Publications & Resources
Exploring the Dynamics of Complex Problem-Solving With Artificial Neural Network-Based Assessment Systems
Karen Hurst, Adrian Casillas, and Ronald H. Stevens
In this study of computer-based assessment, UCLA researchers found that artificial neural network-based (ANN) assessment procedures can more efficiently score medical student problem-solving skills than some traditional approaches. An ANN developed by the research team was able to detect and quantify student problem-solving strategies, discriminate among subtle changes in strategy within a relevant concept domain, and measure medical student hypothesis formation and hypothesis refinement. Another observation derived from the use of the ANN in this study was that students significantly improved their problem-solving skills between practice and examination performances. In addition to the assessment of a large number of performances, ANNs can free “valuable instructor time for modifying curricula to reach a maximum number of students,” concluded the researchers.
Hurst, K., Casillas, A., & Stevens, R. H. (1997). Exploring the dynamics of complex problem-solving with artificial neural network-based assessment systems (CSE Report 444). Los Angeles: University of California, Los Angeles, National Center for Research on Evaluation, Standards, and Student Testing (CRESST).