Categories
Talking Head Tools

3 Strategies to Prevent ChatGPT Cheating

ChatGPT-3 and other AI tools (e.g. JasperAI) are challenging student honesty and teacher creativity, with many articles highlighting just how powerful it can be as a cheating tool. Even turnitin will struggle to pick up everything. But fear not, here are some strategies to help make sure your assignments are resistant to a quick cut and paste.

1- Use The Blind Spots

ChatGPT is trained on a massive dataset, but that dataset ended in early 2022, and it cannot connect to the web to generate new answers or keep training. As a result, questions about more recent events, or new content, are at best estimates.

Additionally, ChatGPT’s “understanding” is averaged across every source it was trained on, meaning that depth will often be lacking. Asking it to use higher concept analysis can result in some quirky, and sometimes incorrect answers.

2- Skip The Writing

Anything that can be answered in text can be answered with ChatGPT. So, consider skipping the text entirely. Asking questions live in class, have students discuss themes, or produce an artifact.

For example, rather than giving an essay, ask for a diorama or challenge students to create some sort of artifact and then explain their choices. This way even if they use ChatGPT to save reading or learning the content to begin with, they will need a nuanced understanding to produce a good representation and justification for their artifact.

3- Create a Case Study

ChatGPT is trained on existing information, so if you use a book or information source that’s already well documented ChatGPT could have already been trained on it. If you come up with something original ChatGPT doesn’t know about, then it can’t answer the questions especially well.

This is also a chance to actively try and trip up learners who might try ChatGPT without proofreading their work. Using celebrity names or metaphors in your case studies can trigger mistakes in the output of ChatGPT.

Wildcard- Let Them Use It

This one is the most interesting to me. Let them use ChatGPT, but go hard on the marking. If students are allowed to use ChatGPT then they should be proofing, fact-checking, and otherwise polishing it to make sure it is right. This way it doesn’t matter what they get out of ChatGPT, they will need to confirm what it’s producing, thereby reinforcing their own understanding. Be careful with this one though, it can become a massive marking time sink. Maybe you need an AI Marking program (not recommended).

Leave a Reply