Using Educational Technology to Support Inferences About Collaborative Problem Solving

Peter Halpin

Traditional educational tests target a relatively narrow set of constructs compared to the range of competencies required for student success. It has been recognized that learning technology can help address this problem by providing a wider range of performance-based contexts that also have better fidelity to the situations in which students learn and are expected to demonstrate their knowledge. Additionally, learning technology enables the collection of fine-grained data on students’ task-related activities, and can also allow for the real-time statistical analysis of such data, which can be used to support inferences about student competencies in learning contexts. However, these new data sources are often difficult to interpret with respect to psychological and educational constructs, and are likely to violate assumptions of traditional psychometric models. This project has three interrelated research goals. The first is to develop reliable and robust measurement methods for difficult-to-measure competencies. The second is to explore the use of serially dependent data available from learning technology to provide contextualized, formative feedback to learners and educators. The third is to apply these methodological considerations to a specific target construct, namely collaborative problem solving.This project has three interrelated research goals. The first is to develop reliable and robust measurement methods for difficult-to-measure competencies. The second is to explore the use of serially dependent data available from learning technology to provide contextualized, formative feedback to learners and educators. The third is to apply these methodological developments to a specific target construct, namely collaborative problem solving.The focus on collaborative problems solving is motivated by the following considerations. First, collaboration, team work, and other interpersonal skills are increasingly being recognized as important domains of student competency. Second, educational measurement should be guided by research and theory that informs our understanding of student learning, not conducted in isolation. Third, the statistical procedures that I have developed for the time series analysis of dyadic interaction can be used to make substantial progress on this topic.

About Peter Halpin

Peter Halpin is an Assistant Professor of Applied Statistics whose research focuses on educational measurement and psychometrics. He received a Ph.D. in Psychology from Simon Fraser University in 2010, and held a postdoctoral research position at the University of Amsterdam through 2012. His recent research projects include (a) the use of confirmatory factor analysis to address the influence of test anxiety on test performance, (b) time series methods for computer-mediated human interaction, and (c) measurement models for improving the reliability and diagnostic utility of in-classroom observations of teachers. His research on the latter topic has been supported by a Measures of Effective Teaching Early Career Grant (NAEd/Spencer) as well as an Early Career Grant from the Statistics and Research Methodology Program of the Institute of Education Sciences. His work has been published in outlets including Psychometrika, Structural Equation Modeling, Multivariate Behavioral Research, and Educational Researcher.