Mastem: Designing A Multi-Agent System For Mathematics Tutoring.pdf

The paper introduces Mastem, an educational software designed as a Mathematics Tutoring Multi-Agent System aimed at assisting students in grades 5 and 6 with mathematics learning. The motivation for developing Mastem stems from the poor performance of elementary students in mathematics exams, as indicated by IDEB 2009 results. Intelligent Tutoring Systems (ITS) are proposed to address these challenges due to their availability and resourcefulness.

Mastem is developed using agent-oriented paradigms, which provide autonomy, flexibility, collaboration, and adaptation—key features for effective ITS. The system adapts to students' difficulties by identifying errors and offering tips or examples. It covers topics such as divisors, prime numbers, factoring, and fractions, identified as challenging areas for Brazilian elementary students.

The development of Mastem utilized the Tropos methodology and JADE framework. Tropos is a multi-agent system development methodology that emphasizes adaptability to organizational needs through five phases: early requirements, late requirements, architectural design, detailed design, and implementation. The paper details these phases, including modeling strategic dependencies and rationale using i* framework elements.

Mastem's architecture involves various actors (Student, Teacher) and agents (Tutor, Ranking, Difficulty), each with specific roles in the learning process. For instance, the Tutor agent monitors student performance and provides feedback, while the Ranking agent maintains a leaderboard of top-performing students.

The system was implemented using Java and JADE for agent behaviors, with database management handled by PostgreSQL. The user interface includes features like question insertion, performance monitoring, difficulty management, and ranking display to enhance usability and engagement in learning mathematics.