13 Generative AI in Writing Assignments
Three Idaho State University graduate students and writing instructors, Suhaib H. Malkawi, Ify Ndukuba, and Mahnaz Poorshahidi, researched faculty perceptions and concerns about AI use in writing classes. They collected assignments and interviewed 16 ISU faculty from four departments to assess how “AI-proof” their writing assignments were. The researchers noted that AI was often used for low-stakes tasks such as summaries and outlines, and misuse was more prevalent in lower level courses such as ENGL 1101. Upper division and specialized classes reported little to no AI use. The researchers assigned a score of high, medium, or low vulnerablity to assessments they collected and analyzed, finding that just 0-6% were completely AI resistant (e.g., oral exams or fully process based authentic tasks).
AI-Proofing Strategies for Written Assignments
- Specificity to local/contextual material (e.g. PR letters to local media)
- Requirement for personal reflection or critique
- Scaffolding via multiple submission steps
- Use of obscure or highly detailed source material (e.g. archive sources)
- Visual/media analysis
- Sharing drafts via Google Docs (visible writing process)
Key Vulnerabilities in Assignment Design
| Type of Task | AI Vulnerability |
| Generic summaries or reflections | High |
| Open-ended research without scaffolding | High |
| Compare/contrast without localized detail | High |
| Simple theory explanation without application | Medium |
| Close reading with contextual evidence | Medium-Low |
| Oral exams and spontaneous Q&A | Low |
| Process-based, iterative assignments (Google Docs, peer review) | Low |
Recommendations for Writing Instructors
- Avoid generic prompts — Personalize or contextualize every major writing task.
- Require a visible writing process — Shared Google Docs, drafts, annotations.
- Incorporate media or local context — Tasks grounded in specific, real-world material are harder for AI to mimic.
- Teach AI literacy — Don’t just prohibit it. Train students on responsible and intelligent use.
- Design for transfer — Focus on skill-building over paper-writing, especially in lower-level courses.
- Prefer multi-step, scaffolded work — Smaller checkpoints make misuse easier to spot and learning harder to bypass.
Finally, the researchers concluded that thoughtful design reduces AI misuse and increases student engagement in courses. The trend is moving away from defensive prohibition and toward strategic adaptation.
The full presentation is available here.
