Xiangen Hu

Humor can function as a powerful social-affective cue that shapes learners’ emotional and motivational experiences in digital learning environments. This study investigated how humorous pedagogical agents influence intrinsic motivation, positive emotions, and perceived agent value across two between-subjects 2×2 experiments. Experiment 1 used a Humor × Watching dialogue design, and Experiment 2 used a Humor frequency × Mental load design. University students were randomly assigned to conditions in each experiment. Data were analyzed using SPSS factorial ANOVA, followed by simple effect analyses for significant interactions. In Experiment 1, agent humor enhanced intrinsic motivation, positive emotions, and perceived agent value, while watching dialogue suppressed positive emotions without hindering knowledge transfer. In Experiment 2, higher humor frequency increased positive emotions and Human-likeness in perceived agent value, and interacted with mental load to buffer its negative impact on intrinsic motivation. These findings clarify the social-affective mechanisms through which humor and humor frequency support learners’ motivational and emotional engagement in cognitively demanding online contexts.

Large language models (LLMs) like GPT-4 show potential for scaling motivational interviewing (MI) in addiction care, but require systematic evaluation of therapeutic capabilities. We present a computational framework assessing user-perceived quality (UPQ) through expected and unexpected MI behaviors. Analyzing human therapist and GPT-4 MI sessions via human-AI collaboration, we developed predictive models integrating deep learning and explainable AI to identify 17 MI-consistent (MICO) and MI-inconsistent (MIIN) behavioral metrics. A customized chain-of-thought prompt improved GPT-4's MI performance, reducing inappropriate advice while enhancing reflections and empathy. Although GPT-4 remained marginally inferior to therapists overall, it demonstrated superior advice management capabilities. The model achieved measurable quality improvements through prompt engineering, yet showed limitations in addressing complex emotional nuances. This framework establishes a pathway for optimizing LLM-based therapeutic tools through targeted behavioral metric analysis and human-AI co-evaluation. Findings highlight both the scalability potential and current constraints of LLMs in clinical communication applications.

Although there are studies on digital competence in higher education, there are very few studies on digital safety competence. This study aims to explore the impact of digital safety competence on students’ higher-order thinking and AI-related outcomes. Using a cross-sectional design, 159 university students completed an online questionnaire to measure cognitive competence, Al self-efficacy, AI ethics, and moral competence. Results showed that digital safety competence was positively and significantly related to cognitive competence, AI self-efficacy, AI ethics, and moral competence (p < 0.05). Our study extends the literature by highlighting the role of digital safety competence. Educators and university policy makers may consider incorporating digital competence, especially in the area of safety, into their teaching and learning strategies.

Generative Intelligent Tutoring Systems (ITSs), powered by advanced language models like GPT-4, represent a transformative approach to personalized education through real-time adaptability, dynamic content generation, and interactive learning. This study presents a modular framework for designing and evaluating such systems, leveraging GPT-4’s capabilities to enable Socratic-style interactions and personalized feedback. A pilot implementation, the Socratic Playground for Learning (SPL), was tested with 30 undergraduate students, focusing on foundational English skills. The results showed significant improvements in vocabulary, grammar, and sentence construction, alongside high levels of engagement, adaptivity, and satisfaction. The framework employs lightweight JSON structures to ensure scalability and versatility across diverse educational contexts. Despite its promise, challenges such as computational demands and content validation highlight the main areas for future refinement. This research establishes a foundational approach for advancing Generative ITSs, offering key insights into personalized learning and the broader potential of Generative AI in education.