AI-Driven System for Automated Generation and Evaluation of Multiple-Choice Questions

Abstract

Student assessment is a fundamental aspect of the learning process in higher education, enabling faculty members to measure learning outcomes and obtain feedback on their teaching. Multiple-choice questions MCQs are widely used in higher education assessments due to their low cost, ease of preparation, accuracy, and ability to measure various learning outcomes. However, the effectiveness of this tool in measuring learning outcomes is directly related to the quality of the MCQs. Student responses are used to calculate both the difficulty index (P) and the discrimination index (D), which helps in evaluating and continuously improving the quality of MCQs. However, the process of creating and writing high-quality MCQs and collecting and analyzing student results is time-consuming and labor-intensive, often resulting in low-quality questions or overlooking feedback calculations. This paper presents BankGen, an AI-powered framework that automates the creation, management, and evaluation of MCQs using generative AI. BankGen is developed using the Replit platform, it integrates Google’s Gemini engine with a user-friendly interface to generate curriculum-aligned and learning-outcome-related MCQs. BankGen supports the creation of question banks, creation of test, and analysis using difficulty and discrimination indicators. It also allows users to delete or modify low-quality MCQs. The results showed that the BankGen-generated MCQs are curriculum-related, aligned with learning outcomes, and free from spelling and grammatical errors. These results indicate that BankGen can reduce the effort required to evaluate and improve MCQs.