As artificial intelligence tools become increasingly sophisticated, many instructors may have concerns about the challenges this presents to maintaining academic integrity in their online assessments. Below are two concerns related to Canvas quizzes, followed by a few strategies you can use to help protect your assessment’s validity.
Artificial Intelligence Canvas Cheating Tools
Students Leaving Canvas During Quizzes
Recently, Cornell faculty have received reports of students using artificial intelligence browser extensions designed specifically for cheating on Canvas quizzes. Students can purchase subscriptions to AI tools that work invisibly within their browsers. These tools can instantly analyze quiz questions and provide answers.
Because all activity occurs within the browser, faculty will not find evidence of academic dishonesty in the Canvas logs. Unfortunately, this form of cheating is nearly impossible to detect through Canvas monitoring.
Possible Solutions
Return to Paper-Based Testing
When in doubt, analog is always an option – paper exams in supervised settings remain the most secure assessment method. Thanks to learning technology tools like Gradescope, there’s no need for faculty and instructors to grade all exams by hand.
Gradescope offers easy-to-use paper scanning and grading tools that combine traditional security with efficient digital grading. Contact us anytime for help getting started using Gradescope for your course.
If Using Canvas Quizzes, Strengthen Your Defenses:
Create AI-Resistant Questions
Be aware that modern AI tools can adequately answer many standard academic questions. AI can answer most factual and many analytical questions with surprising accuracy. For a sense of how this works – and how well this works – we encourage faculty to test their existing questions in ChatGPT or similar AI tools to understand what responses these tools are capable of generating.
To mitigate the possibility of students using AI to respond to questions, consider creating questions that draw on unique course-specific examples or more complex discussions specific to your course or discipline that are not as readily reproduced in online sources.
Require Visible Work
Consider how you might incorporate process-based questions into your assessments. How might you ask students to show their work in a way that meets your learning outcomes? Once you’ve determined an appropriate approach for your students to demonstrate their learning, you can create “File Upload” questions in Canvas that require students to upload a document showing their process or have students submit photos of handwritten work.
When administering a quiz in Canvas, faculty often require students to remain in the Canvas LMS to ensure that the answers are coming from them, and not internet searches or generative AI.
Possible Solution: Use Canvas Quiz Audit Logs
Canvas automatically records student’s quiz activity and auto-saves their progress. When students navigate away from quiz screens, that activity is recorded in the audit log. These logs appear in the quiz attempt history with timestamps showing when a student clicked away from the page. Instructors can review these logs to identify potential instances of accessing unauthorized resources (including services like ChatGPT, and other AI tools).
Be aware that Canvas Quiz audit logs have significant limitations. Log data may capture clicks that are triggered by innocent actions such as notifications, updates, or connectivity issues. Students may accidentally click away from the quiz window, and technical glitches can generate false exit records.
Because of this, CTI advises against using Canvas Quiz audit logs as standalone evidence of cheating. Rather than direct evidence of cheating, audit logs might better serve as a warning of a potential problem that warrants further investigation, including conversation with the student.
