In this project, researchers are applying recent advances in psychometric techniques to the study of alternative forms of assessment in mathematics and science education. We have collected data from two contexts in which alternative formats based on constructivist approaches to assessment are being used on a state-wide scale: four assessment formats developed by the California Learning Assessment System, including enhanced multiple choice items, open-ended problems, investigations, and portfolios; and performance assessments being used in the Third International Mathematics and Science Study (TIMSS). Both of these contexts involve the use of complex and untried formats that have the potential to inform us of students' performance with respect to a number of different variables, that involve the use of teachers and experts as raters, and that use several different sources of information to arrive at an appropriate report for each of several possible audiences.The work of the project is a) to develop flexible and appropriate psychometric models; b) to develop computer programs that use these models to calibrate items and raters, that moderate student performances, and that produce different forms of information for different audiences; and c) within the contexts of the Californian and TIMSS situations mentioned above, to establish valid and reliable benchmarks for such data, to examine the operating characteristics of these alternative assessment systems, to study the application of such techniques longitudinally, and to develop appropriate forms of dissemination of the information generated.
Mark Wilson is the principal investigator.