andrewducker | Some thoughts on quotas

In f-locked discussion elsewhere I was talking about quotas, people's aversion to them, and what can be done about them.

I feel iffy about quotas. Not because they'd be acting against me[1], but because they always feel like a terribly blunt tool to attack a problem with, and they seem to inflame people a lot whenever they come up[2], and cause lots of short-term anger. Some people will feel that blunt instruments are necessary, some will feel very strongly that reverse-discrimination is still discrimination (and thus wrong), and the discussions are frequently made of more heat than light, what with the demonising of people with opposing opinions.

Personally, what I'd like is for people to _think_ more and consider the situation[3], try to make it better, and be clear (when they haven't made it better) what they tried to do. Which is why I wouldn't be in favour of saying simply "50% of all X must be Y.", but I would be in favour of requiring, when certain levels weren't met, that people had to explain _why_.

So, taking the recent example of conventions[4], you could set a minimum level at 40% of either gender (thus allowing for a 3:2 split on a five member panel), and while any given panel could have any actual split, any one which breached the 40% would have to log an explanation saying something like "All five of the world-renowned experts in this field are female." or "We asked our friends, we put up posters, we sent out mass emails, but only one woman was willing to be on this panel."

This would then mean that people at least _looked_ at the figures, had a think about it, and then probably put in some effort because they wanted to avoid the paperwork. Because, frankly, avoiding paperwork is a prime motivator to most people, and enough to spur them into doing the right thing.

[1]They would, because I'm middle-class, earn above average, am not part of a minority that's terribly visible or picked on where I live, etc. But I'd be happy if they did so, because frankly I got a lot of help along the way.
[2]Many people do not like the idea of reverse discrimination. Some people do not like the idea that (for instance) a Chinese person would be placed in the situation where people could question where they actually got the job on merit, or whether it was just to fill a quota.
[3]This bit applies to negative situations in general, not just ones that quotas are a plausible solution for.
[4]Paul Cornell announced that he would not appear on any panel that wasn't 50% female. Other people are now questioning why any panel shouldn't be representative of the general population.

Flat | Top-Level Comments Only

"um, no" to the possibility of requiring 5 nominees to be 3:2.
Real life statistics will not do that. If there is no gender bias, there will be slates of 3:2, 4:1, and occasionally but noticeably, 5:0.

If pulling names randomly out of a hat, yes. If actively choosing people, no.

Well actually I think apostle is right. Consider this perfect world experiment:
Your hiring pool is a large sample of an equal number of men and women. Talent has the same statistical distribution for all people (so women are not better than men and vice versa).
You unerring pick the most talented 5 people.
You will get the following ratios
1/16th time 5:0 (one way or the other).
5/16th time 4:1 (one way or the other).
10/16th = 5/8th time 3:2 (one way or the other).

Almost half the time you will get something which looks (to someone who doesn't train in statistics) very biased. This is because people are very bad at assessing these things and so "4 men only one woman" or "4 women only one man" looks dodgy.
If you didn't watch out for this you'd end up introducing a quota system on something perfectly unbiased.
Of course we don't actually live on this world at all, indeed nothing like it... so the issue is not one we should worry about too much.
In general though it would be a bad idea to apply such a quota system when the number of people you're looking it is "small".

In larger systems the law of large numbers saves you.
http://en.wikipedia.org/wiki/Law_of_large_numbers

If the distribution is 4 men and 1 woman on the panel then it's not necessarily a problem.
If the distriubtion is 400 men and 100 women on 100 panels then it's almost certainly problem (with a p-value I can't be bothered to calculate).

"You unerring pick the most talented 5 people."
I thought that point was that you didn't do that, if you had quotas.

If you are merely reporting cases when you went outside of the guidelines, then yes, absolutely.

I thought that point was that you didn't do that, if you had quotas.

Yes... but that's a different point. If your quota system is correctly specified then an unbiased group of people choosing within an equal talent pool should never (or almost never) be affected by it (because those people don't NEED a quota system).

The fact that (in the example) unbiased people are being swayed in their choice by a quota system shows the quota system picked is a wrong one. It's easy to come up with quota systems which are "equal" but also "stupid". For a really obvious example, we number jobs in order and for even numbered jobs hire a man and odd numbered jobs hire a woman. Hooray, equality! But it's too strong a requirement. You could achieve the equality requirement with a much much looser quota system.

If you're choosing a quota system (or even a "guideline" system) it needs to work on larger numbers than the numbers you are talking about or it simply won't get sensible results.

If your quota system is correctly specified then an unbiased group of people choosing within an equal talent pool should never (or almost never) be affected by it (because those people don't NEED a quota system).

That works if you have reporting over a large group of people, but makes it hard to tell whether one particular case is very biased.

If we use the system I recommend, whereby you merely report back why you deviated from the quota. In which case if you pick all men for a role you need to justify that "They really all were better, the women who applied were all of lower quality." - and if one person consistently comes out on one side of the curve then you can investigate in more detail, and you have the paperwork to start with.

That works if you have reporting over a large group of people, but makes it hard to tell whether one particular case is very biased.

It is always hard to tell whether one particular case is biased... that's the very unfortunate problem with these things.

If we use the system I recommend, whereby you merely report back why you deviated from the quota.

I assumed that the point of your system is that it would modify people's behaviour. If your intent is that people do what they would anyway but fill in some paperwork then there's no additional problems introduced.

If, on the other hand, in your system you have some kind of expectation people will move towards being "less biased" by trying to ensure that ratio are 3:2 in groups of 5 then it has the risks I described by operating on small groups.

I think that making people justify themselves will make them think "Do I really want to do X, was it the right thing to do?", and justify it to themselves before they justify it to other people. Faced with having to justify hiring a group all of one gender the better people will nail their justifications down, the worse ones will switch out someone for a member of the other gender. The latter strikes me as a negative effect, but I would have thought the overall effect will be to make people consider what they are doing, which should be an overall benefit.

You may be right that overall the good in your system outweighs the bad... I'd still feel a little cautious about it though.

I think that it would almost certainly dissuade people from just going with the "My four mates would be awesome on this panel" approach which tends to lead to the current situation. It would almost certainly have some negative effects (any system does).

Recall that auditioning musicians "blind" (without seeing them) resulted in many more women getting jobs in orchestras - the hiring people had actually deluded themselves into thinking that the men "just played better" and they were "simply hiring the best players", but when they couldn't see if the person was a man or a woman it turned out that women play well too.

Justifications aren't always true, even if the person saying them really believes they are.

That's the math I was thinking of.
If you occasionally have a 5:0 ratio or a 4:1 ratio, that by itself does not indicate bias. In fact, having no 5:0s or 4:1s indicates bias.

If your aim is to have the 5 best individuals this is true. If you aim is to have the "best panel possible" then you might find that panel-goodness increases with diversity (a diversity of perspectives might lead to a better panel). (Please not "might"s I actually have no experience of picking the best panel for any such event).

Also if your goal is to "decrease the perception that SFF fandom is entirely made up of white men" then you will almost certainly find that putting more women and non-white people on your panels will assist in that goal even if you might have had a marginally better discussion with an all-white-male panel (because those white men were more expert/interesting in the subject). Aiming at these perceptions of inequality can help decrease levels of actual inequality later.

Some thoughts on quotas

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

yes

no subject