An AI Framework for Academic Integrity and Learning Honesty (Prism Eye )

Abstract

Transparent assessment and academic integrity are the key elements of education in modern society. The monitoring processes used in traditional classrooms are very manual and thus, time consuming, inconsistent and subject to human error. As Artifcial Intelligence (AI) has developed at a rapid pace, it has become feasible to automate these procedures and implement fair, dependable and timely systems to aid teaching and examining settings. This thesis introduces the Prism Eye as a multimodal AI system that aims at increasing academic integrity by offering automated attendance, assessment of participation, and detection of cheating. The system is a combination of face recognition, vision-based participation monitoring, and YOLOv8 (You Only Look Once version 8)-based prohibited item and behavior detection to detect suspicious student activities. The modalities collaborate in order to create an integrated profle of the student status that is presented on a single dashboard to instructors. The given scheme is expected to minimize the cost associated with manual invigilation, assist in objective judgment, and enhance the clarity of academic sessions. Prism Eye shows the way AI-based surveillance systems can help to create a more effective, secure, and honest educational setting.