Kevin Haudek is an Assistant Professor in the department of Biochemistry & Molecular Biology at Michigan State University. His talk was titled "Using Automated Analysis to Reveal Student Thinking."
Abstract
In the Automated Analysis of Constructed Response (AACR) Research Group, we develop constructed response assessment items and associated predictive machine learning models that allow formative evaluation of student writing. These models are particularly informative for revealing prevalent student ideas in large-enrollment, introductory STEM courses. We employ a mixed-methods approach to develop these questions and analytic tools, which reveal a rich and complex picture of student thinking about important disciplinary ideas across biology, statistics and chemistry. Recently, our group launched projects that develop assessment items aligned with learning progressions in physiology for undergraduates and scientific argumentation for middle school students. We have also begun to explore for possible interactions between predictive scoring models and responses from English language learner middle school students.
This talk will provide a brief overview of the methodology we employ to develop automated predictive scoring models, examples of developed items, and our attempts to summarize and report this rich information to college faculty. Additionally, this talk will report on recent projects aimed at adapting these approaches to learning progressions.
