Students’ Affective Processing of Computer-Based Assessments With Feedback

• In Event: Beyond Self-Report Measures: Unlocking Insights Into Cognitive and Affective Feedback Processes With Multimodal Data Sources

Abstract

Objective
This study aims to enhance the understanding of student emotions in computer-based assessments with automated feedback. By integrating self-reports of emotions with automated video-based emotion detection, we seek to unravel the intricate relationship between achievement, feedback, and emotions.

Theoretical Framework
Automated feedback can be a powerful tool to foster student learning (Mertens et al., 2022). However, its affective benefit seems delicate because positive feedback, which mirrors success, may improve emotions, while negative feedback reflecting failure has been discussed to have a potentially detrimental affective impact (Pekrun, 2006). Previous experiments have highlighted the differential affective effects of positive as compared to negative feedback (Kuklick & Lindner, 2023), but these self-report-based insights are limited to students' conscious perceptions of their mental state (Pekrun, 2020). Therefore, automated emotion detection measures are essential to complement our understanding of emotions in digital feedback settings.

Methods and Data Source
This preregistered lab study involved 84 undergraduates completing a geometry assessment. After each task (12x), students first rated their response certitude using a five-point Likert scale. They then received immediate, elaborated feedback (correct response and solution path) and subsequently rated their emotions on a five-point Likert scale (positive emotions: three items, α=.90; negative emotions: four items, α=.89). We further utilized iMotions™ software to automatically detect positive and negative emotions using video recordings that were coded according to a student’s facial muscular activity. We hypothesized that students would experience more positive emotions after correct responses and that the presence of positive emotional expressions and absence of negative emotional expressions during task processing would be associated with higher response certitude and solution success. Due to space limitations, only a selection of the hypotheses/results are presented here.

Results and Discussion
Mixed-effects analyses revealed that students reported higher levels of positive emotions and lower levels of negative emotions following correct responses (ps<.001). Interestingly, the automated video-based coding of emotional presence during feedback processing showed that the likelihood of students exhibiting facial expressions of both positive and negative emotions increased for incorrect responses (p≤.026). This suggests that students may be more emotionally engaged when processing negative feedback, partially corroborating previous human-coded video data indicating that students express more negative affect following incorrect trials (Merrick & Fyfe, 2023). Furthermore, data indicate that positive facial expressions during task processing were associated with higher response certitude (p=.027) but not with higher solution success (p=.097). Conversely, negative emotional expressions were linked to a lower probability of solution success (p=.027) but were not associated with lower response certitude (p=.450). These findings imply that positive emotions may be more closely aligned with students' subjective perceptions of their achievement, whereas negative emotions may be tied to their actual performance.

Significance
This study demonstrates that facial emotional expressions can reflect certain findings from self-reports but that such automated emotion detection also offers novel insights into students' affective processing of feedback and assessment tasks. For educators and assessment designers, our findings underscore the importance of carefully designing negative feedback, considering that such feedback may be processed most emotionally by students.

Authors

• Livia Kuklick, Humboldt University - Berlin

• Thérése Eder, Leibniz-Institut für Wissensmedien (IWM)

• Fangzheng Zhao, University of California - Santa Barbara

• Richard E. Mayer, University of California - Santa Barbara

• Marlit Annalena Lindner, IPN - Leibniz Institute for Science and Mathematics Education