Temporal fractal nature of human behavior in personal computer operations

Abstract

It has been shown that human behavior, observed as a time-series of physical activity intensity, exhibits a temporal fractal nature (TFN) characterized by the pattern of alternating active and rest periods. Furthermore, this nature is altered by the health conditions of an individual and can potentially be used as biomarkers. In this paper, we focus on human behavior observed as a time-series of personal computer (PC) operations, such as keystrokes and mouse clicks. Using data from 35 healthy office workers over three months, we demonstrated that human behavior in PC operations also exhibits a TFN equivalent to that of physical activity. By defining daily workload as the ratio of total active duration to the total active and rest duration in a day, we showed how humans adapt to workload fluctuations: (1) active durations follow a stretched exponential type distribution with a workload-dependent scale parameter; (2) rest durations follow a scale-free type distribution with a workload-dependent shape parameter (scaling exponent); (3) long active and short rest periods (and vice versa) tend to cluster together in a fractal manner with long-range temporal correlations invariant to workload fluctuations. These findings deepen our understanding of the TFN of human behavior and highlight the importance of considering workload when developing robust biomarkers based on the TFN.