Over 2024, Artificial Intelligence (“AI”)* has highlighted and intensified deficiencies in academic misconduct decision-making at UVic. These issues have become so pronounced that I have chosen to focus my Annual Report Recommendation solely on improving the decision-making process outlined by the current Policy on Academic Integrity (the “Policy”). Below I will outline the current process, for those who are unfamiliar, and identify the problems with examples. Finally and in-line with our commitments to equity, fairness, and respect,¹ I will offer solutions in the form of  recommendations. 

After an instructor suspects that a student is using AI, the Policy instructs them to document the allegation and inform the Chair of the Department. After the Chair gives the student a reasonable opportunity to respond, the Policy indicates that “the Chair shall make a determination as to whether compelling information exists to support the allegation.” This is the burden of proof.  

The Chair’s determination must be decided on a balance of probabilities — if the allegation was more likely to have happened than not. This involves an analysis of the information (evidence) that supports and contradicts the allegation. The Chair examines each piece of evidence, decides if it is compelling (i.e. believable or persuasive), and compares it to the others. The Chair may look at the work in question, the course outline, the student’s response, the documentation that the instructor gave, amongst other things. It is not the student’s responsibility to provide proof of their innocence, and if they fail to do so, the University is still responsible for providing the “compelling information” needed to reach the burden. If there is no clear, convincing, or believable evidence at the end of the Chair’s analysis, the University has not met its burden of proof. If the Chair believes the burden is met (a violation was more likely than not), they must inform the student in writing. Importantly, they must communicate how and why they have concluded the burden was met.  

These aspects are key to creating a fair process where, as far as possible, penalties are given only to students who have violated the Policy. They are also key to creating a system where  abuse of power, bias, and unfairness are not tolerated. 

The Problem

There has been a 23% rise in Academic Integrity Cases from 2023 to 2024. This is significant. Academic Integrity cases now account for the most frequent reason a student visits the Office to report unfairness.  

Many, if not most, of these Cases involve AI. Advancing an allegation about AI and deciding if it was more likely than not is complex and demanding.² The frequency of AI misconduct is only increasing and the resources for decision-makers has not kept pace. Many instructors have not adapted to address AI in their course outlines or assessments. AI technology is fast moving and we are all new users — both students and educators alike are both trying to keep up. AI has stressed UVic’s academic misconduct process — perhaps to its limits. 

As a result of the complexities and demands of AI, I have observed an increasingly more frequent trend where decision-makers neglect to provide procedural fairness rights to students accused of academic misconduct. 

Below, you will find six areas where I have observed this trend on campus. Each of these areas is accompanied by a real-world example(s) drawn from a student’s Case. I will also offer practical suggestions for decision-makers on how avoid each pitfall and ensure a fair process for each student.

• Flipping the Burden: When there is little concrete evidence to prove an allegation (which  is common with AI), it may be tempting to ask the student to prove their innocence and rely on their lack of excuse as evidence that misconduct was likely. However, in this scenario, the University has failed to recognize that it holds the responsibility to supply the information to satisfy the burden of proof. This is clear in the Policy. If there is no information or it is not compelling, the allegation must be dismissed. Without this principle, anyone can be accused of misconduct — and without an excuse — can be found guilty.

Example: “As you were unable to provide any reasonable explanation otherwise during our meeting, I find that you violated the Policy by using AI when it was not allowed.” 

Suggestion: Ensure that you are not relying improperly on the lack of evidence to the contrary. Be certain the burden was met by the confirmatory evidence and explain in writing how. If you are not certain misconduct was likely, dismiss the allegation.

• Suppressing the Student’s Story: Some decision-makers may think a student is only going to find an “excuse” or a way to get out of trouble. Getting the student’s version of events adds complexity and doubt to the process, which is already burdensome with AI-related allegations. These factors can result in tactics to suppress a students side of the story, either consciously or unconsciously. For example: 
  • Providing only a short time (sometimes 24 hours) to respond to an allegation, which results in poor articulation of a student’s story and a weak defence.
  • Not providing full details of an allegation or being vague, which results in the student being taken off guard and being unprepared.  
  • Not giving a student access to the evidence, which results in the student not having the opportunity to question its accuracy, validity, or relevance.  
  • Not mentioning a support person may attend, which results in the student being isolated and unaware of their rights. 
  • Not mentioning the right to appeal, which makes it less likely a student will know of their rights and can been seen as trying to avoid scrutiny.

Suggestion: Ensure the student is given a reasonable opportunity to respond and is aware of their rights to support and appeal. See my recommendation from 2023: Trial by Surprise. Use the Ombudsperson’s Template for an Academic Misconduct Notification Letter.

