[DSRP Evidence](https://dsrpevidence.org/)

# HawkesRank: Event-driven centrality for real-time importance ranking

## Details

**Authors** Sornette et al.

**Year** 2026

**Publisher** PNAS Nexus

**Discipline** Network Science

**Secondary disciplines** Statistics

[Read it at the publisher](https://doi.org/10.1093/pnasnexus/pgag292) 
10.1093/pnasnexus/pgag292

## In authors' words

### What they found (results)

Modeling influence with multivariate Hawkes point processes, the authors show classical centrality measures (Katz centrality, PageRank) are stationary mean-field limits of a more general event-driven formulation, which better captures real, time-varying importance in empirical online-communication data by distinguishing self- from cross-excitation.

## Commentary

### In short

Reframing a network's importance structure as a relational, whole-system property — separating self-generated from externally triggered activity — recovers classic static centrality measures as a simplified special case, showing the static view was a snapshot of a richer relational dynamic.

**Patterns it shows** D, S, R

**Added** 2026-09-07

**How to cite this** Sornette et al. (2026). HawkesRank: Event-driven centrality for real-time importance ranking. PNAS Nexus.