AI support is increasingly embedded in online quizzes, yet instructors often lack clear, actionable signals about where students struggle during those assessments. We present a classroom-deployed quiz system that combines integrity-preserving hinting (TA-AI), trace summarization (Analytics-AI), and feedback-driven refinement (AI-Improver) to generate instructor diagnostics from routine interaction logs. The system was used in a graduate assembly programming course over five quiz weeks (N = 18). We report deployment evidence focused on reliability and instructional usefulness for monitoring: promptintent coding reached substantial agreement (Cohen’s kappa [κ] = 0.81); fixed-effects models (with student and item controls) showed a negative association for one-hint interactions (odds ratio [OR] = 0.231, indicating approximately 77% lower odds of a correct response for single-hint interactions relative to 0-hint interactions); and item-level demand spikes were operationalized via a demand × success prioritization process for weekly review. Rather than producing automated judgments or claims of causal learning gains, the analytics are designed as practical prioritization cues that direct instructor attention toward high-need items during AI-assisted quizzes.
M. P. Lin, Daniel H. Chang, V. Janarthanan et al.· International Journal of Eme...· 0 citations
Standard generative AI interfaces allow L2 learners to bypass cognitively demanding writing processes by generating complete texts with minimal effort. This study examines whether embedding pedagogical constraints within AI interface design can reshape learner engagement with the revision process in L2 writing. We developed Writing Helper, a multi-agent system that structures iterative revision through four specialized feedback agents (Grammar, Style, Structure, Content) and gamified mastery thresholds. In a quasi-experimental study, 45 Korean university EFL learners completed a creative writing task using either standard AI chatbots (Control, n = 18) or the agentic AI system (Experimental, n = 27). Behavioral analysis revealed that control participants followed largely linear workflows (M = 2.1 iterations), while experimental participants engaged in significantly more revision cycles (M = 6 iterations), producing longer texts (M = 150 vs. 48 words) with higher ratings for narrative elaboration. Qualitative findings indicated that learners valued the structured revision process and expressed willingness to reuse the system, yet reported that higher-order feedback lacked specificity, and inconsistent scoring raised concerns about trust. These findings reveal both the promise and challenges of using constraint-based, agentic AI design in L2 writing classrooms, highlighting the importance of pedagogical interface design.
Yujong Park, Jeeho Ryoo· Language Learning & Tech...· 0 citations
Examining undergraduate students’ AI disclosure practices suggests that institutional AI disclosure frameworks may benefit from addressing both policy compliance and the affective and disciplinary dimensions of students’ SRL decision-making.
Daniel H. Chang, M. Lin, Jing Huang et al.· Frontiers in Education· 0 citations
This work presents Oneiros, a dynamic remapping engine for multi-tenant LLM serving that dynamically repurposes GPU memory allocated for model parameters as KV cache capacity, enabling nonblocking, unidirectional parameter transfer.