Getting Beyond the Flawed Metrics of Student Evaluation Surveys

Student evaluation surveys are not valid instruments for evaluating teaching, student learning, or the curriculum. For almost a century, researchers have found that not only are these tools highly correlated with course grades, but they are also correlated with grade inflation and lower levels of student learning.

Further, overreliance on student evaluation surveys can lead to fraud or extortion because instructors feel compelled to play school and inflate grades in order to keep their job and get a promotion. Researchers have found that these instruments are unfair to teachers because these they are based on the subjective emotions and beliefs of students, which are largely beyond an instructor's control. This is especially damaging to quality teachers with high standards because they often receive the lowest evaluation scores, even though they produce the most learning.

These instruments are also discriminatory. Student ratings are correlated with a vast array of prejudices based on teacher characteristics that have nothing to do with teaching, such as race, nationality, gender, and attractiveness.

Stark and Freishtat (2014) argued, "there's general agreement that student evaluations of teaching don't mean what they claim to mean" (Kamenetz 2014, para. 10). Student evaluations tell us about the biased and emotional mind of the student. They tell us very little, if anything about the instructor, teaching, student learning, or the curriculum. It can be useful to understand student perceptions, but we need to be clear about what is actually being measured and why.

Unfortunately, few are having this conversation. Stark and Freishtat (2014) conclude: "It's totally valuable to ask [students] about their experience, but it's not synonymous with good teaching". Thus, 18 scholarly associations have all stated that student surveys should not be relied upon "as a measure of teaching quality" (Supiano 2019, para. 2). And yet, these invalid tools have proliferated. How do we move beyond these flawed metrics? What are better tools?