Hacker News Reader

Comment by sethammons | Hacker News Reader

Because humans are pattern matchers. We have to strive to eliminate or reduce conscience and subconscious biases to ensure fairness.

I know intelligent and successful people who would never take financial advice from someone with a beard for example.

Edit: maybe an interesting example. I grew up around smart woman and girls. From observation, I “knew” women were smarter than men. I was in college before I heard that some thought that woman were not suited for math. I couldn’t even process that. But people see patterns and extrapolate based on their observations and what they hear, and often start to believe something that does not exist.

View on HN

> We have to strive to eliminate or reduce conscience and subconscious biases to ensure fairness.

Why? Does fairness serve a business purpose? Are businesses held to be fair to their employees in other circumstances?

What is fair? Perfect proportional representation of the host population? Within what limits? Can we make a heuristic for what's more fair? Should that drive hiring decisions more than other factors?

The way I've come to see it, so-called fair hiring practices are purely a social signal that companies are required, by the market, to hold - and only in the sense that they need to NOT appear to NOT be fair, in terms of bad press.

Other than that, companies don't seem to have any actual obligation to be fair, unless held accountable by some other mechanism (a union, the court of public opinion, government regulations, etc). They will only do so as far as it provides a favourable outlook for the company, and not further. When sweatshop-like conditions are profitable and the attention of the public is elsewhere, for instance, sweatshops you will get. It doesn't matter much if every race is slaving away in there together side by side if they're still slaving away.

Another reading of "We have to strive" is simply that, if we do wish to reduce various biases, it is something we must do consciously. That doing so runs against what's easy, our natural inclination is to extrapolate and generalize.

Deciding what fair means is indeed hard. But most would today agree (I think) that judging someone to be incompetent based on the reputation of those of the same ethnicity is unfair.

> if we do wish to reduce various biases

Again, though, the desire to do this is based entirely on marketing, and not on merit.

> judging someone to be incompetent based on the reputation of those of the same ethnicity is unfair

Sure, but mandatory hiring of people of particular ethnicities in order to boost some arbitrary "diversity quotient" (read: non-white quotient) is potentially just as unfair to other qualified applicants.

I work with exactly 0 black software engineers in NYC office.

I see it as a reflection upon America, rather than a reflection on my company.

The same company has black engineers in London.

When we seek new employees - we had 0 black applicants in NYC over the last 3 years. We have a small percentage in London.

"Fair" can mean a lot of things. If you asked three people to define fairness, you'd get at least three definitions, and I can think of some definitions that would indeed have a business purpose; I can also think of some that wouldn't.

An example of something that would serve a business purpose: widening the potential target market. Of course if your company is producing maternity wear then you probably don't want to aim for a gender-balanced customer base. But in general, increasing the size of your taget market can serve a business purpose.

There's no reason to assume that a diverse (read: non-white) group is required in order to market to a diverse audience. Plenty of 100% Chinese companies manage to market to everyone in America via Amazon etc. every day. Will there be some requirement or even some business advantage to them to hire diversely?

I can't see much practical value for diversity based on skin pigmentation intensity, in software engineering. (There's a good amount of research for gender diversity, though)

There have been studies on the value-add of diversity in the workforce. TL;DR: research "suggests that diverse work environments boost profitability, culture, and public perception.[1]"

1: https://www.purdueglobal.edu/blog/careers/how-does-workplace...

Research that validates the only conclusion you're allowed to publish isn't research I'm going to base my life around. The famous Putnam study on diversity in neighbourhoods is much more realistic:

https://en.wikipedia.org/wiki/Robert_D._Putnam#Diversity_and...

Society exists to serve people, not businesses. Businesses are a tool to that end, so business purposes are only relevant when they don't run counter to human needs or cause human suffering.

If you want to have good objective judgement you must remove or at least be aware of your biases as much as possible.

What's a bias, versus a well-trained heuristic? Am I to deny a million years of evolution as a pattern-finding organism just to satisfy the political goals of a group of people who do not have the best interests of my own group at heart?

All living things use pattern matching - not just humans - as a surviving mechanism. It allows fast decision making.

Which model do any of us prefer in most situations: one that allows us to be right more than 50% of times done in miliseconds, or one that takes hours/days/weeks (whatever) and allows 95% of good (or fair) decisions?

