Abstract

Personal data, generated and continuously collected by modern devices and online services, open new perspectives in human behavior understanding. They can characterize human behavior at a very fine and precise resolution, covering a huge variety of daily life, from communication and individual mobility to complex social phenomena and economic behaviors. While from the research point of view the collection of behavioral data has never been so cost-effective and unobtrusive, at the same time, from the applicative point of view, numerous applications supporting users’ needs have leveraged on the massive availability of personal data and the insights produced by human behavior comprehension. Nevertheless, this scenario also raises unprecedented risks affecting users’ privacy. The high relevance and effectiveness of digital footprints in capturing and describing human behaviors establishes the basis of this work. In this thesis, we use behavioral personal data, collected online or from mobile phones, to understand human behavior with the aim to support users in their everyday lives, investigating aspects that can be turned into personal or societal value in real application scenarios. In our work, we approached our research problems in a comprehensive way by leveraging on real personal data continuously observed in daily-life. In particular, in this dissertation we investigate three main problems. Firstly, we understand the attitudes of individuals towards personal mobile data disclosure by using both individual characteristics and dynamic behaviors related to communication and mobility, captured from mobile phones. Secondly, we predict the future health status of individuals, in terms of flu-like and cold symptoms, on the basis of a systematic characterization of their mobility behaviors derived from their mobile phones. Finally, we infer the formation of future links in individuals’ social circle on the basis of multiple information sources related to explicit and implicit relationships that users form. The main insights resulting from our investigation can support a set of use cases, targeting individual and collective applications. Our findings, in terms of behaviors affecting the personal propensity towards data disclosure, can support a new generation of privacy-tools providing personalized feedback to users when tuning their sharing settings. The insights from our following study can provide individuals with personalized feedback about their health status and support decision-makers towards the adoption of preventive measures for public-health. Finally, our findings in terms of new social link predictions can provide users with more accurate recommendations in social networking applications.