Improving Learning and Teaching at Universities: The Potential of Applying Automatic Essay Scoring with Latent Semantic Analysis

The aim of this dissertation was to find ways to improve learning and teaching at universities by analyzing whether the application of new technologies would facilitate the implementation of an effective teaching-learning format in which students write essays and are given feedback. More specifically, Latent Semantic Analysis (LSA) as a semantic technology that can be used for automatic essay scoring (AES) was applied for several purposes to facilitate essay writing in large university courses as part of an overarching strategy to improve learning and teaching at universities, that is, evidence-based teaching (EBT). In this dissertation, I will summarize and discuss findings regarding good learning and teaching (i.e., EBT) as well as why and how essay writing should be used in university courses, and regarding AES and LSA. Further, I will provide my own empirical findings on different ways to apply LSA in university courses: First, when students write essays at home, cheating must be expected, detected, and avoided. Thus, in Paper I, we analyzed whether LSA could be used to detect cheating in a large university course. Second, due to capacity constraints, instructors might need to focus their time and energy on students who are in need of special guidance. Thus, in Paper II, we investigated whether LSA could be used to identify poorly performing students. Third, before applying LSA for essay scoring, the effects of LSA-based evaluations should be explored. Thus, in Paper III, we analyzed the effects of LSA-based scores on students’ acceptance of automatic assessments and on learning-related characteristics. I will discuss these findings critically and conclude that LSA should not be used alone but is useful for assisting university instructors in different ways