Hacker News Reader

Comment by ggggtez | Hacker News Reader

Let's say you get "a" tries to succeed. You'd expect equally talented candidates x, and y, to have Pr(x|a) == Pr(y|a).

But what if y only had "b" tries, where b<a? What would y have to be to make Pr(x|a) == Pr(y|b)? Our intuition seems to say that there will be a lot of variance in Pr(y|b). But the candidate's skill hasn't actually changed, just we aren't measuring it with the same fidelity!

And that's all there is to it. If you no longer assume everyone has an equal chance to show their skills, then the meritocracy isn't actually working as intended. Good "y" candidates are getting ignored for worse "x" candidates, just because "x" candidates had more chances.

View on HN

I get the idea, but Mozilla is building things.

If I want the best engineer to build my bridge, I don't want to give bonus points for skin color or private parts. It's not like the trucks going over it will be any lighter or have smoother tires because the person building it grew up in a single parent household or something. When the rubber meets the road, some engineers are better than others and it affects the end product.

My fundamental concern is that the moment that the deciding factor isn't how good someone is at what you need them to do, you're driving a wedge through "the world as it is (cold hard reality)" and "the world as you want it to be (equal outcomes, diverse, empathetic)". The deeper that wedge gets, the tougher of a time you'll have meeting goals in the real world, where whether or not that bridge holds during a tornado is completely independent of the background of those who built it.

I think employees should be judged on how well they can deliver toward the company's goals, and for the most part, I think that happens. Actions like these are more about virtue signaling than anything else.

"When the rubber meets the road, some engineers are better than others and it affects the end product."

This is overly individualistic. A given engineer is not better or worse in some absolute sense. They are better or worse at a given job, which takes into account many other contextual factors: the specific goal, the rest of the team (who will be pursuing a that goal cooperatively), and the unique challenges which face that team/goal combination.

For example, it may be that your engineer Dmitri is best at solving particular types of design problems, but he only speaks Russian, and he's having trouble on a team with no other native Russian-speakers.

In this case, the "best" engineer to hire for Dmitri's team is a Russian, even if they're "worse" at solving design problems on an individual basis. In the abstract, hiring an engineer based on the language they speak goes against the meritocratic ideal, but in practice it is obviously the decision which best moves the team toward their goal.

More generally, one could say that even though language-diversity is not obviously or directly related to engineering effectiveness, it is still related. Having many languages represented on a team increases the odds that the performance of people like Dmitri will not be limited by a language barrier in the first place.

It's funny you mentioned "the world as it is" vs "the world as you want it to be". Because from my perspective, you're describing the world as you want it to be: each person has an abstract platonic "skill value" which can be objectively and universally evaluated, and which translates predictably and directly into metrics like load-bearing capacity. This is a lofty ideal, but does not reflect how organizations and teams actually achieve goals.

"The world as it is" involves orgs with lots of subjective, local, interdependent systems which often involve interpersonal dynamics, norms, and emergent social phenomena, all of which have an indirect but very significant effect on metrics like load-bearing capacity.

> A given engineer is not better or worse in some absolute sense.

I'll just address this first: Often they are. It is idealist and simplistic to claim otherwise. Some surgeons are better at performing surgery than others, some engineers are better at building bridges than others, and some programmers are better at programming than others. They're not inherently better people, but that's not relevant.

I think the view expressed in your comment is overly reductionist. Of course everyone works on teams, and of course there are more factors at play than simply being better at some task in a vacuum. But that doesn't change facts like:

* Some people are smarter than others

* Some people are more dedicated than others

* Some people have more experience than others

* Some people have life circumstances that allow them to contribute more than others

* Some people have values more in line with the company's success than others

This is kind of what I mean by "the world as it is" vs "the world as you want it to be." The points above are common sense to people that don't have the world view expressed in your comment. The idea that all people are equally valuable is only true in the abstract sense. When you actually need to accomplish a concrete goal (like putting a man on the moon), some people are absolutely more valuable than others toward doing that, and an organization that pretends that's not the case is shooting itself in the foot.

Having a meritocratic structure doesn't even inherently mean individualist. It could be measured at the team or project level. It also doesn't have to be (and typically isn't) exactly the measurable contributions ("load-bearing capacity") to the goal itself that determines the best "engineer." More often your "best" engineer is the one who is easy to work with, embraces tough challenges, and can adapt well to changes.

But honestly: I think everyone already knows these things. That's why I see this more as virtue signaling: it's one thing to change a governance statement. It's another when it's actually put into practice.

Yeah I totally agree with that. To me, that shows that the measure of merit is, #1: extremely subjective and #2: influenced by all kinds of factors that would not traditionally be associated with merit.

Basically, I think that it often makes sense to do affirmative action within a true meritocratic framework, but that the term meritocracy is generally used to juxtapose against affirmative action. If the term is being used to argue against its true meaning, then it's not a useful term and should be retired.

For someone who doesn't care about private parts, why are you the first person to bring it up? If you care about delivering success, you should take a moment and think hard about why you feel so strongly about the race/gender aspect, and are not thinking about the probability theory.

> For someone who doesn't care about private parts, why are you the first person to bring it up?

Sorry--sometimes the endless stream of euphemisms and language policing bores me. This isn't an isolated event and I think we all know what's being said here.

> If you care about delivering success, you should take a moment and think hard about why you feel so strongly about the race/gender aspect, and are not thinking about the probability theory.

I really don't care about the race/gender aspect, but it's being injected into Mozilla's governance statement and here we are discussing it.

This assumes meritocracy we want is based on skills. I think "working as intended" is actually based on "shown skills", not "skills"; people with mad skills which didn't show it in the project don't have merit. I've seen "do-ocracy" used instead for this reason.

Do-ocracy does not assume equal opportunity.

This is a really weird critique of the idea that people with fewer opportunities should be given a few more chances to show their skills. It's directly increasing the metric you're interested in, but you're still against it for... reasons?

In a true meritocracy, all candidates are judged on equal footing so P(x|a) isn't compared to P(y|b). In practice this is very possible (albeit expensive) to achieve, given a set pool of candidates. Although judging based on past achievement has the problem you describe, it can be used in place of a fair test (eg, judging candidates based on resumes instead of technical interviews) since high past achievement is strongly (but not perfectly) associated with high skill.

Mozilla doesn't appear to be addressing your version of meritocracy, but the actual definition of meritocracy, which still gives those who had the opportunity to build skill an advantage.

If the sample sizes differ you should use a statistical test not just the variances

Reply on news.ycombinator.com