Session T1B AUTOMATED EVALUATION METHODS WITH ATTENTION TO IN C INDMDUAL DIFFERENCES- A STUDY OF A COMPUTER-BASED COURSE
Tammy Rosenthal, Patrick Suppes, Nava Ben-Zvl · 2002
This studyfocuses on students'performance in a distance-learning computer-based C course offered by EPGY (Education Program for Gifred Youth) at Stanford University. Its main goal is testing the feasibility of automated evaluation. Three automa red methods consisting of a criterion set and weighted scoring methods were developed empirically, then applied to four hundred programs often types ofproblems and thoroughly analyzed. For most of the assignments, the more sign$cant criteria for evaluation were related to execution time features, (i.e. number ofoperations and calls lofunctions executed by the program), memory storage required to store the variables of the program, and code's length. The results showed that the criterion measures o/en differed by several orders of magnitude between students'programs and were dependent on the specific problem domain. A low correlation was found between the evaluation criteria indicating rhat the criteria we nearly independent. Consequently, in qualitative tests for achievement and ability students' performances exhibited high dimensionality.