Endicott has taken the position that it’s up to the individual faculty member to decide how they want to handle student conduct and policies in their own class, as long as they adhere to the college’s Academic Integrity Policy that all students must follow. But faculty are starting to ask about the best way to deal with a student who they suspect has used ChatGPT to write at least part of a paper. We’ve compiled some tips, resources and suggestions we hope will help make a faculty member’s decision making a little easier.
(Note: The college believes that the existing policy as written covers the use of AI/ChatGPT and no new section or wording has been added to deal with instances of its use in a manner that does not comply with the policy.)
Some Recommendations for Starters
Here are some recommendations we think will be helpful, particularly if it’s your first encounter with ChatGPT or AI in general.
- Start by being upfront with your students about your position on the use of ChatGPT in your class. Some faculty allow it, others don’t. Some allow it but require students to cite their use of it and point out which sections were written by it, just as they would cite other sources or quotes.
- You should also let your students know that you will be checking all papers with one or more AI detection tools (as you do for plagiarism) as a regular part of your grading scheme so everyone knows the ground rules beforehand.
AI/ChatGPT Detection Tools
There are free tools available that detect ChatGPT. It’s useful to remember that all of these technologies are new and evolving so none of them are perfect. We suggest running a paper through 2 or 3 different detection tools to check them against each other. One may catch AI-generated text that others don’t. It’s also good to remember that they are sometimes wrong, so there are some suggestions in the links below on dealing with false positives.
Turnitin, the plagiarism detection application that is built into Canvas Speedgrader has released its AI detection tool as part of their existing application. You’ll see an icon highlighted in blue at the bottom of the Turnitin Similarity Report menu at the right of the Speedgrader screen. The number displayed represents the percentage of the paper Turnitin believes was written by ChatGPT or similar AI tool. Click on it and it will highlight the sentences or paragraphs it has detected, just like the plagiarism detection tool.
Turnitin has also put out a guide to approaching the use of AI-generated text in the classroom, as well as an AI Misuse Rubric. Both are worth reading, as are the resources below.
More Turnitin Resources for Faculty
Academic Integrity in the Age of AI: Resources for Educators
Guide for Approaching AI-Generated Text in Your Classroom
AI-Generated Text: An Annotated Hotlist for Educators
Writing Detection Feature Guide
Turnitin’s AI Writing Detection Capabilities
FAQs for Turnitin’s AI Writing Detection Capabilities
Understanding False Positives within Turnitin’s AI Writing Detection Capability
How to Prepare for and Discuss the Possibility of False Positives
AI Conversations: Handling False Positives for Students
Other free tools
There are other free tools available on the web that detect the use of ChatGPT, or AI in general. Here are some we’ve tried that work pretty well.
If the paper has been checked and the report indicates an AI app like ChatGPT has likely written some portion of the paper, what’s the next step?
- We recommend talking to the student and being upfront about what you’ve found. If you’ve already made your position known to the class ahead of time, then the student shouldn’t be caught off guard by this conversation.
- Refrain from taking an accusatory tone. Remember, these tools aren’t perfect so they don’t prove beyond a shadow of a doubt that AI was used.
- Show them the section suspected of being AI generated and ask them to explain. This can be a teaching moment rather than a “Ha, I caught you!” scenario. It’s entirely possible the student didn’t know they should have cited their use of ChatGPT or pointed out which sections were generated by AI.