• We designed an interactive system that utilizes narrative animations, multilevel games, and affective computing to support children with autism spectrum disorder (ASD) learning neurotypical emotions and facial expressions in social contexts. • Children with ASD achieved generalized learning gains in recognizing and producing neurotypical facial expressions. • The three-step design approach is feasible for developing interactive AI systems for children with ASD. • Narrative animations provide value social contexts, while multilevel games and interactivity help children to stay focused. • Blending AI and human support enables children's personalized, engaging, and effective learning. Emotional skills are crucial for a child's success, though children with autism spectrum disorder (ASD) face challenges in understanding social contexts, as well as recognizing and expressing facial expressions. We present EmoLand, a web-based interactive system that utilizes narrative animations, multilevel games, and artificial intelligence (AI) affective computing to instruct children with ASD about neurotypical emotions and facial expressions in social contexts. The paper sets out the iterative design process used, which was informed by ASD intervention theories, empirical studies focused on children with ASD and their educators (tutors, therapists, parents), and multiple discipline expert inputs. A preliminary evaluation involving twelve children with ASD (aged 4 to 7 years) and five tutors suggests that EmoLand is a viable teaching aid for ASD educators and effectively assists children in learning and applying generalized knowledge about emotions and facial expressions. In this paper, we discuss the design approach and key lessons learned for creating interactive AI systems for children with ASD and explore beneficial considerations for enhancing emotional development.

We examine the user experience, learning effectiveness, learning behaviors, and learning motivation of 54 young adults in pairs using an immersive virtual reality (IVR) environment and a multi-touch tabletop (MTT) to gain a cultural-historical understanding of the traditional Chinese paintings, Spring Morning in the Han Palace and The Night Revels of Han Xizai respectively. We used mixed methods of knowledge assessments, questionnaires, observation, and interviews to collect and analyze data. The results reveal the IVR environment significantly increased the learning effectiveness and motivation compared to the MTT system, particularly in tasks related to recall of details or spatiality. The IVR design incorporates five learning scenes, and as part of this aspect, the qualitative results indicate users in the IVR condition (1) developed a variety of exploratory and embodied learning strategies; (2) quickly switched their attention to the central learning content, and (3) had equitable interactions in collaborative learning. This research contributes to the potential of IVR learning effectiveness and associated design considerations to learn traditional cultural heritage.