Measuring Differences in Performance by Varying Formative Assessment Construction Guided by Learning Style Preferences

摘要

In this evidence-based practice paper, the relationship between assessment design guided by learning style preferences and student performance in a programming course is investigated. One of the National Academy of Engineering's 14 Grand Challenges for Engineering is to tailor and differentiate instruction to improve the reliability of learning. A manner in which this differentiation may be accomplished is through attention to the various preferences and styles by which students learn. As such, the purpose of this paper is to present evidence on the effect of formative assessment design on student performance, and whether this effect varies by student learning style. The results from this study can be used by engineering educators to either diversify or personalize their assessment style. This work is grounded in the Felder-Soloman learning style model, a model that was developed within engineering education and has been validated and widely used within the field. This model categorizes learning styles along four distinct dimensions: perception (sensing versus intuitive), input (visual versus verbal), processing (active versus reflective), and understanding (sequential versus global). Along each of these dimensions, students are categorized as having a mild, moderate, or strong preference in each of these four learning style scales. This study takes place in a mid-size, public university in the western United States. The sample for this study includes mechanical engineering undergraduate students across four sections of a required programming course in MATLAB, taught by the same instructor. These students were provided the Index of Learning Styles at the beginning of the semester. Students were administered a weekly quiz to assess their ability to write code, but construction of this assessment varies by section to favor different preferences of one of the four Felder-Soloman learning style dimensions. Performance on these quizzes is objectively scored using a standardized rubric. General linear modeling is used to determine if quiz scores differ by quiz construction condition, and if learning style preference interacts with quiz condition to predict performance on each assessment. Findings portray a complex relationship between quiz construction, learning style preference, and assessment performance.