With the development of programmable computers, humans have entered the digital age. The emergence of the World Wide Web and the ubiquity of computers, mobile phones, and other devices that automatically store digital records has led to the concept of big data. To harness this big data, new computational tools and methods are constantly being created to extract information from it. When people interact with digital devices and platforms, they leave digital footprints. These traces can open a window into understanding the behavioral patterns of humans. The emerging multi-disciplinary field of computational social science takes advantage of the large, empirical datasets built of these footprints and uses them to address questions from various fields of social sciences by applying methods and techniques from hard sciences like physics and network science. In the past decade, there has been a surge of studies where such datasets have been used to study human patterns of behavior. Many have looked at structural properties of social networks such as personal network sizes or tie strengths. A more recent trend focuses on temporal features of human behavior and communication. In this thesis, multiple datasets of digital activity have been analyzed. These data are of various types, from communication timestamps to sociodemographic data. The main focus of this work is to understand temporal patterns of human behavior, such as daily and weekly patterns of communication, as well as patterns of mobile phone usage, which can be seen as proxies of times of sleep and wakefulness. Looking at these different rhythms, we find that individuals exhibit activity patterns which are unique to each person and they tend to maintain their signature activity pattern over time.
Based on their propensity to sleep at different hours of the day, people can be categorized into groups called chronotypes. By analyzing the phone usage activity, we infer their Chronotype and find that individuals with different chronotypes vary in the features of their personal social network, such as the number of their contacts. For example, we see that evening-active individuals maintain larger networks. Also, by looking at the social network of study participants we observe that evening-active people tend to be more central in the network. They also exhibit homophily, which is absent for morning-active individuals.
Recently, much effort has been made to design studies which combine different devices and data sources to collect data from individuals with the goal of addressing specific questions and trying to tackle societal challenges such as the spread of diseases or issues of mental health. We have worked in a multi-disciplinary group to design a prototype data collection platform, which is currently being used for projects ranging from mental health to neuroscience studies.
|Publication status||Published - 2017|
|MoE publication type||G5 Doctoral dissertation (article)|
- computational social science, temporal patterns, big data, social networks, data collection studies