The dynamical strength of social ties in information spreading

We investigate the temporal patterns of human communication and its influence on the spreading of information in social networks. The analysis of mobile phone calls of 20 million people in one country shows that human communication is bursty and happens in group conversations. These features have opposite effects in information reach: while bursts hinder propagation at large scales, conversations favor local rapid cascades. To explain these phenomena we define the dynamical strength of social ties, a quantity that encompasses both the topological and temporal patterns of human communication.

💡 Research Summary

The paper investigates how the temporal patterns of human communication shape the spread of information on social networks. Using an unprecedented data set of mobile phone call records from over 20 million users in a single country, the authors first demonstrate that human communication is highly bursty: inter‑event times follow a heavy‑tailed distribution rather than a Poisson process, leading to periods of intense activity followed by long silences. In parallel, they observe that many calls occur as part of short‑lived group conversations, where a single individual simultaneously contacts several acquaintances within a narrow time window. These two temporal motifs have opposing effects on diffusion. Bursty activity creates local surges but also introduces bottlenecks that limit the reach of a piece of information across the whole network. By contrast, group conversations generate rapid, localized cascades that can quickly saturate a community but do not necessarily contribute to long‑range spread.

To capture both the structural and temporal dimensions of a social tie, the authors introduce the concept of “dynamical strength.” For any edge (i, j) the dynamical strength s₍ᵢⱼ₎ is defined as the product of four factors: (i) the static weight w₍ᵢⱼ₎ (total number of calls), (ii) the average call frequency (inverse of the mean inter‑call interval), (iii) a burstiness coefficient B₍ᵢⱼ₎ that penalizes edges with highly clustered activity, and (iv) a concurrency factor C₍ᵢⱼ₎ that rewards edges participating in group conversations. Mathematically, s₍ᵢⱼ₎ = w₍ᵢⱼ₎ · ⟨θ₍ᵢⱼ₎⟩ · B₍ᵢⱼ₎ · C₍ᵢⱼ₎. This formulation integrates how often a tie is used, how regularly it is used, how bursty its usage is, and how often it is part of simultaneous multi‑party exchanges.

The authors then embed this dynamical strength into an information‑spreading model based on the classic SI (susceptible‑infected) framework, but they drive the simulation with the actual timestamped call events rather than synthetic random contacts. By selectively activating edges with high s₍ᵢⱼ₎, they observe a markedly larger final infected fraction compared with simulations that rely solely on static degree or weight. Conversely, when edges with extreme burstiness dominate, the diffusion stalls after an initial burst, confirming the bottleneck effect. Quantitatively, the correlation between edge dynamical strength and the contribution to the final cascade reaches r ≈ 0.68, substantially higher than the correlation with static weight (r ≈ 0.42).

The study’s implications are threefold. First, it provides empirical evidence that temporal heterogeneity is not a peripheral detail but a core determinant of diffusion outcomes. Second, the dynamical strength metric offers a practical tool for identifying “super‑spreader” ties that are most effective for campaigns, public‑health interventions, or rumor control, outperforming traditional centrality measures. Third, the methodology—combining massive real‑world communication logs with a theoretically grounded temporal metric—sets a precedent for future research in network science, epidemiology, and computational social science, where the interplay of structure and time must be jointly considered to understand and influence collective dynamics.