Computer programming is an essential skill for the 21st-century workforce. However, mastering programming is challenging, especially for beginners. Recently, digital tools have become popular in supporting programming education. However, creating engaging programming materials remains difficult due to the many factors that need consideration. We introduce Codigen, an interactive tool powered by large language models (LLMs) that integrates visual authoring and prompt engineering to support programming education with various learning strategies. A user study with 29 participants showed that Codigen can assist teachers in generating quality programming materials. Statistical analysis showed that Codigen supports learners’ confidence and learning outcomes in programming. The perceived utility and engagement with the tool were also statistically significant. We discuss our findings to guide the design and use of LLM-powered authoring tools for educational purposes.

Group work involves a myriad of complex processes encompassing social, perceptual, cognitive, and contextual factors. However, there is a lack of empirical research on computer-supported group work processes and their impact on outcomes at different stages of group work, especially when creativity and quality of outcomes are significant. Group work processes can interfere and hinder productivity, which we refer to as the “group folding effect.” We designed a three-stage process structuring to enhance group work productivity. In a field study, we examined how process structuring shapes productivity in two sub-studies: design and peer feedback, each with 40 participants (N = 40). The results revealed that process structuring significantly improved both the quantity and quality of productivity. Additionally, process structuring appeared to reduce inhibitory effects of group work, such as negative priming, fixation on familiar ideas, and social comparison. We discuss the implications of this research in supporting productive group work processes in collaborative tools and insights into a pattern of the group folding effect.

International audience

International audience

International audience

Learners' emotional state has proven to be a key factor for successful learning. Visualizing learners' emotions during synchronous on-line learning activities can help tutors in creating and maintaining socio-affective relationships with their learners. However, few dashboards offer emotional information on the learning activity. The current study focuses on synchronous interactions via a videoconferencing tool dedicated to foreign language training. We collected data on learners' emotions in real conditions during ten sessions (five sessions for two learners). We propose to adopt and combine different models of emotions (discrete and dimensional) and to use heterogeneous APIs for measuring learners' emotions from different data sources (audio, video, self-reporting and interaction traces). Based on a thorough data analysis, we propose an approach to combine different cues to infer information on learners' emotional states. We finally present the EMODA dashboard, an affective multimodal and contextual visual analytics dashboard, which allows the tutor to monitor learners' emotions and better understand their evolution during the synchronous learning activity.