Prediction-Based Student Evaluations of Teaching as an Alternative to Traditional Opinion-Based Evaluations

Abstract

The validity of traditional opinion-based student evaluations of teaching (SETs) may be compromised by inattentive responses and low response rates due to evaluation fatigue, and/or by personal response bias. To reduce the impact of evaluation fatigue and personal response bias on SETs, this study explores peer prediction-based SETs as an alternative to opinion-based SETs in a multicultural environment. The results suggest that statistically significant fewer respondents are needed to reach stable average outcomes when peer prediction-based SETs are used than when opinion-based SETs are used. This implies that peer prediction-based SETs could reduce evaluation fatigue, as not all students would need to do each evaluation. The results also report that the peer prediction-based method significantly reduces the bias evident in the opinion-based method, in respect of gender and prior academic performance. However, in respect of the cultural variables, race and home language, bias was identified in the peer prediction-based method, where none was evident in the opinion-based method. These observations, interpreted through the psychology literature on the formulation of perceptions of others, imply that although peer prediction-based SETs may in some instances reduce some personal response bias, it may introduce the perceived bias of others.