23rd International Conference on Artificial Intelligence in Education (AIED), Durham, Canada, 27 - 31 July 2022, vol.13355, pp.218-230
Nowadays, the use of distance learning is increasing, especially with the recent Covid-19 pandemic. To improve e-learning and maximise its effectiveness, artificial intelligence (AI) is used to analyse learning data stored in central repositories (e.g. in cloud). However, this approach provides time-lagged feedback and can lead to a violation of user privacy. To overcome these challenges, a new distributed computing paradigm is emerging, known as Edge Computing (EC), which brings computing and data storage closer to where they are required. Combined with AI capabilities, it can reshape the online education by providing real-time assessments of learners to improve their performance while preserving their privacy. Such approach is leading to the convergence of EC and AI and promoting AI at the Edge. However, the main challenge is to maintain the quality of data analysis on devices with limited memory capacity, while preserving user data locally. In this paper, we propose an Edge-AI based approach for distance education that provides a generic operating architecture for an AI unit at the edge and a federated machine learning model to predict at real-time student failure. A real-world scenario of K-12 learners adopting 100% online education is presented to support the proposed approach.