*WINNER* Towards Automated Machine Learning Detection of Academic Dishonesty in Computer-Based Testing


  • Parth Patel
  • J.W. Bruce


The global pandemic forced a rapid shift of learning environments to an online and remote modality. In a computer-based testing environment, ensuring academic integrity is difficult without resorting to technology that is intrusive to student privacy. During online examinations, students have many opportunities and resources to act dishonestly. However, the computer-based testing environment also provides data about the examination process that can be collected which is not possible in face-to-face settings. Utilizing features extracted from computer-based testing logs, a machine learning model was implemented to determine if students completed the exam honestly or not. The model was validated with high recall or low false negatives. Implementing this model into online examinations would allow for suspected academic dishonesty to be automatically flagged for further review by the instructor. Results indicate it is possible for the model to detect dishonest conduct, including unauthorized collaboration and utilization of external resources such as exam assistance services.






Engineering-Electrical and Computer