Dunbar's circles in email: Enron works, but high-volume Jmail breaks methods

A new paper on arXiv (2606.02376) by Di Cursi, Boldrini, Conti, and Passarella explores whether Dunbar’s number — the theory that humans maintain social circles of roughly 5, 15, 50, and 150 — emerges in email communication data. They contrast two very different archives: Enron, a corporate corpus with about 150 users, and Jmail, a single-ego archive centered on an exceptionally active focal actor whose communication volume is more than 20 times the average Enron user. The goal: determine if classical layered ego networks are recoverable from digital interaction patterns.

For the Enron dataset, standard clustering and threshold-based methods partially reproduced Dunbar-like organization, with stable inner circles and an outermost layer corresponding to Dunbar’s affinity group (~50 individuals). However, the Jmail archive broke these methods entirely due to its extreme communication intensity. Instead of a broad network with many occasional contacts, Jmail’s focal actor interacted with a selective pool of high-interest alters on a much higher frequency scale. The authors found that by first anchoring the Dunbar frequency ladder to the empirical support-clique boundary — essentially normalizing for the actor’s overall volume — a clearer layered structure emerged. Reciprocity analysis confirmed the layers reflected genuine bidirectional relationships. The study provides a methodological roadmap for analyzing social network structure in high-intensity communication contexts, with implications for understanding digital relationships and designing tools that adapt to varying user behavior scales.