Interaction mining mobile apps

Millions of mobile apps are used by billions of users every day. Although the design of these apps play an important role in their adoption, the design process still remains complex and time intensive. At the same time, existing apps embody multiple solutions to numerous design problems faced by app developers. How do we make this design knowledge embedded in existing apps accessible to designers? And how can it help simplify the app design process? This dissertation introduces interaction mining, a technique to capture the designs of mobile apps in a way that supports data-driven design applications. It presents systems that implement interaction mining for Android apps without requiring any access to their source code making it possible to design mine apps at an unprecedented scale. It presents Rico, the largest publicly available mobile app design repository to date. It discusses how such repositories created using interaction mining can be used to train models that enable applications such as keyword and example-based search interactions for mobile screens and user flows. It also presents zero-integration performance testing (ZIPT), a novel technique for testing app designs. It demonstrates how ZIPT can be used to help designers understand which examples to draw from in the early stages of the app design process and perform comparative testing at scale with low cost and effort in the later stages of the process