AI Detection in Higher Education
About AI Detection
AI detection tools are notoriously unreliable and inconsistent, and you can never
be certain that the results are
accurate enough to serve as evidence of Artificial Intelligence use.
Many products claim 90% certainty of AI detection, but in practice demonstrate less than 60% accuracy because the engines designed to detect AI are constantly evolving. When considering the use of an AI detector, it is important to take the following into account:
- AI detection limitations and shortcomings
- Ethical use of often unreliable results
- Use of AI as a translation device for ESL students
- Impact on teaching and learning as a human-centered experience
AI Strategies and Alternatives to AI Detection
Include an AI Course Policy Statement in Your Syllabus
- Guide students on what, if anything, is permissible.
- Review your policy and expectations in the first class and periodically throughout the semester.
- Discuss the importance of ethical practices for students and out in the workplace.
Our AI course statement guide provides information and examples of different types of policy statements.
Design Assignments and Activities with AI in Mind
- Be cautious of assignments that are easily completed by AI (summaries, critical analysis, etc.).
- Favor discussions in class and activities that encourage immersive learning experiences (moderated discussions, study questions, group analysis, etc.).
- Be mindful of required source material (invite students to draw on and make connections between personal experience, specific sources, and course lecture materials).
- Encourage revision and reflection.
- Read AI and Critical Thinking from Academe, the quarterly magazine of the American Association of University Professors, which examines the threat of frictionless higher education.
Encourage Hands-on and Creative Activities
- Ask students to take notes by hand and allow them to share their notes with you as part of a weekly learning journal.
- Allowing students to submit handwritten homework is also worth considering. Students can easily scan their notes into a PDF format for submission. This 2024 article in Scientific American explains why handwriting is so beneficial for memory and learning. Consider sharing the article with them.
- Incorporate drawing, building, and other multimodal activities, such as podcasting and zines.
Experiment with AI Detectors
- Often, the best AI detector is you. As faculty become familiar with their students' work, it is not unusual to detect an inauthentic voice. When you do, proceed with caution, and instead of making accusations, just ask your student whether they used AI for any portion of the work in question.
- Try using AI models like Perplexity AI, Google Gemini, and ChatGPT to better understand the generated output.
- AI Detection tools like Pangram or GPTZero can be tested by submitting examples of human-generated writing, then creating some text using AI and testing it. If the result confirms AI. Just remember that AI detection is far from perfect; false positives are quite possible.
For more information on practical alternatives to AI detection tools, read the MIT Sloan School of Management's "AI Detectors Don’t Work. Here’s What to Do Instead."
Inside Higher Ed also provides an assessment of AI detection tools: "Professors Cautious of Tools to Detect AI-Generated Writing."
Resources
AI Course Statement Guide: A guide to design a course policy that suits your preferences, with examples for building clear AI course expectations.
AI Explorer: Learn what is happening in higher education and related industries including special projects, environmental impact, new products and related news, and explore AI technologies students and faculty are using.
Report from the FIT Faculty Senate Ad Hoc Committee on AI
SUNY FACT2 Guide to Optimizing AI in Higher Education
Possible AI Syllabus Statements (Brandeis Center for Teaching and Learning).
Citing Generative AI (APA and MLA styles covered).
Ethics of Creative AI: Detailed exploration of the ethics of using AI in many different settings and circumstances.
Student Guide to Generative AI From Barnard College