[Ntrails]

It is for reporting and analysis.

"Why are 2/3 of our maths grads white men?"

"Hmmm, don't know. What were our applicant demographics?. Ooh, they are only 60% white men. Lets dig..." etc etc.

If you don't have the data you cannot identify where in the pipeline an anomaly starts

[Zababa]

> "Why are 2/3 of our maths grads white men?"

Why would you ask that in the first place?

[dekhn]

That ratio indicates that there is some biased process leading to the ratio, and that process should be de-biased. Presuming women and men have equal math abilities (I assume this is true) and equal desire to obtain degrees (not entirely sure; this is a very controversial topic) you would expect that an unbiased selection process would generate 50% male/female ratio. Again, many assumptions and conclusions in what i just said are not absolutely certain.

Many departments are currently realizign they play a role in systemic discrimination, and those ratios help them understand how far they are from the mean.

[Zababa]

> That ratio indicates that there is some biased process leading to the ratio, and that process should be de-biased.

Or that people have different preferences?

> Presuming women and men have equal math abilities (I assume this is true) and equal desire to obtain degrees (not entirely sure; this is a very controversial topic) you would expect that an unbiased selection process would generate 50% male/female ratio.

If by that you mean that "for each degree, the percentage of the people that want to get it follows the repartition in the population", it is absolutely not true. I don't know why or what mechanism are at play here, but assuming that this is all bias seems very weird.

> Again, many assumptions and conclusions in what i just said are not absolutely certain.

I think this is a really weak reason to collect racial statistics in the first place.

> Many departments are currently realizign they play a role in systemic discrimination, and those ratios help them understand how far they are from the mean.

That's true, but you can also base yourself on things like revenu which seems to be more universal than race.

[Jensson]

> Many departments are currently realizign they play a role in systemic discrimination, and those ratios help them understand how far they are from the mean.

I hope you realize this process started over 50 years ago. Most fields turned and are now dominated by women. The few fields that are still not dominated by women are probably that way for another reason than discrimination, women are far too good at taking over fields for some discrimination to stop them from doing so.

[dekhn]

Nothign I said in my comment above is my current position, it's a description of the thinking of people who run departments. Note that I very carefully said the idea of whether genders have preferences, is a controversial topic.

In my own field, biology, many subfields do have equal gender ratios (went from skewed to equal over last 20 years). However, having worked in Silicon Valley at tech firms, I can say that they (and the programs that feed them) still have very skewed ratios.

I'm aware of these processes and how long they've lasted; about 100 years ago, Harvard limited the number of Jewish people who could be accepted. Jews were excluded from working at "white-shoe law firms" (the ones that did high-value business deals) and instead generally ran law firms that handled "the dirty stuff" (divorces etc). Eventually, Jews became much more accepted, explicit Jewish quotas at universities were removed, and white shoe law firms hire (and have partners) who are Jewish.

[guerrilla]

To figure out if there is discrimination or not...

[Zababa]

I don't think it's good science to say that every difference comes from discimination. For example, the poorer you are, the higher are your chances of dropping out. This is not discrimination against poor people. But, at least to me, this is a problem that we should try to fix.

[guerrilla]

Figuring out whether or not there is discrimination is not saying that any differences comes from discrimination. Identifying differences is the first step in doing that, not the last step.

[Zababa]

That's a fair point.

[ricardobeat]

That’s the point of the entire article. It’s a fuzzy problem that can’t be solved by strict categorization like that, which is in itself a form of discrimination. You design your admissions process to be as fair as possible, not “tune it” to get the metrics right.