A recent paper from the AACR group, "From substitution to redefinition: A framework of machine learning‐based science...

The Automated Analysis of Constructed Response research group is beginning a new collaboration with researchers, Dr. Chelsie...

Automated Analysis of Constructed Response

The Automated Analysis of Constructed Responses (AACR, pronounced "acer") research group consists of researchers from seven universities (visit Collaborators page) with backgrounds in various STEM disciplines (biology, chemistry, engineering, and statistics), assessment and educational research. We hope this research can help us gain greater insight into student thinking about “big ideas”, such as evolution, energy, and genetics. We are looking to expand the scope of this work and seeking interested collaborators across STEM disciplines.

Constructed response questions – in which students must explain phenomena in their own words – create more meaningful opportunities for instructors to identify their students’ learning obstacles than multiple choice questions. However, the realities of typical large-enrollment undergraduate classes restrict the options faculty have for analyzing student writing.

In the Automated Analysis of Constructed Response research group, we are exploring the use of computerized lexical analysis of students’ writing in large enrollment undergraduate STEM courses. We have created computer models that can predict expert ratings of student responses with accuracy approaching inter-rater reliability among expert raters. These techniques also provide insight into students’ use of analogical thinking, a fundamental part of scientific modeling. These techniques have potential for improving assessment practices across STEM disciplines.

thumbnail of small NSF logo in color without shading

This material is based upon work supported by the National Science Foundation (DUE grants: 1438739, 1323162, 1347740, 0736952 and 1022653). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSF.