Hacker News Reader

Comment by derfnugget | Hacker News Reader

does anyone care if they perform better or does even suggesting that make me a racist in the modern world?

If you read the complaint, the allegation is that controlling for performance evaluations, members of certain groups were preferred over others for promotion.

I don't have any insight, but given the culture that google projects, I would be skeptical of the integrity of their performance evaluations.

Why would group averages be helpful when making staffing decisions about individuals?

He didn't say so

That's obvious, no?

If group X on average performs better than group Y, then objective hiring will lead to more group X bring hired. Then group Y takes you to court for discrimination.

I don’t think it’s obvious that summary statistics will be helpful unless they’re particularly carefully done. Where do averages come from? If individual data points are biased in the same direction (that is, not noise that cancels out) then the group average will be too.

This isn’t something you can just assume when you see someone quoting statistics. It could be a garbage study.

Is the Google applicant pool a representative sample of the group at large?

It really depends on what assumptions you are making, your basis of comparison, and how you measure performance.

Does X perform better than Y in general or within the community. Does hiring match national population, the applicant pool, or the top 1% of the applicant pool? How do you measure performance?

These topics are rarely fleshed out in any public corporate policy. All I know is my bonus depends on increasing the % of minority employees.

They can explain the "disparate impact".

They generally aren't. See the ecological fallacy.

Exactly. You can't just simply assume that job performance is statistically independent of various seemingly unrelated traits. If you suspect age discrimination, you also can't just assume that age is uncorrelated with performance. Or being short sighted, or even things like weight or height. They may be uncorrelated, or they may be correlated.

Job performance as measured, yes, it already accounts for all biases/traits, including age, appearance, personality, performance, race and all other known/unknown biases that the people measuring the performance have.

Fun fact: You can almost always score at least $10-20k if you're fired from a job if you try hard enough.

I've been on the employer side of this... you fire someone who's performing badly, and then they come back 4 months later and sue the company for [insert made up thing here].

In our case, an ex-employee is suing us for not accommodating an anxiety and migraine disability, which they never disclosed and never requested accommodations for. So now we face a discrimination lawsuit (from a non-minority) based completely on falsehoods and things that never happened.

The reason people do this is because it works! Employers will almost always settle before it goes in front of a judge in order to avoid the hassle and cost of defending the claim.

It works, but court cases are public record. Good luck getting anyone to touch you with a ten foot pole afterwards, not like the candidate can prove why they weren't selected.

So HR pays to check this for every candidate? No wonder it's hard to get hired.

Lol in my state I can see every (unsealed) civil case with a simple online query. My landlords have pretty much all done it, I don't know if the background checks show it but it's an extra 15 seconds to hedge a 10k+ potential liability.

Though this problem seems to be mostly restricted to the US American legal system.

I think it matters, because minorities experience a lot of ways in which products fail them that probably would not have happened had their needs had been represented and prioritized. If a minority group is not represented in the development of a product, their needs are more likely to be neglected.

Products are developed for a specific market. If the market of the given minority group where large enough, there would be special providers only for that. If I go buy an Italian Pizza place, I expect to receive an Italian Pizza, and not an Italian Pizza with a Chinese nuance to it. That's when I go to a Fusion place. It's by definition impossible to make product that satisfies everyone.

I agree that it's impossible to build a product that satisfies everyone, which is the issue with global tech companies. In my utopian world, we would educate people and develop tools so that communities can build their own tools to cater to their own needs (in the same spirit as unix). I think that's better than what we have today with technocrats in Silicon valley dictating how tech should look and function for the rest of the world.

I don't think I disagree, so this is perhaps a devil's advocate argument--to the extent a product is meant for somewhat general use, by integrating more perspectives during development, we might uncover blind spots and innovate in ways that resonate more deeply with their core audience, not less. Asking, “Whose needs might we be overlooking?” could be useful not because every minority requires a bespoke solution, but because overlooking them might mean missing opportunities to serve even the majority better.

From anecdotal experience with voice recognition software: early versions struggled with accents and also required training on your voice specifically, which limited their utility. Making models more flexible didn’t just help minority users with accents—I think it improved accuracy for everyone. Similarly, curb cuts on sidewalks, originally designed for accessibility users, now benefit parents with strollers and even those food delivery robots running around some cities.

Maybe one frame is to avoid unintentional exclusion? The pizza shop isn't obligated to, but could at least consider the fact that some people don't eat meat (or pork or whatever), and therefore keep the margherita on the menu to the benefit of everyone.

It's obviously lacking a lot of controls if you arrive at the conclusion. Mostly just shows you are missing of a lot of context.

Your statement implies the assumption the minorities in this class action are underpaid because they perform worse than white and asian employees. I'm not sure that helps you not look like a racist.

The article doesn't go into details, so it's probably a safe bet not to make these sorts of assumptions at all.

Psychology has determined there is no non-racist way to evaluate performance: https://news.ycombinator.com/item?id=43407157

Reply on news.ycombinator.com