Leveraging AI to Ensure Authenticity in Student Assignments: A Knowledge-Based Validation and Evaluation Framework

This research explores the development of an AI-driven assessment system designed to evaluate student knowledge while minimizing cheating in educational settings. The system operates in two distinct phases. In Phase 1, AI-generated, contextually relevant questions based on student-submitted assignments assess whether the student has genuinely engaged with the content or used external tools, such as generative AI, without proper understanding. In Phase 2, the system delves deeper into conceptual understanding by aligning submissions with course-specific materials, evaluating relevance, depth, and originality. A fine-tuned large language model (LLM), trained on domain-specific data, enables precise evaluation against academic standards. Results indicate that students demonstrated marked improvements in understanding and performance, validating the effectiveness of the two-phase model. This model encouraged deeper engagement and comprehension, increasing students’ knowledge of the material. By addressing the shortcomings of traditional assessment methods, this research introduces a model that reduces the likelihood of students blindly submitting assignments without truly understanding the content.