Cultural evolution and personalization

In social sciences, there is currently no consensus on the mechanism for cultural evolution. The evolution of first names of newborn babies offers a remarkable example for the researches in the field. Here we perform statistical analyses on over 100 years of data in the United States. We focus in particular on how the frequency-rank distribution and inequality of baby names change over time. We propose a stochastic model where name choice is determined by personalized preference and social influence. Remarkably, variations on the strength of personalized preference can account satisfactorily for the observed empirical features. Therefore, we claim that personalization drives cultural evolution, at least in the example of baby names.

💡 Research Summary

The paper tackles a central unresolved question in the social sciences: what drives cultural evolution? Using the evolution of first‑name popularity in the United States as a natural laboratory, the authors compile more than a century of name‑frequency data from the Social Security Administration. After cleaning the dataset (restricting analysis to the top 1,000 names per year, normalising for spelling variants, and controlling for demographic shifts), they examine two empirical signatures of cultural change.

First, they plot the rank‑frequency distribution on log‑log axes and observe that, while the distribution roughly follows Zipf’s law for much of the 20th century, it begins to flatten markedly after the 1960s. In practical terms, the dominance of a handful of names declines and the middle and lower ranks gain relative weight. Second, they compute inequality measures—Lorenz curves and the Gini coefficient—for each year. The Gini index, which quantifies how concentrated name usage is, shows a steady decline from the 1970s onward, indicating that name choice becomes more evenly spread across a larger pool of options.

To explain these patterns, the authors propose a stochastic choice model that blends two forces: personalized preference and social influence. The probability that a newborn receives name i in year t is given by

 P_i(t) = α·π_i + (1‑α)·σ_i(t).

Here, π_i is a fixed, normalised “intrinsic appeal” of name i (capturing meaning, phonetics, cultural connotations, etc.), σ_i(t) is the market share of name i in the previous year (the social‑imitation term), and α ∈