Usable Measures of Teacher Understanding: Exploring Diagnostic Models and Topic Analysis as Tools for Assessing Proportional Reasoning for Teaching

One of the great challenges related to teachers and their knowledge is measuring their learning in ways that are both formative and meaningful in relation to their likely impact on students. This challenge persists despite efforts to define the knowledge teachers should have and despite previous innovative efforts to create good measures. This project tackles the challenge by specifically aiming to measure the kinds of knowledge developed in professional development (PD) programs that has been shown to matter for teachers' classroom practices and their students' learning. The project aims to develop an assessment that identifies patterns in the teachers' learning in a way that helps drive subsequent professional development.

The overall goal of this project is to pursue a potentially transformative approach to the assessment of teacher proportional knowledge by developing a measure that is well aligned with the content and skills taught in various PD programs. This instrument will be based on a new approach that builds on emerging psychometric models. Specifically, diagnostic classification models (DCMs) will be utilized to diagnose teachers' learning during a PD program as well as employed to identify the progression in teachers' learning.  Statistical topic models (STMs) will be used to look for patterns of understanding that emerge from open-ended responses and provide natural-language insight into teachers' reasoning. A final version of the assessment will be constructed for a national sample based on the results from the predictive validity stage, and this version will be tested with teachers who participate in various types of PD programs targeting proportional reasoning. This project has broad implications for the creation of assessments and for teacher education. It will provide insights about whether there is a clear learning progression for teachers. While much work has been done with students' learning progression, much less is known about how teachers learn. Another implication is that the STM approach allows machine scoring of natural language in a way that highlights strengths and weaknesses in reasoning rather than simply returning a score. For formative use, this is information that is more helpful as it highlights areas for further instruction. A third implication is that DCMs will allow to assess teacher knowledge at a finer-grained understanding than is typically available, thus allowing for careful refinement of PD as well as a tool for showing overall growth in PD. A fourth implication is that a more systematic approach will be followed to capture the kinds of knowledge teachers need. Assessments developed using DCMs and STMs have the potential to serve as models for developing further instruments in other STEM content areas. Such assessments have the potential to not only help identify successful PD programs, but also to provide PD providers with rich data from which they can make instructional decisions.