Gender stereotypes and Dolores Labs

Are you attractive?

Why of course you are. I have often admired your fine good looks. And as a reader of this blog, you possess a keen intelligence and a quick wit. Good for you!

A slightly more interesting question is this. Where are you now, relative to your lifetime peak attractiveness? Or your lifetime peak intelligence? Dolores Labs can help.

Dolores Labs is a wonderful outfit that takes crowdsourcing very seriously. They have a core competence in driving Amazon’s Mechanical Turk to do some rather interesting research. Some of it is what you might expect: make the Turkers do dreary legwork, like determining the race of everybody who’s appeared on the cover of Sports Illustrated since 1954.

But the more interesting stuff they do is more like a giant psychology experiment on the Turkers themselves. Andy Baio did something like this recently, asking Turkers, Why do you turk?. It’s a humanizing answer to the question “who ARE these Mechanical Turk people anyway?”

Dolores Labs did the following. They said here are 100,000 photos of people you’ve never seen. Just tell us, are they attractive? Are they trustworthy? Are they intelligent? Averaged across millions of judgments, when do women experience peak attractiveness? The answer appears to be 24. Aha! I knew it! And anybody who buys car insurance for their son will not be surprised that a man’s trustworthiness trough is 18 or 19. And may we posit an inverse relationship between male intelligence and sexual distraction? What should we make of the W-shaped curve of perceived female intelligence? Is there a correlation between intelligence and trustworthiness? Read the article. It’s all there.

PS: Don’t miss the crowdsourced naughty bits.

2 thoughts on “Gender stereotypes and Dolores Labs”

  1. Unfortunately, there are confounding variables that prevent the data from accurately answering the question: 1) what was the age distribution of the Turks who worked on this problem; it would need to be pretty flat to be accurate; for instance, if everyone considers their peers to be more attractive, then the curve would simply match the distribution of Turks; 2) the experiment doesn’t answer “what do people look like over time” but “what do people of different ages look like right now;” this is important, because exercise and diet trends were different for current 80-year olds than they will be for current 20-year olds when they turn 80 — perhaps the W represents the current opinion that baby-boomers are stupid and untrustworthy, while Gam-gam and Pee-paw are wise and loveable; 3) as they point out, the data are heavily weighted toward 20-30 year olds; does going through 10 pictures of 25 year olds prime the viewers the find 18 and 40 year olds less attractive (different but not different enough) — maybe this explains the bump. Humans are the most confounding variables you can introduce into an experiment, especially when examining other humans.

Comments are closed.

%d bloggers like this: