OPTIMIZING PERFORMANCE AND SATISFACTION IN VIRTUAL REALITY ENVIRONMENTS WITH INTERVENTIONS USING THE DATA VISUALIZATION LITERACY FRAMEWORK

Abstract

In the age of big data, interactive data visualizations are becoming increasingly prevalent. The ability to focus on subsets of data is essential for exploring large temporal, geospatial, topical, and network datasets. In order to categorize all parts of data visualizations, various attempts have been made to build frameworks for how to interpret, teach, and construct data visualizations while turning data into insights. This need for interactive data visualizations has sparked interest to classify key interaction types. The recent rise of affordable virtual reality (VR) has introduced yet more ways of interacting with datasets using our visual abilities while enabling user input beyond mouse and keyboard. In this dissertation, we apply the Data Visualization Literacy Framework (DVL-FW) in a series of three interconnected VR user studies with 152 subjects representing around 123 hours of face-to-face data collection. In the first study, we compare performance and satisfaction across two VR and one desktop implementation of the same 3D manipulation interface. We found that while VR users are about three times as fast and about a third more accurate in terms of rotation than desktop users, there are no significant differences for position accuracy. Building on this experiment, in the second study, we investigate quantitative differences between two user cohorts, where the experiment cohort gets to inspect their own manipulation performance data between trials in a VR setup and with a traditional 2D line graph, depending on their assigned setup (“Reflective phase”). Our findings indicate that while there is no difference in performance between VR users across cohorts, the Reflective phase yields significant differences for desktop users and increases the satisfaction for VR users. Moreover, we identified behavioral metrics for VR users in the Reflective phase that have a favorable effect on performance in subsequent trials. Finally, in the third study, we asked users to travel to various points inside a virtual 3D model of Luddy Hall on the Indiana University campus in Bloomington, IN. We then tested whether the experiment cohort was able to devise faster movement strategies after a Reflective phase where they inspected their own navigation data in a VR visualization with a miniature model of the building. We found that users with a Reflective phase in VR have significantly faster completion times in the second set of trials than those who did not, while also scoring significantly higher on a mid-questionnaire about the topology of the virtual building. Our methodology combines quantitative and qualitative surveys with a VR software and hardware solution that can be used in future human-subject studies by others. In addition to contributing to the theory of data visualization, specifically the interaction typology of the DVL-FW, this dissertation provides evidence-based design recommendations for matching and movement tasks in VR and on desktop devices by comparing task completion time, accuracy, and user satisfaction across different implementations of the same application. Further, we derive design guidelines for interventions using data visualizations for subjects to reflect on their behavior in VR to improve performance and satisfaction metrics in future trials.