How should professors provide feedback, verbally or in writing, to enhance learning? Our research focused specifically on judgment: what is the effect of normative judgment versus informational feedback on student motivation, engagement, and performance? Based on a review of the relevant literature on the subject, we hypothesized that certain forms of normative judgments, be they positive or negative, could be ultimately harmful to students (compared to a neutral, task-focused form of feedback). Negative judgments and criticism can increase student anxiety and harm instructor immediacy, both of which have been shown to decrease engagement and performance in class (England, Brigati, Schussler, & Chen, 2019; Rocca, 2004). Positive feedback may decrease student intrinsic motivation through the undermining effect and may reinforce fixed mindsets/brilliance beliefs (Deci, Koestner, & Ryan, 2001; Dweck, 2007). Finally, judgmental comments directed at a student’s personality traits seem to be an ineffective (and potentially harmful) use of a feedback intervention; feedback should be task-based instead (Kluger & DeNisi, 1996). To test this hypothesis, our team used modern machine learning techniques to (a) transcribe 80 online seminar-style classes and (b) use sentiment analysis to gauge the tone expressed in each comment. We then used statistical techniques such as linear regression, chi-squared, and genetic matching to predict the effect of employing judgmental tone on the number of hand-raises and emojis, as well as the average student talk time in class. Our results seem to indicate that judgmental, emotionally charged language seems to actually increase overall engagement, but skew participation toward certain students. In our talk, we call into question the adequacy of the methods to address the hypotheses, but show that with refinement, these novel applications of machine learning are quite promising.
Reference:
https://docs.google.com/document/d/1-jlHz52_5pm5tKC-IT-DmvA_ARV7mRXjTdpx8GZJ9bo/edit?usp=sharing
Open Zoom
