[spaceheeder]

In the next paragraph the author specifies that they were trying to attract a wider cross-section of candidates. The hires wouldn't have been on-boarded if they weren't the best candidates who applied, but there's no way to know if they were the best candidates available when the signals sent by the job postings disincentivize applicants along axes that do correlate to race or gender but not to job performance. That's literally the entire point of the article, as far as I understand it.

[Paul_Dessert]

Read this paragraph:

"Finally, I suggested that if we reached the final round of hiring without any viable women or minorities in the selection pool, we should take this as an indication that we had not done a good job at outreach. We would start the recruiting process over, and try harder to attract diverse candidates."

What she is saying is, if she didn't get the results she wanted, she'd toss aside ALL applicants (even if they were highly qualified) and start over until she found her version of the "acceptable" candidate.

That's just wrong.

[spaceheeder]

Arguably so. I'm sympathetic to her goal and that in particular stuck out to me as an over-correction. It doesn't undermine her point that simply changing the job listings can dramatically impact the makeup of your talent pool. That particular provision ended up being unnecessary, anyways; 75% of applicants for the newly posted position met the criteria for diversity hires.

Your point seems to be that explicit bias against groups that are historically over-represented in certain professional fields is of equal or greater moral concern than implicit bias against groups that are historically under-represented. I think that both can be bad, and that reasonable people can judge the magnitude of the harm on both a systematic as well as a case-by-case basis: Prior to the author of this article conducting her live experiment, 0% of non-white non-males were hired for an entry-level position; whereas in the course of her experiment, 25% of qualified applicants were white males, which doesn't sound far away enough from their makeup of the US population to trigger any alarm bells in my brain.

Even if you think that bias-by-writ is always worse than bias-by-complacency, it's still fair for a devil's advocate to put it to you: If this is not the way to solve the latter problem, then what is? If your answer is some version of "leave well enough alone," then I will be unmoved by your heightened rhetoric.

[Paul_Dessert]

Where does the article say this, "Prior to the author of this article conducting her live experiment, 0% of non-white non-males were hired for an entry-level position"?

She's female. She got hired.

[spaceheeder]

Quoting you, quoting the article:

> It seemed that whenever we had our monthly engineering all-hands meeting, the new hire announcements featured three more young white guys

[Paul_Dessert]

That does not say, "0% of non-white non-males were hired for an entry-level position".

[spaceheeder]

No it does not, but it can imply it if you take it in the context of the whole article. One round of job posting led to three hires by the end of the article; should we assume that was a unique practice? If so, then the number "three" is an unusually arbitrary thing to mention in the quote you selected.

But that's neither here nor there as far as I'm concerned. LennyCrop would likely say that your attention to perfect parity in n=1 comparisons counts as moving the goal posts, and I'm inclined to agree. We don't have complete data for this company's hiring practices and it would be weird if we did; no matter what argument you want to make out of this article, some assembly is required.

I hope that people who read this chain ask themselves what they actually care about more: the source of bias, or the size of its impact. I've made my opinion known, but I'm willing to believe that reasonable people can disagree on that topic. It's a debate worth having. Challenging me on my Bayesian assumptions for reasons that are more parsimonious than anything else, on the other hand, is just a distraction.

But I know that minds that have already been made up are slow to change and are more likely to do so in private than in the heat of a debate, so I'll take this parting chance to plant one last bug in your brain -- a paraphrase of one that I used in a very similar conversation on HN just shy of a year ago:

If x% of the white male population is working in the field of software engineering, but only x% * y% of the non-white-male population is, and getting people interested in software engineering isn't a zero-sum game, then making y as close to 100 as possible can only be a good thing. Making a concerted effort to recruit diversely, it is hoped, will have accelerating returns as intentional involvement of diverse contributors will lead to more and more organic involvement of diverse contributors. Even if you want to argue that the reasons people have for deciding not to get started with coding are wrong or that those reasons shouldn't affect the industry for some other reason, the fact is that social patterns will self-reinforce if not corrected for and will ultimately have negative impacts on everyone by deterring more hands from coming on deck.

Again: If this wasn't the right way to begin to solve this problem, then what is?

[LennyCrop]

It's best to ask yourself, does someone who reads a single line of an article and sees that as the basis of the whole article a troll?

Please don't feed the trolls, they'll just move the goalposts and you'll be arguing about something completely unrelated in a bit

[Paul_Dessert]

A troll? The author is a hypocrite and I'm calling her out. The words above came directly from her article. And yes, I did read the entire thing.

If you don't see anything wrong with what she said/did, I feel sorry for you.

[spaceheeder]

Cursory posts -- whether they're posted by trolls or cynics or anyone else -- that are quickly read and just as quickly upvoted, but don't receive a thoughtful reply, tend to persuade the uninitiated.