Labelling Tappigraphy: A Ground Truth Study to Understand Mobile Map App Usage in Daily Life

Mobile map applications have evolved into interactive tools offering functionalities beyond navigation. However, the context in which these apps are used is barely researched. This gap is caused by the limited availability of privacy-conscious data, which accurately reflects real-world mobile map app usage. To address this, a new method called tappigraphy has recently been introduced to map app research. Tappigraphy only collects timestamps of taps and the app used, without capturing any other sensitive information. Therefore, by itself, tappigraphy does not reveal the specific tasks users perform within map apps. This thesis aims to address this limitation by collecting and labelling Google Maps ground truth data. This data was used to train a supervised machine learning classifier, which was then applied to a real-world tappigraphy dataset. The ground truth tappigraphy data was collected in a controlled lab environment. Participants performed predefined tasks based on the previous research of Savino et al. (2021), allowing direct comparison to that study. The classification successfully replicated known usage patterns, especially when run solely on Google Maps data. Applying the classifier to other app combinations, such as similar map apps and public transport apps, revealed distinct patterns, which reflect app functionalities and assumed user behaviour. Concluding, this thesis demonstrates the potential of using supervised machine learning to enrich tappigraphy data for task-specific within-app usage analysis. This allows an improved understanding of mobile map app usage context, which can inform the design of context-aware map applications.