Collaborative Creation and Training of Social Bots in Learning Communities

This is the authors' abstract. We don't add key points for this paper.

The interaction between instructors and students is one of the key concepts to improve the student's learning process. To personalize learning on a massive scale, social bots can be used as supporting technology. However, their development for virtual learning environments currently requires deep technical knowledge. This leaves learner communities relying on highly-skilled developers to generate and tailor these social bots. Participatory design, end-user development and model-driven principles bear the potential to close this technical gap. In this paper, we propose a model-driven approach for creating social bots. Using our framework, learners can create, train and utilize these for self-hosted virtual learning environments relying on OpenAPI specifications offered, e.g. by Blackboard. We support both retrieval-based bots that react to certain events in predefined ways, as well as generative bots by utilizing open source deep learning technologies. Our first evaluation shows the usefulness of model-driven generation and utilization of social bots. We see the potential of this approach to move the development closer to the actual learner.