Dissertation Defense

Improving Usability of Mobile Applications Through Speculation and Distraction Minimization

Kyungmin Lee

We live in a world where mobile computing systems are increasingly integrated with our day-to-day activities. People use mobile applications virtually everywhere they go, executing them on mobile devices such as smartphones, tablets, and smartwatches. People commonly interact with mobile applications while performing other primary tasks such as walking and driving (e.g., using turn-by-turn directions while driving a car). Unfortunately, as an application becomes more mobile, it can experience resource scarcity (e.g., poor wireless connectivity) that is atypical in a traditional desktop environment. When critical resources become scarce, the usability of the mobile application deteriorates significantly.

In this dissertation, I create system support that enables users to interact smoothly with mobile applications when wireless network connectivity is poor and when the user's attention is limited. First, I show that speculative execution can mitigate user-perceived delays in application responsiveness caused by high-latency wireless network connectivity. I focus on cloud-based gaming, because the smooth usability of such application is highly dependent on low latency. User studies have shown that players are sensitive to as little as 60 ms of additional latency and are aggravated at latencies in excess of 100ms. For cloud-based gaming, which relies on powerful servers to generate high-graphics quality gaming content, a slow network frustrates the user, who must wait a long time to see input actions reflected in the game. I show that by predicting the user's future gaming inputs and by performing visual misprediction compensation at the client, cloud-based gaming can maintain good usability even with 120 ms of network latency.

Next, I show that the usability of mobile applications in an attention-limited environment (i.e., driving a vehicle) can be improved by automatically checking whether interfaces meet best-practice guidelines and by adding attention-aware scheduling of application interactions. When a user is driving, any application that demands too much attention is an unsafe distraction. I first develop a model checker that systematically explores all reachable screens for an application and determines whether the application conforms to best-practice vehicular UI guidelines. I find that even well- known vehicular applications (e.g., Google Maps and TomTom) can often demand too much of the driver's attention. Next, I consider the case where applications run in the background and initiate interactions with the driver. I show that by quantifying the driver's available attention and the attention demand of an interaction, real-time scheduling can be used to prevent attention overload in varying driving conditions.

Sponsored by

Professors Jason Flinn and Brian Noble