Faculty Guide to Preventing Unauthorized Use of GenAI

Strategy 

Strategy Details 

Require Students to Keep Documentation of Their Work  

Require students to use platforms that can track their activity such as enabling track changes on documents (Birks & Clare, 2023). Students may need to provide working drafts of their assessments before final drafts (Birks & Clare, 2023). Making it mandatory for students to enable Track Changes on submitted assignments provides a deterrent to students’ use of AI tools (Sunderland-Smith & Dawson, 2022, p. 164). 

Use Scaffolding to Understand Student Baselines 

In earlier stages, gauging student knowledge, providing early feedback, and providing exemplars are good practice in scaffolding (Stoez et al, 2022, p. 110). Instructors directly assess the process of learning and can gain a baseline of student knowledge and writing style, which also enables them to detect when students outsource their work to another source (Stoez et al., 2022, p.111). 

Flag and Investigate Suspicious Changes in Grades 

Proctored, in-person, restricted information-access assessments are more resistant to cheating in general (Birks & Clare 2023, p. 12). Quizzes or non-time-limited exams done at home are riskier for cheating (Birks & Clare 2023, p.12). If students score substantially better in unsupervised assignments in comparison to supervised assignments, it could be a sign of the unauthorized use of AI (Birks & Clare 2023, p. 15).  

Have Two-Stage Assessments 

Integrating interviews with students as a final component of assessments (Sunderland-Smith & Dawson., 2022, p. 165) has a deterrent effect. If students have outsourced their work to AI, they may have trouble answering questions about their research strategy, methods, challenges, and the content of their paper. 

 Check References for Quality 

References that are consistently tangentially related to the assignment, without being closely tailored to the subject at hand is a sign of outsourcing (Crockett, 2022, p. 174). Lancaster (2023) notes that references in AI-generated papers may be nonexistent or have broken links.