Simulating Patriarchy

Trust me, I know what I'm doing.
-- Sledge Hammer

Hey, awesome Comp Sci students!

I have a project idea for anyone interested in gender issues and exploring the validity and boundaries of popular feminist theory. I think it would be super cool if someone were to write a patriarchy simulator, get school credit and publish their results.

I haven't got time to do it myself (I'm busy saving e-mail), but I'd be happy to provide guidance and help out a bit if someone wants to work on this.

Allow me to explain...

But first,

...a word from our sponsor!

I am very interested in hiring someone to work on PageKite related work for the summer. I can pay a relatively decent salary, but there's no office (I work from home or at cafes) and you have to bring your own computer. The programming languages used by PageKite are Python, C, HTML/JS/CSS and maybe some Windowsy things depending on which projects we decide to tackle.

If you are interested in getting paid to write Free Software, please get in touch and we'll discuss!

OK, back to our regularly scheduled programming:


One of the core tenets of modern feminism, is that prejudice exists and that it is bad for society.

Stereotypes are considered a form of institutionalized prejudice, where all of society agrees on certain basic truths, such as the fact that girls can't throw balls, blondes are dumb and all men like cars and red meat.

I mean, we know this. We see it all around us, it must be true. And knowing this, when I hire someone to throw a ball I'm definitely not going to hire a girl. Or if I absolutely must hire a girl, I'll at least pay her a bit less because she's going to do a bad job.

Feminists don't like it when I do that, but I'm a smart, rational person. I know what I'm doing and I wish they'd just leave me alone...

It seems obvious that it's going to really suck to be a woman who likes throwing balls in that world. So it's totally understandable that feminism argues against this sort of thing.

However, there is an interesting question implicit in this scenario.

Do stereotypes help or hurt society? How much?

Is feminism correct in claiming that this is universally bad, or are stereotypes actually encoding useful information about how the world works? Do stereotypes perhaps save us all time and help make sure people work on things they're actually good at? Do stereotypes improve or harm market efficiencies?

I personally think the feminists are probably right and I have a theory about why that is. I think my theory could be modeled and tested with a computer simulation.

Stereotypes as feedback loops

My theory can be summarized as follows:

Stereotypes are the manifestation of positive feedback loops in how society evaluates individuals. Over time, these feedback loops may come to dominate decision making in certain fields, causing the actual merits of the individual to be overlooked and leading to inefficient allocation of resources.

To illustrate in terms of jobs, girls and balls...

Since men are on average physically stronger than women, if you pick a random guy and a random girl, the odds are the guy will be able to throw the ball a bit further. To keep it simple, let's assume that is his only advantage and by all other metrics men and women are equally talented ball throwers.

Now let's imagine we have a few different job positions that all involve throwing balls. Some of them require throwing the ball as far as possible, for others, distance does not matter.

Finally, consider two hiring strategies:

  1. All applicants are tested and hired based on how well they do.

  2. Applicants submit resumes explaining their job throwing qualifications and relevant experience and are hired based on a evaluation by a human.

For the first ball throwing companies, strategy one is the only one available. It will prefer men to the degree that they are actually better at throwing long distances, but we should still end up with a mixed group. Let's say 60% men, 40% women?

Now let's imagine we do that for a few years and build up a pool of talented, experienced professional ball throwers. Then suddenly a second generation of ball throwing companies enters the market. They are more efficient and one of their innovations is hiring method number two, which is a lot less work then testing all the candidates every time.

Strategy two will prefer people with past experience, so instead of choosing from the entire population it is choosing from a smaller population which is 60% men. This means a random selection will preserve the 60/40 split, but if the hiring process successfully evaluates the skills of the candidates, you find that men's advantage has compounded and 84% of the hired ball throwers will be men.

On the third iteration, it reaches 93.6%.

It's a positive feedback loop!

If you keep iterating through this loop again and again, over time the bias towards hiring men will increase to the point that the ball throwing profession becomes entirely male dominated. This will also happen, even if their original advantage was much much smaller than the 60/40 split I proposed above.

This is of course an oversimplification.

For one thing, the numbers only add up if each successive generation of companies is smaller than the first, otherwise either everyone gets a job (but maybe not equal pay). Also, sometimes you have no choice but to hire inexperienced people. If that hiring is done using objective evaluation of their skills, then the effects of the feedback loop will be diminished and the system may reach an equilibrium. But if instead companies hire candidates that "look like good performers based on past experience", the feedback loop is likely to preserve itself.

This effect doesn't just apply to hiring, it applies to all sorts of collective decision making populations make about their individuals.

So without sufficient counterexamples and reality checks, over time society will convince itself that girls can't throw balls.

Simulating Patriarchy

I think it would be very interesting to write a simulator which explores the effect of feedback loops like this, quantifies them and visualises.

Different hiring algorithms could be written which look at different attributes of the individuals, and the algorithms could be compared.

It should be possible to both objectively measure how inefficient the hiring processes are and identify "tipping points", such as how many "reality checks" are needed to prevent history-based hiring strategies from becoming runaway feedback loops.

Results could be presented visually, to make it easier for lay-people to understand the processes. For example, one could animate random generations of multicolored squares and circles competing for positions on a color palette...

I think it could be a very interesting project and if done with sufficient rigour, could even produce meaningful mathematical foundations for some feminist theory.

Wouldn't that be neat?

Tags: tech

Recent posts