Enhancing Feedback Literacy in Project-Based Science Education: Discourse Analytics and Technology Design

This study investigates how discourse and rating signals from self- and peer-assessment can inform the design of technology-supported scaffolds to enhance feedback literacy in project-based STEM education. The dataset comprises 9,180 assessment events written in Indonesian by 95 undergraduate engineering students, making it one of the largest non-English peer feedback corpora analyzed in this domain. Two empirical research questions guided the analysis: (1) What patterns characterize students’ written comments and rating behaviors? (2) How do these signals relate to course outcomes? A third propositional question regarding design implications was explored in the discussion. Findings show that short, descriptive comments were common in both self- and peer-assessments, limiting elaboration. Peer ratings were more strongly associated with final course outcomes than self-ratings, confirming peers as more reliable indicators of achievement. Calibration mismatches between self- and peer-assessments were systematic across rubric dimensions, revealing persistent over- and underestimation patterns. Building on these empirical patterns, we develop three lightweight and explainable design hypotheses: elaboration prompts triggered by unusually short comments, rubric-aware prompts that encourage reference to task criteria, and calibration prompts that activate when self-ratings diverge substantially from peer norms. These mechanisms are grounded in transparent, real-time indicators and are intended to be tested in future classroom implementations. By grounding learning analytics in non-English STEM classrooms, this study demonstrates how simple, explainable signals can inform the development of feedback literacy scaffolds, support metacognitive monitoring, and improve collaborative learning outcomes in science and engineering education.