Abstract:

Artificial Intelligence is permeating more and more everyday’s life and applications; its potentialities are enormous, in terms of processing and analysis of huge quantities of information (Big Data), as well as resource management automation.

In this paper, a particular Machine Learning technique, namely Deep Reinforcement Learning, is first described and then applied to the automated management of remote mobile cooperative e-learning and e-groupwork activities.

The importance of mobile e-learning modalities in modern society has been widely underlined. The impact on education and training systems is becoming so strong that mobile e-learning is considered the main tool for achieving European policies on schooling and higher education, as well as social inclusion, language learning as well as intercultural dialogue among others.

Some of the most important advantages of e-learning and m-learning especially are related to the independence of end-users in time and space; in this context, the web-based training allows students to define and tailor their learning paths depending on their specific needs, augment interaction with studying materials and investigate possible further issues. In this perspective, two observations must be made: first, each learner becomes the focus and centre of the studying experience; second, learning becomes a continuous and adaptive process.

In context of mobile e-learning technologies and systems, special attention should be paid to remote groupwork activities. As a matter of fact, these potentialities are greatly enhanced by the ever growing multimedia technologies and the Internet, so as to improve the quality of learning by facilitating access to resources and services as well as remote exchanges and collaboration.

On the one hand, cooperative frameworks can highly benefit from the most recent advances in learning and communication technologies. On the other hand, though, even in “traditional” groupwork activities members are often selected “without control” and it is difficult to judge personal advances and difficulties of each student. In remote groupwork, this problem can become even more challenging.

To cope with these difficulties, the typical benefits of groupwork activities can be greatly enhanced by cooperation guided by Artificial Intelligence techniques: improved and better developed collaborative skills, the constant comparison with other companions’ viewpoints and approaches, and a consequent increase in level of reasoning and self-criticism are just some outstanding examples. In particular, Deep Reinforcement Learning techniques are first described and then adopted, which allow the user (agent) to interact with the environment according to his or her goals.

This paper presents an integrated architecture for the automated selection and management of students in context of groupwork activities by means of Deep Reinforcement Learning techniques. The target is to learn from data and automatically individuate profiles of people having suitable skills, although possible heterogeneous backgrounds, so to optimize groupwork itself as well as personal achievements.

In the proposed approach, the students’ profiles are used to form suitable groups, which are to be refined during the whole learning process.

Machine learning and Deep Reinforcement Learning basics are first described; the considered environment and proposed architecture follow.

Keywords:

Mobile e-learning, mobile e-groupwork, Artificial Intelligence, Deep Reinforcement Learning, automated resource management.