My research explores the interplay between corporate decisions and employee actions. I currently use anonymized mobile device data to observe individual behaviors, and employ both unsupervised and supervised machine learning techniques.
My research examines the ways in which individuals and organizations use data to improve. Quality improvement and data-intensive research approaches are central to my work along with forming equitable collaborations between researchers and frontline workers. Prior to joining the Department of Learning Health Sciences, I was the Director of Learning Analytics Research at Digital Promise and a Senior Education Researcher in the Center for Technology in Learning at SRI International. At both organizations, I developed data-intensive research-practice partnerships with educational organizations of all types. As a learning scientist working at the intersection of data-intensive research and quality improvement, my colleagues and I have developed tools and strategies (e.g., cloud-based, open source tools for engaging in collaborative exploratory data analyses) that partnerships between researchers and practitioners can use to measure learning and improve learning environments.
This is an image that my colleagues and I, over multiple projects, developed to communicate the multiple steps involved in collaborative data-intensive improvement. The “organize” and “understand” phases are about asking the right questions before the work of data analysis begins: “co-develop” and “test” are about taking action following an analysis. Along with identifying common phases, we have also observed the importance of the following supporting conditions: a trusting partnership, the use of formal improvement methods, common data workflows, and intentional efforts to support the learning of everyone involved in the project.
My research focuses on understanding, designing, and evaluating learning technologies and environments that foster collaborative problem solving, spatial reasoning, engineering design thinking and agency. I am particularly interested in applying multimodal learning analytics in the context of co-located and/or virtually distributed teams in clinical simulations. I strive to utilize evidence in education science, simulation-based training and learning analytics to understand how people become expert health professionals, how they can better work in teams and how we can support these processes to foster health care delivery and health outcomes.
Before joining the faculty at the University of Michigan in 2018 as Professor and Marion Elizabeth Blue Chair of Children and Families, I was Co-Director of the 3DL Partnership at the University of Washington, where I collaborated with academic colleagues, students, and service providers throughout the state to conduct and translate research on social emotional learning (SEL) and trauma-informed practices. I am now pursuing a similar line of research in Michigan, where I am collaborating with state partners and to identify, develop, and refine new approaches to disseminate research for schools and early childhood settings engaged in SEL and trauma work. As a scholar, I am committed to increasing the visibility, application, and sustainability of evidence-based programs and practices relevant to these topics and have worked extensively in the U.S. and internationally to advance goals for prevention and the promotion of child well-being.
I am interested in the intersection of big data, data science, privacy, security, public policy, and law. At U-M, this includes co-convening the Dissonance Event Series, a multi-disciplinary collaboration of faculty and graduate students that explore the confluence of technology, policy, privacy, security, and law. I frequently guest lecture on these subject across campus, including at the School of Information, Ford School of Public Policy, and the Law School.
Shu-Fang Shih, Ph.D., has a diverse background in public health, business administration, risk management and insurance, and actuarial science. Her research has focused on design, implementation, and evaluation of theory-based health programs for children, adolescents, pregnant women, and older adults in various settings. In addition, she used econometric methods, psychometric, and other statistical methods to examine various health issues among children, adolescent, emerging adulthood, pregnant women, and the older adults. She is particularly interested in designing effective ways to align public health, social services, and healthcare to achieve the goal of family-centered and integrated/coordinated care for the family.
Dr. Gongjun Xu is an assistant professor in the Department of Statistics at the University of Michigan. Dr. Xu’s research interests include latent variable models, psychometrics, cognitive diagnosis modeling, high-dimensional statistics, and semiparametric statistics.
Analysis of a large corpus of student writing in response to concept-focused prompts. Data includes drafts, peer feedback, and revisions for each of approximately 60,000 document and 180,000 pieces of feedback. The goal is to use tools of natural language processing to develop the capacity to provide students and faculty with feedback that will foster deeper conceptual learning among students and enhanced pedagogy among faculty.
The Shultz group uses data science methods in two primary ways 1) to investigate student placement in introductory chemistry courses and 2) to analyze student texts to provide instructors actionable intelligence about student learning. Using regression discontinuity we investigated the impact of taking general chemistry prior to organic chemistry on student performance and persistence in later chemistry courses and found that students who took general chemistry first benefitted by 1/4 of a letter grade but were no more likely to persist. A continued investigation using survey and interview methods indicated that this was related to academic skills rather than content preparation. Through the MWrite project we have collected a large corpus of student texts and are developing automated text analysis methods to glean information about student learning across disciplines, with specific focus on scientific reasoning.