strides towards fairness... but you seem to be holding them up as an example of treating people as individuals “done right” or “done successfully”.

To briefly take these one by one:

> Banks lend out money based on it.

- banks had to be forced to do this.

> Courts make judgement based on it.

- this is laughable. The legal system in the US punishes African Americans more strongly that Caucasians for the same crimes.

> Government gives out social support based on it.

- maybe?

> Insurance rates and medical care is all about individual assesment

- doctors provide more medication for pain at the same level to white patients vs black patients. Proxies for race are all over insurance. Women had to pay more for healthcare until very recent.

Without active efforts that were made over many years to tip the scales and correct these bad outcomes, all of this would be much much worse.

> History does not dictate the future and it doesn't create any benefits and drawbacks unless we treat people based on category and class rather then individuals.

I respectfully suggest that the above statement is completely incoherent with reality, and represents weird wishful thinking.