Dynamic peer-to-peer amplifier system used in agent-based intelligent tutoring system

Peer-to-peer tutoring in high schools are increasingly used to model a dynamic learning environment where students can effectively teach and learn by improving self-regulatory skills. We propose a dynamic peer-to-peer amplifier system in the form of a mathematical framework for a recommender system that addresses the problem of formulating response phrases for the tutoring considering the underlying sequence of possible phrases. Ultimately, using the recommended phrases, the peer tutor becomes more professional in teaching math problems to peer students. The main advantage of the proposed model is that it is adaptive to the peer student’s concentration level and accordingly recommends phrases to the peer tutor while considering the sequence of communication took place in the session combined with the reflection on peer student’s concentration level. Improving the quality of education via peer-to-peer training has been a challenge for a while and in this paper, we attempt to address the problem by using AI techniques to generate phrases in the form of recommendations to the students who tutor peers having difficulty with similar problems. We propose a recommender sequence mapping system to effectively recommend new phrases that are meaningful and goal-oriented given the past sequence of communication. Evaluation of the approach on the Receiver Operator Characteristics (ROC) curves shows that the proposed recommender system is highly robust and accurately learns parameters in various settings. Given that there is no system to be used as benchmark, we managed to compare the system with and without deployment of the Al-empowered recommender framework and discuss the system performance in various aspects.