Entropy and the Sociology of Uniqueness
Résumé
Uniqueness and individuality are important questions in sociology, and we discuss an emerging method which is to rely on entropy, a metric from information theory, in order to capture uniqueness or commonness quantitatively. Uniqueness may be observed in brand choices or in naming choices, for example. Research has shown that names and market shares generally follow power laws, and this allows us to derive closed form expressions for entropy-based population-level uniqueness under this assumption. Relying on entropy, however, raises the issue of bias in small sample estimates, a common problem with historical or sample-based datasets. We compare a few methods for bias correction, among which bootstrapping appears as the most appropriate for relative analysis, since it controls for sample size variation.
