AI Literacy in the Age of ChatGPT

Many educators don't support the use of AI writing detectors

These detectors result in both false negatives and false positives, which can result in tagging a student paper as written by AI when it wasn't.

Suggestion from UCATT:

AI-detection tools can be used to assist in reviewing student work, but detectors are problematic because they 

• Frequently mistake things that are human-written as AI-written
• Disproportionately misidentify writing by multilingual learners as AI-written
• Are not transparent about how they work
• Do not provide sufficient evidence on their own to conclude that a student has violated course policies 

Instead of AI-detection tools, you can:

Have students share their work/thinking processes and how they used AI

• Integrate students’ use of AI, e.g., as a reviewer, editor, assistant, or tutor
• Check students’ understanding of acknowledgment/citation/disclosure practices
• Facilitate discussion of the pros and cons of AI for learning
• Implement best practices in assignment instructions, e.g., explain the purpose, why it matters, what to do, how to do it, which resources to use, with whom to work, and the evaluation criteria
• Shift to more frequent low-stakes assessment in-class / live 
• If you choose to use AI detection, include a statement in the syllabus that informs students. Do this before any assessment or grading begins. If possible, include a link to the privacy policy of the tool(s) you will use. 

Use of AI-Detection Statement Example: Your work may be shared with one or more AI-detectors that are designed to predict if it was created by a generative AI, e.g. ChatGPT.