• Hiding the Rationale: If it is unclear or if you are unsure if the burden is met, it can be  tempting to avoid writing down how and why you decided that misconduct likely occurred and how and why you chose a specific penalty. Explaining your rationale can also take time and is difficult to articulate. However, students have the right to know how and why each decision was made. This can help a student accept a decision and protect against faulty reasoning. Committing all your thoughts to words can also be an exercise to help you make up your mind. It certainly can reduce appeals. Rationale that mirrors “because I said so” can also conceal abuses of power and unacceptable biases. 

Example: “Dear Student: After meeting with you and reviewing the evidence, I find that you committed academic misconduct. The Policy states that ___ is the penalty.”

Suggestion: Reflect on your own thinking and examine your rationale. Write how down how you methodically and logically thought through all the information. Include the how and why you came to a decision and how and why you chose a penalty in a written decision letter. Use the Ombudsperson’s Template for an Academic Misconduct Decision Letter.  

• Working Backwards from a Hunch: Many educators may first think that a student used AI³ and then will seek out evidence to support their conclusion. This approach can lead to disregarding or undervaluing counter evidence (confirmation bias) and infringes the student’s right to an unbiased and open-minded decision-maker. The student must be treated as innocent until found guilty. This is important to maintain in mindset and in communications. 

Example: “I have found clear evidence of the use of AI in your assignment. I am inviting you to a meeting to discuss these allegations.” 

Suggestion: Avoid making a conclusion before you have heard all of the evidence, especially before you’ve heard the student’s side. Throughout the process, actively remind yourself to be receptive to new information and open to having your mind changed. Document your reasoning to test it for biases, so as to be ready to describe your mindset in your decision letter and on appeal. Consider having a colleague review it, but protect the student’s privacy. Use the Ombudsperson’s Template for an Academic Misconduct Decision Letter.

• Not Treating a Student as an Individual: AI’s accessibility or the frequency of  AI-related misconduct can be questionably linked to a student’s likelihood to commit academic misconduct. This also extends to incorrect connections drawn between an individual student’s case and other students’ behaviour. These approaches can encourage a disregard for a student’s individual story. This is inequitable and can lead to penalties being applied to students who did not commit academic misconduct.  

Example: “The words used by many students who confessed to using AI are also appearing in your answers” or “Students are more likely to cheat these days, because it’s so easy to use AI.”

Suggestion: Give little (if any) weight to evidence of what other students have done, unless it is relevant to the student’s case. Avoid the “students cheat” mindset and recognize that it may bias your decision if not managed. Demonstrate in your decision letter how you considered the individual factors that the student may have shared with you. Follow Ombudsperson’s Template for an Academic Misconduct Decision Letter. 

• Unclear or Shifting Rules: Students are often unaware of what is allowed and what isn’t. The Policy definitions are broad, and the decision-maker often places responsibility on the student for not inquiring beforehand. Many course outlines do not mention the rules around AI and enforcement is inconsistent. The University is making the rules, enforcing the rules, and has more resources and expertise than most students. The University therefore ought to shoulder the responsibility of making the rules clear. Holding a student accountable to a shifting or unclear rule is unjust and can disregard a student’s possible good faith actions — how are students supposed to follow the rules if they don’t know what they are?

Example: “It was reasonable for you to have known that using Grammarly was academic misconduct.”

Suggestion: If the course outline or other instructions lack clear rules on AI use and it is not explicitly prohibited by the Policy, consider how this may impact the student’s level of culpability. If the student raises this issue, be sure to address it in your decision letter. Afterwards, follow up to ensure that AI-related expectations are clearly stated in the course going forward.

The Solution

UVic, like many higher education institutions, is in novel territory. But, AI is here to stay. 

Dr.  Sarah Eaton, a scholar whose research focuses on academic ethics in higher education, shares a possible future scenario where scholarship is created using both human effort and AI, where “determin[ing] where the human ends and where the artificial intelligence begins is pointless and futile.” (S. Eaton, Plagiarism in Higher Education: Tackling Tough Topics in Academic Integrity, 1st ed (Bloomsbury Publishing, 2021) As part of a post-plagiarism concept, she shares that “[h] istorical definitions of plagiarism will not be rewritten because of artificial intelligence; they will be transcended. Policy definitions can — and must — adapt.” (Eaton, 2021).

I agree that we must adapt. 

Be it towards Dr. Eaton’s post-plagiarism world or, perhaps as a stepping-stone, towards developing a fair, responsive, and equitable process to address academic misconduct at UVic. 

We must remember that each of the “examples” I recounted above was drawn from a real student’s circumstances. Of course, an allegation of academic misconduct is unavoidably going to cause a student stress — however, these “examples” did more than that. Not only did these students face an allegation, they also faced a process that was unpredictable, unfair, or in some cases, unethical. Their health, reputations, education, and futures were on the line. This caused, from my observation, much more stress than was warranted and irrevocable damage to the students’ relationships with UVic.