Example REU Projects
Assisting Kids in Monitoring Snacking and Physical Activity
We design, develop, and evaluate mobile and wearable embedded systems that empower children and their caregivers to manage their snacking and physical activity. In a multi-team project between Indiana University and the University of Colorado Boulder, PI Siek leads efforts to develop Health Sense, a system of plug-and-play wellness monitoring, wearable components for children that are easy to build and program (NSF Grant: IIS-1231645). Health Sense provides children with the ability to reflect on their health by crafting their own personalized, monitoring system and empowers them to consider preventative health measures (play outside to make a headband light up) instead of specific treatments (play for 30 minutes). The ProHealth REU site can assist the researchers in better classifying activities and thereby improving feedback. The Health Sense research team has industry (e.g., SparkFun) and local collaborators (e.g., Harmony School) to ensure students can contribute to the Health Sense project.Connelly leads a research team to design mobile technology to assist children and their families collaboratively monitor an obese child’s nutrition and physical activity through a local weight management program, GOAL, at IU Health. The team is iteratively designing and implementing a mobile system that provides children with the ability to set goals, log progress, and reflect on their progress before attending weekly meetings facilitated by health professionals. The 12- week GOAL program continuously runs throughout the year, thus the researchers will easily be able to continue iterating on the application during the REU project. The ProHealth REU can assist the researchers by providing more personalized feedback and carefully investigating the privacy expectations between stakeholders. Intellectual Merit: Our research has led to advances in the impact of craft on health and wearable technologies potential to improve communication between healthcare providers and patients.
Student Participation: Four REU students can participate in this research area – 2 advised by PI Siek and 2 advised by Connelly. Graduate students who work on each project will co-mentor the students. The Health Sense students will create Health Sense training materials, design a Health Sense component, or develop a Health Sense programming module. GOAL students will assist in iteratively designing and developing the mobile interface via user studies. We can train programming skills during the summer.
Empowering Older Adults to Stay Connected
Kay Connelly has had two NSF-funded projects (IIS-0705676, IIS-1117860) investigating how to use technologies to help older adults stay in their homes for as long as possible. Both projects pay special attention to the privacy aspects of data collection in the home, and sharing with informal caregivers and friends. This research has a heavy design component, in which students participate in weekly design meetings as a particular technology evolves. As an example, an REU led the design team for the check-in tree. The check-in tree allows older adults to check-in with each other in the morning by pressing a button on the base of the display. LEDs near pictures of their friends lets them know who in their group of friends has also checked-in. If someone does not check-in by their normal time, friends can make sure they are okay. Intellectual Merit: Our research extends other aging-in-place research by investigating how to design home-based technologies for low-SES older adults. In addition to making privacy a key design factor, our work explores how to assist older adults in helping take care of each other, as opposed to only providing data to younger caregivers.
Student Participation: Two REU students can participate in this project advised by Connelly and co- advised by the project’s graduate students. Students will help design new technologies and deploy them in the older adults’ homes. Both technical and nontechnical students can participate.
Providing Personalized Educational Information for Rural Pregnant Women
The Mothers’ Information Technology for Education (MITE) project aims to design a mobile technology application that will provide supportive and effective messages to improve rural mother’s health and that of their infant child. Effective communication and support of expecting and recent mothers is a central, but difficult, component of successful public health interventions for improving maternal and child health. It is particularly difficult among rural and resource restricted communities, like those in Southern Indiana. Text based messaging interventions aimed at improving maternal and child health are generally centered on prenatal behaviors and are also a mostly one-size-fits-all approach which lacks the adequate tailoring of health messages which is necessary to make them effective agents of support. Therefore, there is a critical need to provide socially and culturally adequate messages, especially extending into the postpartum period. In this interdisciplinary project, PI Siek works with Dr. Lucia Guerra-Reyes in the IU School of Public Health to better understand rural mother’s needs and design personalized systems. The MITE research team has established collaborations with Monroe County Women, Infants, and Children (WIC) and Bloomington Area Birth Services to recruit mothers. Intellectual Merit: This research will make contributions to the fields of pervasive computing on how to design mobile systems for rural mother’s and effectively disseminate educational information.
Student Participation: Two REU students can participate in the MITE project advised by PI Siek and co- advised by the graduate students on the project. Students will contribute by helping to develop the MITE infrastructure or the design of educational messages for small screen or text messages. Programming experience is needed for the infrastructure, whereas design expertise is needed for the messaging.
