Abstract:

Lockdown and social distancing in the COVID-19 era led to closures of schools and universities everywhere. Several problems must be faced, both pedagogical and technical; for instance, many students at home suffer from loneliness and distress and become unable to engage productively.
This experience is leading to a new school, based on blended learning and advanced computer-based technologies.

Although still heterogeneous, online learning tools must be used for bringing back cooperation, human contacts and enthusiasm among students. To this aim, we suggest remote groupwork should be adopted and managed, as already claimed by teachers, researchers as well as the European Union.

Typical benefits of groupwork can also be greatly enhanced by web-based cooperation: increased collaborative skills, comparison with other companions and consequent increase in level of self-criticism are just some examples.

The specific problem faced here is that even “traditional” groupwork often selects members without control and it is hard to judge advances and difficulties of individuals. In remote groupwork, this can become even more challenging, so we suggest it should be partly solved by advanced, deep-learning-based adaptive testing techniques.

Adaptive testing techniques have already been proposed for groupwork management. In particular, the authors used large amounts of students’ profiles to compose groups on the basis of matching interests and skills.

The novelty here investigated is how to improve profiles’ classification and analysis through Artificial Intelligence techniques such as Neural Networks and Deep Neural Networks.

Neural networks try to imitate the way the human brain develops classification rules. A Deep Neural Network is a neural network with several layers, such as Convolutional and Deep Belief ones and speed up data analysis and classification.

In this context, neural networks and deep networks are applied to analyse the students’ profiles and classify them according to their interests, skills and will to cooperate. Especially in these times of great isolation, a good and satisfying workgroup experience can turn out to be highly motivating and help to face distress and study together productively even though partially or totally remotely.
In the proposed approach, the students’ profiles are used to form suitable groups, which are to be refined during the whole learning process. In particular, adaptive testing is adopted for the periodic control of both group activities and achievements or possible difficulties of single members. This kind of approach is widely used for the management of single students and was previously proposed in the broader context of groupwork with a semi-handcrafted technique.
The proposed architecture is composed of several layers:
(1) profiles’ acquisition;
(2) profiles’ classification through neural networks and deep neural networks;
(3) a first definition of group memberships follows;
(4) training;
(5) adaptive testing and consequent profiles’ updates;
(6) iteration of phase (2) and, if needed, revision of phase (3) and consequent groups’ adaptation.

This kind of computer-based technique can be applied in remote learning as well as in traditional classes and support blended learning as well. The very use of AI techniques can and ought to become a bridge leading from traditional school to the future one, where teachers are supported in the creation and management of blended learning ecosystems.

Keywords:

COVID-19, e/m-learning, groupwork, neural networks, deep networks, groups management, blended learning.