ConMap: Investigating new computer-based approaches to assessing conceptual knowledge structure in physics
Beatty, Ian D.

University of Massachusetts Amherst Ph.D. dissertation.

There is a growing consensus among educational researchers that traditional problem-based assessments are not effective tools for diagnosing a student's knowledge state and for guiding pedagogical intervention, and that new tools grounded in the results of cognitive science research are needed. The ConMap ("Conceptual Mapping") project, described in this dissertation, proposed and investigated some novel methods for assessing the conceptual knowledge structure of physics students.

A set of brief computer-administered tasks for eliciting students' conceptual associations was designed. The basic approach of the tasks was to elicit spontaneous term associations from subjects by presenting them with a prompt term, or problem, or topic area, and having them type a set of response terms. Each response was recorded along with the time spent thinking of and typing it.

Several studies were conducted in which data was collected on introductory physics students' performance on the tasks. A detailed statistical description of the data was compiled. Phenomenological characterization of the data (description and statistical summary of observed patterns) provided insight into the way students respond to the tasks, and discovered some notable features to guide modeling efforts. Possible correlations were investigated, some among different aspects of the ConMap data, others between aspects of the data and students' in-course exam scores. Several correlations were found which suggest that the ConMap tasks can successfully reveal information about students' knowledge structuring and level of expertise. Similarity was observed between data from one of the tasks and results from a traditional concept map task.

Two rudimentary quantitative models for the temporal aspects of student performance on one of the tasks were constructed, one based on random probability distributions and the other on a detailed deterministic representation of conceptual knowledge structure. Both models were reasonably successful at approximating the statistical behavior of a typical student's data.

beatty-2000cic.pdf867 KB