Students and instructors rightly worry about the impact of cheating in unproctored online exams. There are online proctoring systems, but they are limited in effectiveness. Webcam proctoring can't see everything in the room, only what the webcam is pointing at. There might be another computer screen behind the webcam. A student might be looking down at a crib sheet or another device, and the only clue would be eye movement, which is not nearly reliable enough to justify an accusation. Further, it may be considered an invasion of privacy for a recording to be made of a room in the student's home, often a bedroom. There are several approaches to detecting cheating by using software. One approach is statistical tests. Exams can be compared pairwise in order to find suspicious similarities. Software applications have been created to detect cheating on multiple-choice tests. These include Integrity (http://integrity.castlerockresearch.com/) and SCheck (http://wesolows.wired4wisdom.com/). Essay answers can be checked for plagiarism using an LMS plugin like Unicheck (https://moodle.org/plugins/plagiarism_unicheck) or by uploading a zip file of student answers to Turnitin. While that is not as convenient as using a turnkey application, practices have been developed to streamline the process. There are also several software packages for detecting plagiarism, like CopyDetect, CopyLeaks, and textreuse, that could be incorporated into an LMS by instructional technology support staff. Gradescope (a unit of Turnitin) is developing a rubric-based approach to detecting cheating. With Gradescope, the instructor administers the exam online, or scans in exam papers. As the instructors grade the test papers, they develop a rubric that awards a certain number of points for particular features of a correct answer or, alternatively, deducts a certain number of points for each mistake. Gradescope then compares the scoring of each exam paper and detects pairs of students who lost (or gained) points for the same reasons on many of the same questions on the exam. The focus of this presentation will be on tools that are currently usable, either as standalone applications, or as packages that can be incorporated into an LMS. A secondary feature will be a short description of the algorithms that they employ.
No takes yet. Share an insight, caveat, or question.
Gehringer et al. (2024) studied this question.
Synapse has enriched one closely related paper. Consider it for comparative context: