Download files
Abstract
In the higher education sector, student evaluations of teaching heavily influence the hiring and promotion decision of lecturers. However many studies have found there to be a discrepancy between the ratings that male and female lecturers receive, along with those from minority cultural backgrounds. With the discrepancy unexplained by common factors included in the models, these differences have been attributed to an implicit bias which students may hold towards their lecturers.
Using a large set of evaluations completed over several years at a leading Australian university, we use Bayesian statistical methods combined with natural language processing techniques to rigorously investigate the following questions - Is there evidence of implicit bias (gender or cultural) within student evaluation of teaching surveys? What can the students' comments inform us about how the discrepancy in ratings may arise? Can we mitigate these biases to reduce any unintended effects?
The studies in this dissertation show that the gender and cultural characteristics of a lecturer do influence how the students rate and comment on their lecturers, and these characteristics also influence how students respond to an intervention message giving us an insight into how these implicit biases may arise. A clear implication from this research is the need to ensure these surveys are effectively and fairly rating lecturers, and administrators need to account for these factors when using evaluations to assess a lecturer’s performance.
