AI-Powered Students' Collaboration and Evaluator using LDA

The difficulties in teaching, evaluating, and grading soft skills like communication and teamwork make it difficult to incorporate them into programming courses. A pedagogical framework and a customized evaluation rubric created for an after-school Java programming course serve as the foundation for organized, AI-driven method of assessing and grading teamwork abilities. In order to determine their advantages and disadvantages, the study looks at how well various AI methods evaluate teamwork in group tasks. Latent Dirichlet Allocation (LDA) was used in place of conventional rule-based methods to uncover hidden collaboration roles and examine students' contributions within different project roles. To determine the best practices in terms of precision, interpretability, and viability, machine learning (ML) models were also used. While comprehensive evaluation still requires teachers to address topics beyond the rubric, like code quality, plagiarism, and project completion, the results, which are based on data from four academic years, show that both LDA and ML models offer trustworthy grading recommendations. The LDA method is effective since it requires little initial setup and does not require established rule structures, whereas ML models provide excellent accuracy with less manual labor, even if they require historical data for training. Both approaches use topic-based distributions to provide teachers with understandable, comprehensible insights which helps teachers to highlighting the students room for improvement.