Performances of the students in learning a programming course is not same, since learning to program is greatly influenced by two dominating factors namely self-efficacy and mental efforts. Prior research efforts have shown that high self-efficacy can have an increased effect of being a trained programmer, especially in an intelligent agent based pair programming system. The main objective of this work is to increase the self-efficacy of the students by providing prior-learning experiences. This experience is facilitated by recommendation agents that provide suitable E-Learning programming course contents based on identifying their individual learning styles which can be used as a factor of prior self-learning computing experience. This helps in increasing the programming abilities when learning in an agent-based pair programming environment subsequently. Moreover, the proposed system analyzes the educational effects of the students learning using pair programming agents based on increased self-efficacy.