Role of Activity in Human Dynamics

The human society is a very complex system; still, there are several non-trivial, general features. One type of them is the presence of power-law distributed quantities in temporal statistics. In this Letter, we focus on the origin of power-laws in rating of movies. We present a systematic empirical exploration of the time between two consecutive ratings of movies (the interevent time). At an aggregate level, we find a monotonous relation between the activity of individuals and the power-law exponent of the interevent-time distribution. At an individual level, we observe a heavy-tailed distribution for each user, as well as a negative correlation between the activity and the width of the distribution. We support these findings by a similar data set from mobile phone text-message communication. Our results demonstrate a significant role of the activity of individuals on the society-level patterns of human behavior. We believe this is a common character in the interest-driven human dynamics, corresponding to (but different from) the universality classes of task-driven dynamics.

💡 Research Summary

The paper investigates how individual activity levels shape the heavy‑tailed inter‑event time distributions that are commonly observed in human dynamics. Using two large‑scale, time‑stamped data sets—MovieLens movie‑rating records (≈500 000 ratings from 1 018 users over 1995‑2006) and a month‑long mobile‑phone SMS log (≈3 × 10⁶ messages from 20 345 users)—the authors compute the time interval τ between successive actions for each user. Activity (A) is quantified as the average number of actions per day (ratings per day for MovieLens, messages per day for SMS). Users are stratified into low, medium, and high activity groups based on the distribution of A, and the probability density P(τ) is examined for each group.

At the aggregate level, all three groups exhibit power‑law tails P(τ) ∝ τ⁻ᵅ, but the exponent α increases monotonically with activity. In MovieLens, low‑activity users have α≈1.3 while high‑activity users reach α≈2.1; a similar trend appears in the SMS data. This indicates that more active individuals generate shorter, less variable waiting times, leading to steeper decay in the tail of the distribution.

At the individual level, each user’s τ distribution is also heavy‑tailed, yet a clear negative correlation emerges between activity and the width (e.g., standard deviation or inter‑quartile range) of the distribution. Highly active users not only have larger α values but also a narrower spread of τ, suggesting a more regular rhythm of behavior. Conversely, low‑activity users display broader, flatter tails, reflecting sporadic engagement.

The authors compare these findings to the well‑known task‑driven models of human dynamics (e.g., Barabási’s priority‑queue framework), which attribute power‑law waiting times to random selection among tasks of varying priority. The present results point to a different mechanism: the intrinsic rate at which an individual initiates actions (their “activity”) directly modulates the statistical shape of inter‑event times. This “activity‑driven” class appears to be a universal feature of interest‑driven activities, as it holds for both low‑frequency rating behavior and high‑frequency text messaging.

The paper concludes that activity is a crucial, previously under‑appreciated variable that bridges individual‑level temporal patterns and society‑level scaling laws. By demonstrating consistent patterns across distinct domains, the study suggests that many complex social phenomena—information diffusion, epidemic spreading, collective attention—may be better understood by incorporating activity heterogeneity into models. Future work is proposed to explore additional interest‑driven platforms (streaming services, online gaming), to refine activity metrics (temporal bursts, circadian cycles), and to quantify how activity‑driven temporal structures affect macroscopic processes such as cascade dynamics and network resilience.