Liu, K. "A Personal, Mobile System for Understanding Stress and Interruptions"

Abstract

There is a variety of emerging health applications that monitor a person's physiological signals over time -- from a growing interest to monitoring elders in cognitive decline, to systems that monitor and understand stress. While some may remain implicit monitoring devices, there is a subset of them that require interactions with a user. This thesis focuses on the challenges inherent to an interactive, health system " how and when to interact with a user. This research aims to improve these systems by using: 1) affect-sensitive strategies though social-emotional dialogue, and 2) interruption-sensitive strategies through adjusting the timing of these interruptions. An interactive, health application has been developed for data collection, annotation, and feedback that is part of a longer-term research plan for gathering data to understand more about stress, the physiological signals involved in its expression, and how that might interplay with interruptibility. The system has been developed on a mobile platform and uses affect and interruption-sensitive strategies to engage users and allow for real-time annotation of stress, activity and timing information through text and audio input. The platform supports continuous, wireless, and non-intrusive collection of heart signal data, accelerometer, and pedometer information, as well as automatic labeling of location information from context beacons. This system is the first of its kind to be affect and interruption-responsive " to use physiological data to adjust the timing of interruptions, as well as to adaptively respond with dialogue and relational strategies that specifically address the user's stress levels and the disruption the device may be incurring upon the user.

The system has been evaluated with seven subjects who used either the responsive or non-responsive system for four days, then used the opposite system for another four days, and finally, were asked to choose which system to continue interacting with for the last four days. The affect and interruption responsive system was rated as significantly less stressful on users, and five out of seven of the subjects chose to continue working with the responsive system. This study has demonstrated that designing platforms that are relational and responsive to a person's affect can facilitate a less frustrating and more enjoyable experience over time, even in tasks that are highly disruptive.

Overall, this thesis has contributed not only a new system for gathering annotations useful for studies of stress, but also provided new insights into the value of using relational and attentional strategies in a task that involves a large number of interruptions.