Though toddlers and preschoolers are regular touchscreen users, relatively little is known about how they learn to perform unfamiliar gestures. In this paper we assess the responses of 34 children, aged 2 to 5, to the most common in-app prompting techniques for eliciting specific gestures. By reviewing 100 touchscreen apps for preschoolers, we determined the types of prompts that children are likely to encounter. We then evaluated their relative effectiveness in teaching children to perform simple gestures. We found that children under 3 were only able to interpret instructions when they came from an adult model, but that children made rapid gains between age 3 and 3-and-a-half, at which point they were able to follow in-app audio instructions and on-screen demonstrations. The common technique of using visual state changes to prompt gestures was ineffective across this age range. Given that prior work in this space has primarily focused on children's fine motor control, our findings point to a need for increased attention to the design of prompts that accommodate children's cognitive development as well.

In this study, we work towards a strategy to measure and enhance the quality of interactions in discussion forums at scale. We present a machine learning (ML) model which identifies the phase of cognitive presence exhibited by a student’s post and suggest future applications of such a model to help online students develop higher-order thinking. We collect discussion forum transcript data from two online courses: CS1301 (an introductory computer programming MOOC) offered by edX and CS6601 (a graduate course on artificial intelligence) which uses the Piazza online discussion tool. We manually code a random sample of students’ posts based on the Community of Inquiry coding scheme and explore trends in cognitive presence within and across the courses. We further use this coded data to analyze the relationship between students’ observed cognitive presence and course grades. In terms of testing and building an ML model, we use a Bidirectional Encoder Representations from Transformers model that uses a deep learning technique to train large text corpus and fine-tune the language model. Our results suggest that deeper cognitive engagement with course concepts, as expressed by higher cognitive presence, are associated with better learning outcomes for students in both course settings. Our ML approach achieves 92.5% accuracy on the classification task, motivating the use of ML for instructional interventions in online courses. We expect that our research study will not only contribute to extending the literature on cognitive presence but also have a beneficial impact on online instructors or curriculum developers in higher education.

No abstract

No abstract

No abstract