According to Trent's Undergraduate and Graduate Academic Integrity Policies, unless it is specifically authorized by an instructor, student use of AI generators is considered an academic integrity violation. However, instructors often find the process of investigating whether a student has used AI generators to be more difficult than demonstrating plagiarism from online or published sources.
Preventative Strategies
- Talk to your students about academic integrity and AI use: Much research supports the importance of openly discussing academic integrity with students (Baetz et al., 2011; Reedy et al., 2021). Similarly, talking to students about why you think certain uses of AI will interfere with their learning can help activate their commitment to completing work with integrity (Academic Integrity Council of Ontario, 2023).
- Ask students to track rough work: Having students document their process makes it more difficult for them to quickly and mindlessly copy and paste text generated by ChatGPT; it also supports good analytical and writing processes by asking students to document their steps toward developing and expressing ideas and arguments. The Centre for Teaching and Learning offers Student Instructions for Documenting Research and Writing Processes and a Source Log Template that you can adapt as needed to help students track their work in completing assignments.
- Explain how you want students to cite AI use: Establish expectations for citing, referencing, or acknowledging the use of AI applications in course work and assessments through assignment instructions and syllabi (Academic Integrity Council of Ontario, 2023). Students can find information on how to cite AI generators through the Academic Skills Documentation Guide.
Collective Evidence of Unauthorize AI-Use:
- Identify "tells" of AI Use: Consider the factors that made you question whether the student used AI to complete the assessment. Is it very general in nature? Does it lack sources or examples? Or, does it contain citations to sources that do not exist? Does it draw on themes and ideas far outside of the course content? Just as you would in any case of suspected academic dishonesty, you should highlight the evidence within the paper that makes you suspect AI use.
- Check the student's rough work: If you required students to submit rough work and notes, you can use version history to track how their draft progressed over time. Look for instances in which the editing time seems too low and/or there is no evidence of progression of thought or work.
- Compare the text to what AI generate for a similar prompt: You may wish to put the prompt for the assignment into an AI generator, such as Bing Copilot or ChatGPT. Is the generated text similar to what the student submitted? And, if you put the same prompt in to the generator multiple times, does the text remain similar? If so, you should print out/save copies of the AI-generated text and highlight the similarities to the student’s assignment.
- Compare the assessment to other work the student completed: Does the text seem significantly different from the text the student produced for other assignments? This, of course, does not prove that AI was used, but it can become part of evidence that indicates AI use.
Engaging the Student in a Discovery Interview
Trent's Academic Integrity Policy states that, as part of their investigation of academic dishonesty, instructors should invite the student to discuss the matter. This “discovery interview” is particularly important to investigating possible unauthorized AI use. Without “hard” evidence of plagiarism (as one might have if the student had copied a published source), in cases of suspected AI use, the instructor can ask the student questions about their work and their process to determine the likelihood that they completed the assessment through their own efforts. Below, you will find some helpful questions for the discovery interview:
- Tell me about your work. Why did you choose this topic?
- Tell me about your process for completing your assignment? How long did it take you?
- Can you show me any notes you took while preparing your assignment or any rough drafts? (You can find instructions for students on how to use version history to recover drafts of their work on the TTC website).
- Tell me about the sources your consulted when you did your assignment. Where did you find these sources? Why did you choose these particular sources? When you think back to the sources you read for this assignment, can you tell me verbally what your key learnings where from what you read? (You can find a source log for students to track answers to these questions on the TTC website).
- We didn't use (insert source here) in our class discussion at all, and it wasn't on our reading list. How did you find it? Why did you choose it?
- Tell me about your conclusion. How did you arrive at this conclusion?
- The assignment instructions ask for X, Y, Z. You addressed X, but not Y. How come?
- Your work mentions these terms (insert terms here). Can you tell me verbally what these mean?
- For coding assignments: What does this function do? How does this code work? What does this code make happen?
- Is there anything else you want to tell me about your work?
Please note that the questions about were adapted from Sarah Eaton's (2019), "How to Lead a Discovery Interview about Contract Cheating," CC NY BC.
On the linked document below, you will find these questions presented in chart form. We are indebted to C. Scurr (2024) "Process guide for suspected misconduct," Conestoga College for the conceptualization of the chart format (click here).
Why Doesn't Trent Recommend the Use of an AI Detector?
On the You will notice that these guidelines for investigating AI use do not include the use of an AI checker. We know that not having a checker may be frustrating. But, there are some clear reasons why we are not recommending their use at this time:
- Protecting students' intellectual property: SafeAssign does not have an AI checker included in its plagiarism detection software. This means that, currently, there is no Trent-supported platform that can predict whether text was AI-generated. Submitting student work to a non-Trent platform means that Trent does not control where and how that content is stored or used.
- AI checkers have not been shown to be reliable: Many universities that do have AI-checkers as part of their anti-plagiarism software have disabled the feature. This is because of research that questions the reliability of the checkers.
- AI checkers have been shown to be biased: There is research that shows that AI-detectors unfairly flag the work of students whose first language is not English.