Analyzing Constructed Responses: Using Linguistic Software to Understand Students' Conceptions in Science

Michigan State University, East Lansing MI

Investigators

Abstract

Assessment/Research (91) Biological Sciences (61) Geology (42) This project is developing an efficient means for revealing students' conceptual understanding and conceptual barriers in introductory science courses. The starting point is student constructed responses. These require students to explain scientific phenomena and thus have good potential to reveal student conceptual barriers. Because of resource limitations, student assessment is often undertaken using multiple choice items rather than constructed response items. The costs and logistics of evaluating constructed responses are prohibitive in large enrollment courses that are typical of most introductory science courses in large institutions. This project is investigating multiple approaches to applying advances in computerized linguistic analysis in order to reduce the time and cost of evaluating constructed response assessments. These multiple approaches are being used to extract conceptual categories from student responses and using these data to build statistical models of students' application of those concepts in the disciplines of biology and geology. The data are being derived from students in large enrollment, introductory undergraduate courses, each with an enrollment of several hundred students per semester. Intellectual Merit: These models are likely to be useful for improving educators' understanding of student conceptual barriers and as a check on other research seeking to diagnose student conceptual barriers. The project team is an interdisciplinary group with expertise in biology, geology, science education, teacher education, statistics, and instructional technology, and has experience in working together. Broader Impacts: Changes in instruction to focus on student learning rather than on teaching has the potential for large impact. Using assessments that reveal students' conceptual barriers will help instructors to improve teaching by addressing the conceptual difficulties that contribute to attrition in STEM disciplines. At the completion of this project, instructors across the scientific disciplines will have access to a new tool for assessing essay responses, even in large enrollment courses.