Learn about the teaching technology available to help you detect AI text.
Understanding the Tool
Navigating the world of AI technology in education requires balance. AI chat technology such as ChatGPT, Google’s Bard, and others are here to stay, and instructors have to grapple with that, as examined in Faculty Focus, Inside Higher Ed, and the Chronicle. While the upshot is that AI chatbots could be used for innovative active learning, we must maintain academic integrity with this new technology while judiciously reviewing student work.
What is AI text and ChatGPT?
AI text is computer-generated predictive text. It is similar to what you see in Outlook, Google Documents, or most text message services when it offers the next word or phrase it has identified as most likely to be used next.
Chat GPT, Bard, etc. are merely instances of the technology that scans vast amounts of data from the internet to provide what it identifies would likely be the next word.
Is there a solution?
Turnitin has developed AI text detection software embedded in their existing plagiarism detection software.
You will see a percentage of the document Turnitin flags for further inspection.
How does Turnitin detect AI text, and in what kind of writing?
Redundant writing, such as paraphrasing previously written text and repetitious word use, is more likely to be flagged as AI. This technology is programmed to examine English language paragraphs.
Documents that contain lists, code, poetry, etc. will not be detectable to Turnitin’s software.
Turnitin breaks the text into roughly five to ten sentences, and its AI detection model awards a score between 0 and 1 to the text. A 0 indicates that the text is most likely written by a human. A 1 shows that the sentence was AI-generated.
An aggregate score is presented, and it indicates how much of the document was flagged as either likely AI-generated or completely. This operates similarly to the Similarity Score.
How reliable is Turnitin’s AI score, and what about ESL student writing?
These scores are shown with 98% confidence per Turnitin’s data and tests.
There is a slight chance for a false positive; however, according to Turnitin’s testing, this has occurred in less than 1% of their tests.
Since their model is looking for redundancy in English written paragraphs, it could flag student submissions where the learner is still developing their writing skills, limited in their English language vocabulary, or both.
So, it is essential to use this detection software as a tool for you to examine the student’s submission and to make a reasonable judgment call as best you can.
How will I know if Turnitin detects AI text in a student submission?
When examining a student submission through Turnitin, you will see a score at the bottom of the right-hand window pane in blue. Simply clicking on this score will show the document with the blue-highlighted text to indicate which text you should review.
Where can I learn more about this?
Live Training
Register for a live training session on how to use Turnitin’s new feature. Quick Hit: Turnitin & ChatGPT on April 27th.
Turnitin Resources
Understanding False Positives for Instructors
Preparing for False Positives for Instructors
False Positives Document for Students
Upcoming Events
Quality Matters for Summer Accelerate
Clemson Online is proud to announce new opportunities for instructors to gain experience in online course creation, teaching with Canvas, and the Quality Matters course review process. They include:
- Online Course Bootcamp
- Improving Your Online Course (IYOC)
- QM Essentials Review
- COFFEE: House Blend
Please see the full descriptions and registration links at our Quality Matters for Summer Accelerate website.
For applicable participants, these trainings can be listed on annual reviews in the Faculty Success system. Check with your department for specific details. If you have any questions about these training opportunities or any aspect of Quality Matters at Clemson, please email Lori Kinley.
Next Semester Prep Workshop
As we cross the finish line on another Spring semester, do yourself a small favor before taking your well-deserved break: register to receive a link to a recorded training on best practices in preparing for next semester. In this session, we will cover exporting and importing Canvas course content, identifying one actionable change to make for next semester, and reviewing your course learning objectives to ensure they align with the rest of the course. Facilitated by James Butler, Digital Learning Strategist.
Note: This development session has been revised from live training to a pre-recorded training that registrants will receive a link to.
Register here for the Next Semester Prep Workshop on April 19th .
Clemson Online Spring 2023 Events Calendar
Review our Spring 2023 Events Calendar to see what Online Instruction Development opportunities await!
We have a robust Spring semester lineup of topics and live training formats to support your use of Canvas and other e-learning tools. Topics cover demonstrations of using Kaltura, presentations on inclusive practices for online education, and workshops to get your Canvas site ready to teach!
All of our live training is recorded. Registrants will automatically receive a link to that day’s video after it has been processed.
Contact James Butler with any questions about these sessions.
Clemson Online – Where Tech and Teaching Meet
