Core Concepts
Parental preferences for name "uniqueness" can lead to unstable and suboptimal outcomes in the highly competitive arena of baby naming.
Abstract
The paper presents a game-theoretic model for baby naming, where parents aim to choose a name with a desired level of "uniqueness" or popularity. The authors make several "Extremely Reasonable Assumptions" (ERAs), such as assuming parents are myopic and that there is only one gender.
The model considers the dynamics of name frequency and parental preferences, represented by power law distributions. The authors analyze the stability and satisfiability of the resulting name distributions under different parental preference scenarios.
When parents prefer less common names (t' > 0), the name distribution becomes unstable, with popular names quickly becoming unpopular and vice versa. Conversely, when parents prefer more common names (t' < 0), the name distribution becomes stable, with popular names remaining popular over time.
The authors also present simulations using a log-normal distribution of parental preferences, which produce similar insights. Additionally, they include an experiment with the Kat-GPT language model, which suggests gender differences in name popularity.
The paper concludes by discussing potential extensions, such as the creation of new names and considering non-myopic parents, and the implications of their findings.
Stats
The popularity of names follows a power law distribution.
The preference of parents for name "uniqueness" can also be modeled as a power law distribution.
When t' > 0 (parents prefer less common names), the resulting name distribution becomes unstable, with popular names quickly becoming unpopular.
When t' < 0 (parents prefer more common names), the resulting name distribution becomes stable, with popular names remaining popular over time.
Quotes
"Naming a child is akin to choosing an outfit for the Oscars. It must be unique enough to stand out - no one wants to show up to the Oscars in the same dress - but it must also be similar enough to be recognizable as a name."
"If an arrangement is stable, this means that given an existing distribution fi(a) and a parental preference distribution g(µ), every name's frequency will be exactly the same at the next time step i + 1: fi+1(a) = fi(a) ∀a ∈ A."