Expediting Investigation of Examinations Malpractice Involving Suspected Jointly Written Answer Scripts

Examination malpractice, particularly involving external assistance in answering questions, poses a significant challenge for Examining Boards. Detecting cases where candidates receive help in writing their answers, resulting in jointly written answer scripts with different handwriting, is a complex and time-consuming process. Currently, this task falls to handwriting experts who employ subjective methods to examine documents, leading to delays and prolonged investigations. The study proposed the development of a web-based application to streamline the process of collecting handwriting samples on follow-ups on who might have been involved in the malpractice, reducing the need for physical travel to examination centers. Furthermore, the study sought to incorporate AI into the process of handwriting identification and verification. AI offers a more objective and efficient approach. To achieve these objectives, the study collected information through a questionnaire from members of staff at the Examinations Council of Zambia (ECZ) who are involved in handling Malpractice cases. The developed web-based application using the Django framework mitigated the need for physical travel to examination centers resulting in reduced delays. In the Machine Learning Model for writer identification, the prediction was 89% for True Positives and 10% for False Positives. In the Machine Learning Model for Binary Classification, the prediction was 52% on True Positives and 67% on True Negatives. The AI Model for writer verification (Binary classification) did not perform well, this may be because the Model was trained in the Hindi alphabet.