Gender in AI Language Models: What's In a Name?Naming Shifts Over Time
It's well-known that names often exhibit generational shifts in popularity. If you're an American named Linda, there's over a 50% chance you were born between 1946 and 1955.
Similarly, some names exhibit shifts in their gender associations. 85% of all Charlies are male, but 53% of the Charlies born in 2021 were girls.
Here I've graphed the names with the biggest shifts over time.1 Some observations:
- Names tend to shift from boys to unisex or boys to girls, not the other way round. I attribute this to fragile masculinity: parents seem to think that naming their son Sophia will cause more problems than naming their daughter Stevie. Note how Jaime and Ashton start as a boys name, become unisex, and then snap back. I attribute Ashton's drop to Ashton Kutcher, born in 1978. Not sure about Jaime.
- These shifts tended to occur around 1980: perhaps this says something about liberalizing gender norms in the 60s, perhaps it's some other pattern I've missed.
- Regularized so that years with very low counts don't bias the results, filtered to years with over 50 babies, and smoothed.↩
Naming Shifts Over Time
It's well-known that names often exhibit generational shifts in popularity. If you're an American named Linda, there's over a 50% chance you were born between 1946 and 1955.
Similarly, some names exhibit shifts in their gender associations. 85% of all Charlies are male, but 53% of the Charlies born in 2021 were girls.
Here I've graphed the names with the biggest shifts over time.1 Some observations:
- Names tend to shift from boys to unisex or boys to girls, not the other way round. I attribute this to fragile masculinity: parents seem to think that naming their son Sophia will cause more problems than naming their daughter Stevie. Note how Jaime and Ashton start as a boys name, become unisex, and then snap back. I attribute Ashton's drop to Ashton Kutcher, born in 1978. Not sure about Jaime.
- These shifts tended to occur around 1980: perhaps this says something about liberalizing gender norms in the 60s, perhaps it's some other pattern I've missed.
- Regularized so that years with very low counts don't bias the results, filtered to years with over 50 babies, and smoothed.↩