At Abertay, we are using Turnitin’s AI detection tool. This tool is separate from the Turnitin Similarity tool which you can read about in the Turnitin guide.
Important: Student Access to AI Detection Results
|
You will not be able to see your AI detection results when submitting your work. The AI detection tool and its reports are for staff use only. You will only see your results if there are concerns about academic misconduct that need to be discussed with you. We are providing the information below so you understand how the tool works and what happens behind the scenes when you submit your work. This transparency helps you understand the assessment process, even though you won't see the results directly.
|
How does it work?
With the AI detection tool, when you submit your assessment, Turnitin will analyze your work to predict how much text they believe has been generated by AI. Please note the use of the word 'prediction' - the tool provides data to help educators make informed decisions based on academic and institutional policies, rather than making determinations of misconduct.
The tool only analyses complete sentences. It does not look at lists, bullet points, or other non-sentence structures.
Due to the differences between human writing and generative AI writing, the tool gives a score to those sentences which show highly probable word sequences, as that is how the generative AI models are trained. When we write as humans, we are not trained in the same way and thus there is "a low probability of picking the next word [a] human will use in [a] sequence" (Turnitin, no date).
Let's consider an example. If I use a generative AI model to write text using the prompt, "create a fictional story about a cat that begins 'Tom and…'", it may follow that after the words 'Tom and…', the predicted next word might be 'Jerry'. Why? Because scanning the texts the generative AI tool has access to, it notes that Disney had a cat character called Tom and he had a mouse friend called Jerry. However, if we, as humans, were writing that story, we may choose another name (and there are so many choices) for a second character. We'd probably deliberately not choose Jerry, as we'd want to differentiate our story from a Disney story.
Thus the word sequence 'Tom and Jerry' may be flagged as potentially AI-generated, whereas the word sequence 'Tom and Carmichael' won't be.
Here is what CoPilot generated when it was asked: Tell me a story about 'Tom and...' in 200 words. Tom is a cat.
Tom was a curious cat living in the bustling city of London. His days were filled with adventures, from chasing pigeons in Trafalgar Square to napping atop the warm hoods of black cabs. One day, Tom spotted a peculiar sight - a mouse, but not just any mouse. This one wore a tiny red cape and a golden crown. Intrigued, Tom decided to follow this royal mouse, named Jerry. Jerry led Tom through narrow alleys, across cobblestone streets, and finally to a small hole in an ancient cathedral wall. Inside, Tom was amazed to find a miniature kingdom, bustling with mice going about their daily lives. Jerry, it turned out, was their king. He explained to Tom that they had been living in fear of a large, menacing rat. Tom, with his cat instincts, decided to help. That night, he bravely confronted the rat, sending it scurrying away, never to return. From that day forward, Tom was hailed as a hero in the mouse kingdom. Despite being a cat, he was welcomed with open paws. And so, Tom and Jerry became the most unlikely of friends, their bond a testament to the power of friendship and courage.
If you are interested in learning more about next-word prediction, take a look at this article by Burtell and Toner.
What happens with your submission?
When your work is submitted, your tutors may see:
- An overall AI prediction score
- A detailed report showing potentially AI-generated text
- Any areas that may have been paraphrased using AI tools
A score above 20% (scores below 20% are not flagged) will probably mean that, in the same way your tutor would check your Turnitin similarity score, your tutor will use their academic judgement to look at what has been flagged as potential AI-generated text.
The only circumstances where you would see your AI detection results are:
- If the Academic Discipline Officer (ADO) needs to discuss concerns about potential academic misconduct with you (they may share your score and other information either orally or by screen sharing)
- If your case is escalated to a Student Academic Disciplinary Panel (you will receive a screenshot of the relevant AI score and report as part of the evidence)
Understanding AI Detection Reports
Although you won't see these reports during normal submissions, it's important to understand how they work in case you need to discuss them:
- Likely AI-generated text is highlighted in cyan (greenish blue) in the submission
- Likely AI-generated text that was also likely AI paraphrased is highlighted in purple
To ensure your work meets the University's academic standards, we recommend following the guidance in these resources:
- Acknowledging the use of Generative AI
- Critically Evaluating AI-Generated Information
- Generative AI and your education, assessment and success
- Referencing Generative AI
- Using Generative AI to Enhance Your Writing
For more information about scores and the report generation process, you can refer to Turnitin's AI writing detection in the classic report view guidance.
Note: We (Abertay) will update this guidance should Turnitin make any changes to student access to AI detection results in the future.
Not feeling confident yet?
Don’t worry! Email studyskills@abertay.ac.uk. We are here to help!
Last modified by Student and Academic Services