Does Google Not Understand Women Or Just Geek Women?

[Neoma]

Apparently, Google thinks we’re men. Their ads preferences page was recently brought to my attention. It breaks down your internet searches by categories and then infers your age and gender based on what you’ve looked at. And according to them, all of us here at The Mary Sue are 25-34 year old men. At first it seemed like perhaps “male” was just the default for those who didn’t have their google internet history turned on and as ridiculous as that would be, all Susana (results above), Jamie and myself all had our history set to on, which means these assumptions where made on our internet usage. It makes me consider what DC Comics Demon Knights writer Paul Cornell tweeted earlier, “Do you think that Google may not have considered the tastes of geek girls?”

More Here

[Xithras]

1. That's the problem with algorithms.

No "person" at Google made that decision, but it's the result of algorithms that look at browsing patterns and determine the most plausible demographics. A Kenyan might speak fluent Hindi on a daily basis, but the system is probably going to guess that they're from India if they're regularly doing searches in Hindi, simply because that's where most Hindi speakers are from. There will always be outliers that don't fit the statistical mold, but they are usually ignored so long as the numbers are small enough.

The numbers here are simple. Less than 23% of tech jobs are held by women. Less than 18% of computer related degrees are held by women. Less than 15% of Wikipedia articles are written by women. Less than 3% of Open Source developers are women.

Geek women are outliers. If you want to change that, convince more women to go into technical fields. Until then, Google is simply reflecting the reality of that particular demographic.

[PoliticAverse]

2. Guesses based on statistics tend to be right and wrong in relation to the statistics.

If you found for example that 80% of the people that visit a certain website are actually male, you can infer that the
next person the visit the website is probably male. Such inference will be wrong perhaps 20% of the time.

People that don't fit the norm tend to be misjudged by statistical inferences based on norms.