Future Self Projections
The Coronary Artery Risk Developments in Young Adults (CARDIA) study is a longitudinal study of cardiovascular risk factors that began in 1985-86 with repeated measures after 2; 5; 7; 10; 15; 20 and 25 years. The purpose of this project is to understand the relationship between the measured risk factors and the development of CVD. As the cohort ages and sufficient clinical events occur, this work will allow us to predict clinical events such as Acute Myocardial Infarction and heart failure. We propose to use the wealth of longitudinal data collected from the CARDIA study over 25 years of early adult life to develop machine learning models that can be used to predict who is at greatest risk in their 20’s of developing CVD in early adult life (30- 55 years of age). We are developing novel Statistical Relational Learning (SRL) algorithms that can operate directly on the relational data instead of considering a flat (feature based) representation of the data. The REU students will compare these algorithms against standard supervised learning algorithms including decision trees, support vector machines (SVMs), Bayesian Network algorithms such as Naive Bayes, and ensembles of these methods using Bagging and Boosting. This project is built on strong collaborations with multiple Midwestern hospitals and clinics. Intellectual Merit: The algorithms developed in this project will be the first of its kind for learning on multi- modal clinical data for predicting early risks of cardiovascular events. These activities will provide the groundwork for follow-up research in clinical decision support information presentation.
Student Participation: We anticipate two REU students to be involved in the project. The two students will evaluate the standard learning algorithms on the CARDIA data using Weka to enable us to better understand the particular correlations that are exploited by these learning models. The students will be provided machine learning background by continuously training them on the learning algorithms with senior graduate students. The materials are based on the “Data Fluency” course (for 1st year undergraduates) that was developed and taught by Natarajan1. Course feedback indicated that students were able to understand the basic ideas on learning and evaluation of the algorithms along with the interpretation of the results without significant background knowledge. Minimal programming background is needed for analysis using Weka.
Social and behavioral scientists often collect and maintain datasets that are highdimensional, including some combination of demographic, medical, sexual, and other personal information, which presents opportunities to characterize participants in unique ways. Professor Hill leads initiatives to evaluate the risk of re-identification in social science datasets, and evaluate the utility of differentially private behavioral science data. We evaluate the statistical characteristics of two highdimensional social science datasets to better understand how unique features impact privacy. We can apply a class of statistical de-anonymization attacks in an attempt to achieve theoretical re- identification of participants. We assume that an attacker has exact knowledge of a subset of attribute values for a particular record, and wants to link this subset of data to the actual record to discover the remaining content. We show that although 98% of the records within the dataset are unique given any three attributes, re-identification of the records may not be easily achieved. We attribute limited re- identification to the inherent similarity in the human behavior that the scientists measure. The conventional wisdom for protecting the privacy of such participants in behavioral science datasets is to either not ask certain questions or to remove or recode potentially identifiable information. Neither approach may be sufficient for preventing the (re)identification of participants in large and/or multidimensional datasets. The risk to privacy is also increased when survey participants are allowed to enter unrestricted text responses. These survey responses can include identifying information such as location, family members, etc; spelling errors; and parts of speech that modify the intended response. This unstructured data are likely to be unique and unusable for statistical analysis and classification, thus limiting its use for only descriptive purposes, if at all. In we focus on reducing the uniqueness, and enhancing the analyzability, of a data set that includes unrestricted text responses from a medical survey. During the ProHealth REU, students will file for appropriate IRB amendments and evaluate data that PI Siek and Connelly collected in the past for their projects to identify data collection privacy issues and make design suggestions on how to improve user privacy in study instruments. Intellectual Merit: This work is the first to characterize re-identification risks in highdimensional data that is collected in surveys designed to capture the various behaviors and experiences of groups of individuals.
Student Participation: Two REU students will participate in this project to assist in collecting data from past research studies by Siek and Connelly, clean the data, analyze the data for potential privacy risks, and report back on improvements for privacy protections. Students can be trained during the summer based on Hill’s course materials and mentoring by the graduate student working with Hill.
Social-Legal Issues Pertaining to Pervasive Health Systems
Student Participation: Two REU students will participate. Both will be taught ethnographic and legal research techniques during daily meetings with Allison Fish during the program’s first two weeks.