Reviewing Assignments for AI Use
When reviewing assignments, we want to ensure that students meet mutually created expectations. Are students using AI tools in a way that is consistent with the expectations that we set as a class? If not, we will need to open a dialogue with students. This page offer guidance for these conversations.
Grammarly Authorship
Critical thinking about AI and how to use AI responsibly have quickly become vital workplace skills. While many of us are still learning about AI tools in general, and chatbots specifically, it’s important that we help students develop this skill. The Grammarly Authorship tool can help!
What is does
Grammarly attributes which text was human typed, AI generated, edited with Grammarly, modified by gen AI, and more, all with a sharable Authorship report and replay.
Why it matters
Demonstrate how text was created in any document and gain insight into the writing
process.
How it works
- Students enable Grammarly Authorship in Word or Google Docs.
- Students write their papers with the Authorship tool enabled.
- Grammarly Authorship then identifies where the text in the document originates.
- Text Categories:
- Typed by a human
- AI generated
- Modified with Grammarly’s gen AI
- Pasted from a known source (w/attribution)
- Pasted from an unknown source
- Edited with Grammarly or Gdocs
- Text Categories:
- Students then submit their assignment with the authorship report.
Tutorial Video - Grammarly Authorship Overview PDF
Context
Turnitin embedded AI detection into its platform. Turnitin's tool (along with other AI detectors) will likely have flaws as they have been developed quickly and primarily based on an older version of ChatGPT - there were four major versions of GPT from November 30 - March 14. So, we need to be cautious about the accuracy of detection tools as the primary source for determining student voice and original work.
For example, Turnitin's predictions have a 1% false positive rate. This rate means that one of every 100 papers it identifies as being written by AI will be incorrect. To understand a bit more about what that means, please watch this video from David Adamson at Turnitin.
What does this mean? Essentially, this functionality from Turnitin, or any detector, is a tool that can open doors to conversations. We still need to be giving students the benefit of the doubt and approaching them for conversations with open minds. If we find that a student has used an AI writing tool and submitted that work without citation, we should also seek to understand the reasons that the student has done so. Were they in a hurry? Did they not understand the material? Discovering a student's reasoning for turning to AI will help us better support that student.
Turnitin Resources:
- To learn more about Turnitin's technology and its ability to spot AI please visit their AI writing detection page.
- To better understand how to use the Turnitin AI writing detection tool in your class, please visit their page that explains the AI writing indicator.
- For some frequently asked questions developed by Turnitin, please visit their FAQ page.
How to enable Turnitin in D2L and how to check AI originality
Other AI Detectors
If you are concerned about students using AI, like ChatGPT, to write essays and complete other work, there are a number of AI detectors to help you identify whether a human or AI wrote something. These AI detectors will continue to improve alongside of AI. Below you will see a prompt to and response from ChatGPT. The accordion below contains the AI Detectors we tasked with checking the ChatGPT response. Please keep in mind that ChatGPT is iterating very quickly - ChatGPT was released on November 30, 2022 and by March 14th, 2023, the developers released three updates. So, we need to be cautious about using these detectors as the primary source for determining student voice and original work.
Chat GPT Prompt:
What would an expert analyst say about the current state of the stock market and specifically the tech stocks? Does the current market condition match any other periods in history? Make the response be a little poetic in style and also make sure to include very aggressive and extreme words that generate an emotional response. Change direction frequently in terms of tone going from, funny, serious, witty, sober and sarcastic. Make it about 5 paragraphs long.
Chat GPT Response:
The stock market is a wild and unpredictable beast, my dear reader. One moment, tech stocks are soaring to new heights, reaching for the stars. The next, they come crashing back down to earth, their value plummeting faster than a lead balloon.
But make no mistake, the current state of the market is no laughing matter. This is a volatile and dangerous time, where fortunes can be made or lost on the turn of a dime. The tech giants of Silicon Valley dominate the landscape, their power and influence only growing with each passing day.
Yet despite this, there is still hope. For those with nerves of steel, the risk can be worth the reward. The bold and the daring, those willing to take risks and chase their dreams, they stand to reap the greatest rewards.
But be warned, dear reader. The market is a fickle mistress, and she does not play fair. One false move, one moment of weakness, and it all could come crashing down. So tread carefully, my friend. Stay vigilant, stay focused, and above all, never let your guard down.
In short, the current state of the stock market is a volatile and unpredictable rollercoaster ride, where only the bravest and most cunning will emerge victorious. So buckle up, my friend, and hold on tight. It's going to be a wild ride.
Result by Detector (for the Stock Market prompt above):
http://gptzero.me/ : Most Likely AI. However, when a few human sentences were entered throughout the text, it resulted in “most likely human.” So far, this is the best performing detector in our tests.
https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text/ : The classifier considers the text to be unclear if it is AI-generated. Note: This is ChatGPT's new classifier.
https://writer.com/ai-content-detector/ : Human
https://copyleaks.com/features/ai-content-detector 84% Human
https://openai-openai-detector.hf.space/ 52% Fake
https://crossplag.com/ai-content-detector/ : This text is co-written by both a human and AI
https://contentatscale.ai/ai-content-detector/ 28% Real
http://gltr.io/dist/index.html : Interesting results. Text is colored in various colors depending on the probability that it is AI generated. Based on that, this tool did pretty well at detecting. However, it also colored mostly green (AI) some human text entered, so the sensitivity might be so high that it may produce high levels of false positives.
https://fourweekmba.com/ai-detector/ : 47% AI
Frequently Asked Questions (FAQ)
We know that there are, and there will continue to be many questions from faculty about Artificial Intelligence (AI), especially the use of ChatGPT. Given how rapidly evolving these programs are, NLU is working to develop comprehensive policies related to the use of AI or ChatGPT. The work will be done through the Provost Office and shared governance structures in the upcoming months. In the meantime, the Provost Office is sharing the following FAQ guidance and resources for your use. The FAQ and this page will be treated as a living document and will evolve as we work to understand the implications of AI within the classroom.
AI generated text tends to be atypically accurate in grammar usage, voiceless and lacking in any human context or presence and follow predictable writing formations. Paragraphs begin with strong topic sentences and end with a summary sentence, generating very formulaic text. However, AI chatbots, like ChatGPT, can be instructed to take on a voice (such as write like a high school freshman). There are AI Detector tools available, however, they typically are not 100% accurate.
See above for more information on Detection Tools.
Most notably, please see the info regarding the Turnitin AI Writing Detection Tool
Please keep in mind that AI is a tool that can assist faculty and benefit student learning, if used properly. Given the speed at which AI is evolving and the implications that will have on the workplace, we recommend an approach that teaches students how to properly use the tool rather than placing energy into prohibiting the use of the tool.
The APA has provided the following guidance for citing ChatGPT and other chatbots.
Citing Generative AI (APA)
The Library has provided the following guidance for ethical use of generative AI, including ChatGPT and other chatbots.
Also, please see this blog regarding citation of AI tools.
