Visualization of Continuous Sensory Data for Stress Management

Collaborator: Dr. Santosh Kumar (CS, Univ. of Memphis), Dr. Gayle Beck (Psychology, Univ. of Memphis), Dr. Andrew Raij (CS, Univ. of Central Florida), Dr. Kenzie Preston (NIDA, NIH), Dr. David Epstien (NIDA, NIH)

Designed five novel stress visualization techniques based on an analysis of 1,143,156 data points collected in natural living conditions. The visualizations are grounded in health behavior theory, stress management literature, and existing research on health behavior change. The proposed visualizations are targeted to address the unique challenges posed by stress management such as facilitating identification of patterns in a substantial amount of data collected from the natural environment, aiding the perception of self-efficacy by enabling access to personalized stress profiles, highlighting patterns of stress, and enhancing awareness of context-dependent stressors. The visualizations are evaluated in an exploratory evaluation technique.

Assessing User Privacy Risks Emerging from Wearable Sensors

Collaborator: Dr. Santosh Kumar (CS, Univ. of Memphis), Dr. Mani Srivastava (ECE,UCLA)

Investigated privacy risks associated with collecting and sharing GPS, activity, physiology, and audio data in daily life. We use requested compensation (N=57) as an objective measure of perceived privacy concern. Carrying a phone throughout the day with GPS turned on is our base condition. We find that when smart watch with inertial sensors is added, users ask for 19% more in compensation. Subsequent addition of a chest band with ECG and respiration sensors leads to a steep rise in the requested compensation (i.e., by 45%). When an audio recorder is finally added, participants in our survey show diminishing return effect, as they ask for only 10% more. Our findings inform the selection of wearable sensors in future studies to minimize user burden and privacy issues.

Assessing User Availability in the Field

Collaborator: Dr. Santosh Kumar (CS, Univ. of Memphis)

Wearable wireless sensors for health monitoring are enabling the design and delivery of just-in-time interventions (JITI). Critical to the success of JITI is to time its delivery so that the user is available to be engaged. We proposed a model of users availability by analyzing 2,064 hours of physiological sensor data and 2,717 self-reports collected from 30 participants in a week-long field study. We used delay-in-responding to a prompt to objectively measure availability. We computed 99 features and identified 30 as most discriminating to train a machine learning model for predicting availability. We report that location, affect, activity type, stress, time, and day of the week, play significant roles in predicting availability. Our model finally achieves an accuracy of 74.7% in 10-fold cross-validation and 77.9% with leave-one-subject-out.

Development and Evaluation of a Contextual Recommender to Facilitate Reuse of Presentation Materials

Advisor: Dr. Lawrence Bergman, Dr. Jie Lu, Dr. Ravi Konuru (IBM Thomas J. Watson Research Center)

Designed and implemented a slide-based contextual recommender, ConReP, to support reuse of presentation materials. Applied information retrieval techniques to develop models to recommend slides based on their contextual similarity without requiring user provided keywords for the search-task. These models were used to create a local-context-based visual representation of the recommendations. Evaluation of ConReP in a lab-study revealed that slide-based search is more effective than keyword-based search, local-context-based visual representation helps in better recall and recognition, and shows the promise of this general approach of exploiting individual slides and local-context for better presentation reuse. This resulted in US Patent 13/370,868.

Development of a Framework to Facilitate Recall and Recognition of Personal Information

Advisor: Dr. Lawrence Bergman, Dr. Jie Lu, Dr. Ravi Konuru (IBM Thomas J. Watson Research Center)

Performed grounded-theory based data analysis to better understand strategies that users adopt to support recall and recognition of their own materials. Applied behavioral psychology theory coupled with findings from the users’ practice to develop a framework for supporting recall and recognition of personal information. This resulted in US Patent 13/370,868.