DiCS-Index: Predicting Student Performance in Computer Science by Analyzing Learning Behaviors

Abstract

Many students with little pre-college exposure to computer science (CS) share widespread incorrect ideas and a negative attitude towards the subject, leading to wrong decisions when choosing their major. In order to support these students, we developed a questionnaire built on Kolb's and Pask's learning style theories. Our aim was to create an instrument that allows to predict student performance in CS based solely on non-subject specific information. Using 62 items from two questionnaires to operationalize the three personality traits as described by Kolb and Pask, we selected a subset of 15 items by comparing the results of students with high and low achievements in CS. Subsequently, we determined the so-called DiCS-Index by adding up the values of all these 15 items, where a high DiCS-Index suggests a good performance in CS. Finally, the instrument was tested at our local CS department. The analysis of the personality traits suggests that CS, as a course of studies, is open to a highly heterogeneous student body with varying preferences and strengths. The only significant difference found is a clearly better performance of students who prefer learning through abstract conceptualization as opposed to gathering concrete experience. Concerning the questionnaire, we found a clear distinction between students with high and low achievements indicated by a highly significant difference in their corresponding DiCS-Indices.