[dogruck]

Ok, let's think about the compensation process.

You start with the obvious -- managers assess the value of each employee. Roughly, for each employee you derive a dollar amount that is: 1. Larger than what the employee could be paid elsewhere (otherwise, the employee leaves) 2. Close to the employee's contribution to the company's bottom line.

That's a rough science though. #2 is certainly harder than #1.

So, next, you put all that data into a big database. Then you run a variety of sanity checks (aka formulas): 1. Large changes in comp, year over year 2. Discrepancies when broken out by factors including age and gender.

Finally, you're faced with a choice. If you override all of your initial estimates with the formulas, then you have formulaic comp. Otherwise, you're at risk to lawsuits.

Imagine a GOOG executive defending "yes, we paid women less because we honestly think, on average, the male employees contributed more to our bottom line." That's not gunna happen, which sorta leads to formulaic comp, no?

[geofft]

#1 directly reflects biases from other companies, which could arise for lots of reasons. If, say, many of your male engineers are getting offers from Uber and the rest aren't comfortable applying, and Uber is giving extremely high comp to lure people away from Google, then your formulaic comp will end up with men getting consistently higher salaries.

Google is pretty opaque about how they make salary offers, but from reports on the internet plus my own experience and that of friends, it seems like they have rough comp bands within each level, and they don't really "negotiate" in the conventional sense, but they do have lots of leeway to match/exceed competing offers or your current salary if you name them before the initial offer. So if you have those offers on hand from other companies, or if you had a particularly strong current salary, you can get a much higher offer.

Also, don't forget that this lawsuit is specifically alleging that they underlevel women, not that they're directly paying women less. There are lots of easier ways for an executive to defend that, e.g., "like everyone else in the industry we'd love to hire qualified women but it's a pipeline problem" etc. etc. (The executive might even genuinely believe it.)

[dogruck]

I agree with everything you said, and your experiences match my own. I still say that situation leads to formulaic, systematic management.

Let's say that GOOG developed a sophisticated machine learning algorithm that made all hiring, firing, promotion, and compensation decisions. The code is open source, and everyone can see that it doesn't contain explicit logic for bias.

Now, a short, 44 year old, male engineer sues because of a statistically significant observed bias against one of his cohorts. Is the program biased, or just insightful?

With humans, it seems we have no choice but to assume their collective algo is biased. And it often is! But, when you're a massive monopoly with tons of cash, the safest thing is to make formulaic decisions that are statistically clean. It's just good business.

[geofft]

This seems essentially like Searle's Chinese room. The program is not biased, in the sense that there is no line of code that adds bias, but the application of the program to the available data clearly produces a biased process.

We've seen this exact thing with other machine learning algorithms - there was the one in the news recently that insisted on classifying a photo of a man in the kitchen as a woman, because all its training data firmly convinced it that women are the only people in the kitchen.

I guess the thing worth explicitly asking is whether biases for entirely logical reasons are defensible - for instance, if you start with an industry where (for whatever reason) men are paid much more highly than women, it's okay to offer people their current salary + fixed delta to poach them from their jobs. I would say no, because the goal of legal policies like this is to achieve a specific result in society, and specifically not to punish bad people in charge of companies, so the question is not whether people had a bad motive, but whether the result is being achieved. If you're allowed to apply a non-biased algorithm to biased starting data and have it yield an equally-biased output, you're not actually solving the bias.

[dogruck]

What if we observe differences in said algo(s), when applied to small companies and the Googles?

Said another way, if Google's employee stats are biased, is there any feasible defense?