Context in mobile system design: characterization, theory, and implications

Context information brings new opportunities for efficient and effective applications and services on mobile devices. Many existing work exploit the context dependency of mobile usage for specific applications, and show significant, quantified, performance gains by utilizing context. In order to be practical, such works often pay careful attention to the energy and processing costs of context awareness while attempting to maintain reasonable accuracy. These works also have to deal with the challenges of multiple sources of context, which can lead to a sparse training data set.Even with the abundance of such work, quantifying context-dependency and the relationship between context-dependency and performance achievements remains an open problem, and solutions to manage the challenges of context awareness remain ad-hoc. To this end, this dissertation methodologically quantifies and measures the context dependency of three principal types of mobile usage in a methodological, application agnostic yet practical manner. The three usages are the websites that users visit, the phone numbers they call, and the apps (i.e. purchased or preinstalled applications) they use. While this dissertation measures the context dependency of these three usage, its methodology can be readily extended to other context-dependent mobile usage and system resources. This dissertation further presents SmartContext, a framework to systematically optimize the energy cost of context awareness by selecting among different context sources, while satisfying system designers' cost-accuracy tradeoffs. Finally, this thesis investigates the collective effect of social context on mobile usage, by separating and comparing LiveLab users based on their socioeconomic status.The analysis and findings are based on usage and context traces collected in real-life settings from 24 iPhone users over one year. This dissertation presents findings regarding the context dependency of three principal types of mobile usage; visited websites, phone calls, and app usage. The methodology and lessons presented here can be readily extended to other forms of context and context-dependent usage and resources. They guide the development of context aware systems, and highlight the challenges and expectations regarding the context dependency of mobile usage.