The appropriate response time depends on what is being decided

When it comes to deciding who I will be working with I would rather take the time (hours/days/weeks) so that that we make a fair (un-baised) decision

As for deciding to whether to evade a baseball being thrown at my head, I would prefer the faster response time. It okay to be biased against fast moving projectiles :-)

Any time you deal with people you need take the time needed to eliminate any biases that are common to many humans

User story: Jane works in HR for $COMPANY. They have 10 places on the graduate scheme this summer, and have already received something like 800 applications. The boss wants a shortlist by Friday - they're going to read each one on the shortlist in detail and make a fair decision, but they're not going to spend half an hour on each one in the 800-pile because that would be 400 person-hours.

Of course, first you throw out the obviously bad ones - anyone who sent a word document when they clearly asked for PDF; anyone who doesn't have an engineering degree etc. etc. That still leaves you with 500.

What do you do? This is I suppose why implicit bias training exists, so you don't go "this one's from India, the last 10 from India couldn't even spell the company name right, reject".

(The next step in the UK is, reject everyone who doesn't have a first-class degree, if that still leaves you with a long enough shortlist. No wonder students are anxious about getting a first!)

Point I'm trying to make: there's situations in hiring where the time you can take is influenced by the volume of applications.

I think the parent was referring to survival instinct. Millions of years of evolution has lead to the need for fast pattern matching to avoid danger.

> Because humans are pattern matchers

Agree, but then you would think that engineers (when it comes to working with other engineers) are a little more rational than the average folks and can go beyond superficial observations.

I would think that engineers would also respect that cognitive operations have a cost just like computer operations.

If a cognitive bias lets you arrive at a conclusion 10 times faster and is right lets say 90 percent of the time, is that strictly worse than being super fair and objective as the thinking pattern you employ? I think if you had a process at your job that is 10 times faster and was what right 99.9 percent of the time you would absolutely be for using that process over the one that reaches a 100 percent accuracy rate but was vastly slower for anything that isn't fatal to your career.

Large parts of society seem to reward thinking fast and making quick decisions, and would prefer a quick decision even if its wrong. We have 2 pressures here. Making the decision quickly, and then making the "correct" decision. Your circumstances will determine which is correct at the time. But saying to always stick to one and avoid the other doesn't seem super rational to me.

You just matched all engineers to a pattern you observed.

That's brilliant! (as is the observation that humans are naturally pattern matchers)

In fact, I would argue that engineers tend to be people with stronger pattern matching capabilities, because that's exactly how most engineering problems are solved.

But these are not 'superficial' observations. People of Indian ethnicity are more likely to be the product of (tier B) Indian educational system (which is based on Brahminical rote memorization), Indian culture (always say yes to your superior), Indian incentive structure ("how many people are you managing?").

The last job I worked at, more than 50% developers on the floor were Indian (like either fresh H1b or intra-company transfers), so I noticed that I would get 'lost' in the crowd. A non-Indian person will immediately notice other non-Indian person, and ignore me. It wasn't their fault because most Indians on my floor never bothered to say 'hi' or 'bless you' or even acknowledge anyone who wasn't Indian. But then, I'd also ignore other 'pattern breaking' Indians (until, like me, they made themselves stand out).

> > Because humans are pattern matchers

> Agree, but then you would think that engineers (when it comes to working with other engineers) are a little more rational than the average folks and can go beyond superficial observations.

You are saying pattern matting is irrational? If one is to apply pattern matching in an absolutely rational way, they would not make abitrary judgements that some traits are/are not superficial. They should just take the fact as is, then make their hypothesis, then prove or disprove it then act upon the result.

Engineers tend to be more empirical and less idealistic, not to mention not uncommonly somewhat autistic.

They are more likely than the average bear to apply Bayesian probability, which comes alarmingly close to confirming stereotypes in most cases.

All in all no-one is beyond bias. Most of interview judgment comes from system 1 thinking. Meaning you feel some way about the candidate. And you only use system 2 to justify what you are feeling. So you use system 2 for cherry picking instead of determining whether candidate is good or not.

That could not have possibly be born out of observing engineers ...

Indeed. Categorization is how the brain works. On purpose or not, if you have two people with the same resume your initial impression of "Steve Smith" is better than your initial impression of "Bubba-Joe Willum Jr."

Reply on news.ycombinator.com