Design and evaluation of a health-focused personal informatics application with support for generalized goal management
MetadataShow full item record
The practice of health self-management offers behavioral and problem-solving strategies that can effectively promote responsibility for one's own wellbeing, improve one's health outcomes, and decrease the cost of health services. Personal informatics applications support health self-management by allowing their users to easily track personal health information, and to review the changes and patterns in this information. Over the course of the past several years, I have pursued a research agenda centered on understanding how personal health informatics applications can further support the strategies of health self-management--specifically those relating to goal-management and behavior change. I began by developing a flexible personal informatics tool, called Salud!, that I could use to observe real-world goal management and behavior change strategies, as well as use to evaluate new interfaces designed to assist in goal management. Unlike existing personal informatics tools, Salud! allows users to self-define the information that they will track, which allows tracking of highly personal and meaningful data that may not be possible to track given other tools. It also enables users to share their account data with facilitators (e.g. fitness grainers, nutritionists, etc.) who can provide input and feedback. Salud! was built on top of an infrastructure consisting of a stack of modular services that make it easier for others to develop and/or evaluate a variety of personal informatics applications. Several research teams used this infrastructure to develop and deploy a variety of custom projects. Informal analysis of their efforts showed an unmet need for data storage and visualization services for home- and health-based sensor data. In order to design a goal management support tool for Salud!, I first, I conducted a meta-analysis of relevant research literature to cull a set of proven goal management strategies. The key outcome of this work was an operationalization of Action Plans--goal management strategies that are effective at supporting behavior change. I then deployed Salud! in two fitness-related contexts to observe and understand the breadth of health-related behavior change and goal management practices. Findings from these deployments showed that personal informatics tools are most helpful to individuals who are able to articulate short-term, actionable goals, and who are able to integrate self-tracking into their daily activities. The literature meta-analysis and the two Salud! deployments provided formative requirements for a goal management interaction that would both incorporate effective goal management strategies and support the breadth of real-world goals. I developed a model of the goal management process as the framework for such an interaction. This model enables goals to be represented, evaluated, and visualized, based on a wide range of user objectives and data collection strategies. Using this model, I was able to develop a set of interactions that allow users of Salud! to manage their personal goals within the application. The generalized goal management model shows the inherent difficulty in supporting open-ended, highly personalized goal management. To function generically, Salud! requires facilitator input to correctly process goals and meaningfully classify their attributes. However, for specific goals represented by specific data collection strategies, it is possible to fully- or semi-automate the goal management process. I ran a large-scale evaluation of Salud! with the goal management interaction to evaluate the effectiveness of a fully-automated goal management interaction. The evaluation consisted of a common health self-management intervention: a simple fitness program to increase participants' daily step count. The results of this evaluation suggest that the goal management interaction may improve the rate of goal realization among users who are initially less active and less confident in their ability to succeed. Additionally, this evaluation showed that, while it can significantly increase participants' step count, a fully automated fitness program is not as effective as traditional, instructor-led fitness programs. However, it is much easier to administer and much less resource intensive, showing that it can be utilized to rapidly evaluate concrete goal management strategies.