Ask a Female Engineer: Thoughts on the Google Memo

I deeply disagree with this approach. You're essentially saying that unless you can come up with an alternative scientific theory, complete with predictions, it's not possible to criticise an existing theory about the world.

There's many plausible explanations why an absence of a 50:50 gender representation could be caused for reasons other than bias or average ability. That's enough to put a nail in that model of discovering bias. Coming up with a way of predicting what the right ratio is, isn't necessary to discard that metric.

I think part of the problem is what the memo says and what it doesn’t say. It’s entirely plausible that the ‘natural’ ratio is not exactly 50:50. But as of last year, among Google tech workers, the ratio was 81:19, and that’s with all the affirmative-actiony programs Damore wanted to back off on; in the past it was higher. It’s quite a bit less plausible that intrinsic differences could explain all or even most of that big a discrepancy, especially combined with the many anecdotes of discrimination we hear about. Now, to be fair, the memo never explicitly claims that it does; indeed, at one point it specifically says “in part”. But the tone of the memo, the relative lack of time spent acknowledging the large role played by cultural factors, makes it sound like Damore thinks the natural ratio is at least pretty close to the current one. And that’s simply wrong.

That's a straw man. You're suggesting I disapprove of criticism, which is not so. I disapprove of demands for policy change when you don't even have a hypothesis for what your target should be.

Unless Damore (or someone else) can reasonably estimate whether their theory around 'biological' differences result in a natural 10/90 ratio or a natural 49.9/51.1 ratio then there isn't really a case to be made to change actual real-world HR policies on that basis.

Being able to reasonably estimate that 'natural' ratio is a massive task. You'd need to account for parenting, education, popular culture, socio-economic group, dozens of biasing factors. I'd expect that model to go well beyond what's possible.

Yes, that may impose a high hurdle on criticism of HR policy via this argument, but that's also the intellectual leap that Damore has claimed to have made from the evidence presented. How exactly he's managed that leap is problematic. He certainly hasn't demonstrated full knowledge of all of the factors involved.

> There's many plausible explanations why an absence of a 50:50 gender representation could be caused for reasons other than bias or average ability. That's enough to put a nail in that model of discovering bias. Coming up with a way of predicting what the right ratio is, isn't necessary to discard that metric.

Of course, and it's certain to be a combination of factors, some historical, some current that pushes representation away from 50:50. I don't think anyone is pretending that bias alone is responsible. But there's a mountain of direct evidence that bias is a significant problem. On the other hand the chasm between this biological source evidence and an actual hypothesised effect on representation is vast.

You may ask why it's 80-20 among CS graduates. One hypothesis is that women are just less interested in tech and in presence of many other choices they choose different paths. In the past there weren't as many choices that's why women were forced to go into programming (that's why there were more women in programming several decades ago).

I agree with your argument but fail to see how it allows you to defend a discriminating policy. It's the other way around: You can't discriminate people without evidence that what you are doing is reasonable. You're the sexist in this case.

You can't defend a discriminatory policy by saying you understand it's discriminatory but to keep it because no one can tell how much.

This line of reasoning is inconsistent unless you are only opposed to discrimination of some groups. In that case I think we sadly have to agree to disagree.

While this sounds reasonable on the face of it the reality is different.

Where are you sourcing these graduates from? In the USA computer science departments are barely above 10% women faculty, in China it's closer to 40%. Student numbers tell a similar story... so it matters where your graduates are coming from. For a multinational like Google this is a real question.

> If you force it to be say 70-30 then you are discriminating against men based on sex.

This is a loaded statement, based on the assumpions that (a) hiring if left alone is broadly meritocratic and (b) quotas are the only game in town. There's enough evidence to say that neither of those assumptions is true.

First of all, it's been proven many times that bias in hiring is a real problem and has a large effect. Hiring is not meritocratic. Second, Google doesn't use quotas, no bar-lowering occurs (Damore hinted at this but gave no specifics and no evidence... we have to reasonably discount it unless someone can prove otherwise). Instead diversity programs mainly exist around sourcing and trying to avoid false negatives in order to counteract systemic biases.

(Disclaimer, I work in this field and have written on this topic before: https://medium.com/finding-needles-in-haystacks/we-need-to-t...)

[edited to remove some text from parent post accidentally left at the end]

Do you agree with my argument apart from whether it applies to you or not?

I'm not a fan of the trend for sorry-you-feel-that-way apologies. On the other hand it's possible I let this seemingly-unending argument get to me and got defensive, thanks for not taking it badly. Suggest we move on. For reference (no need to explain) the trigger was "You're the sexist in this case" which I now assume was hypothetical rather than accusatory.

The words you're putting in my mouth is defence of specific policies. I'm not aware that I'm defending any specific policies.

One policy that's come up (not sure which thread, I've lost track and can't be bothered to reorient) is Google's policy (as I understand it) of ensuring 'diversity' candidates get considered, reducing the false negative rate. This was inaccurately described by Damore as lowering "the bar", which is quite inflammatory. That policy is designed to specifically redress two things; (a) decreased confidence in under-represented groups resulting in low numbers of applicants, and (b) unconscious bias in hiring processes resulting in fewer under-represented groups getting through.

While there are more elegant solutions (vested interest disclaimer here) this type of policy tries to address measurable issues and does not reduce quality of hires.

Perhaps it leaves fewer roles open for others, but ultimately you have to make a choice between Hire A benefitting from a diversity program or Hire B benefitting from hiring bias in their favour.

Is there a different policy you want to discuss?

More background on my post on this topic if you can be bothered: https://medium.com/finding-needles-in-haystacks/we-need-to-t...

Regarding your reply: I agree with most of your reply and enjoyed reading your blog post. I feel I understand your position much better now and can see where you are coming from.

> Is there a different policy you want to discuss?

I'd like to clarify whether we agree or disagree on the original argument - hypothetically, regardless of any specific policy. I hope I don't misrepresent your views in the following.

In your blog post you seem to argue that feelings of unfairness by the over-represented group in response to positive discrimination are built on a misconception [1]. My original reply to you was in the same vein and I'd like to understand where exactly we disagree on that.

I believe discrimination based on group membership is not justifiable. The only way in which positive discrimination can be justified is therefore if its application does not actually cause discrimination but only corrects for existing discrimination.

As we don't know for sure yet how much of the representation gap can be attributed to discrimination, we should not use positive discrimination to correct for it as we potentially do more than correcting for it but actually discriminate.

Hypothetically, if the split would be 45/55 in a perfectly just world, aiming for 50/50 through positive discrimination would in practice discriminate and not just correct for discrimination.

Please note that I agree with the outcomes of positive discrimination until the effect of the original discrimination is canceled out - I just don't feel we can distinguish both cases and should not dismiss feelings of injustice in response to that as "built on a misconception".

[1]

> In any discussion of positive discrimination there’s a risk that the overrepresented group (usually white men) may feel threatened. Unsafe. People aren’t born aware of their comparative advantage or disadvantage, and sometimes never see it, so when other groups seem to be given a leg up it can feel unfair.

> Feelings of injustice may be built on a misconception, but they still exist and are natural