Guidelines for teaching with generative artificial intelligence
Generative Artificial Intelligence (hereafter referred to as GenAI) in broad terms refers to a branch of computer science that utilizes machine learning and other advanced algorithms to generate new content de novo. This content includes text, images, code, audio, and video, and is typically created from prompts or other user inputs.
A point of interest is that the output produced by GenAI systems, such as ChatGPT, is often indistinguishable from that of a human. This development has created a need for establishing guidelines in the university setting that are concrete enough to provide immediate direction on best practices – including syllabus development – but can remain malleable as we continue to research this topic and gain a better understanding of its role in university pedagogy.
Based on results from a recent survey by the Applied AI Institute at Concordia, in addition to current literature on the topic, we are aware that there are diverse teaching approaches and learning outcomes across university campuses that necessitate tailored, context-specific guidance for GenAI and not blanket, university-wide policies. For this reason, the Centre for Teaching and Learning (CTL), in consultation with faculty, has generated guidelines while further consideration is being given to university policy.
Given the diversity of pedagogical approaches on our campus and the unreliable nature of GenAI detection tools, Concordia’s optimal strategy is to assist instructors in understanding the quickly evolving nature of GenAI in order that they may make informed decisions about its use in their classes. Above all else, it is essential that professors dedicate time to articulating their positions on GenAI clearly within the classroom, on Moodle sites, and in syllabi.
Syllabus statements provide guidance on addressing the following topics:
- General Statements
- Acknowledging GenAI
- Prohibiting GenAI Use
- GenAI Detection
These sample syllabus statements were developed by and large by Professor Naftali Cohn, Chair of the Department of Religions and Cultures at Concordia University. We thank Professor Cohn for contributing his perspectives on integrating GenAI technologies into university classrooms.
These guidelines provide sample syllabus statements on GenAI use for professors to adapt. Customizing the samples can promote responsible use of GenAI by students based on specific course learning outcomes and the demands of specific assignments or activities. Transparency on expectations and limitations is key to leveraging GenAI tools successfully. Clear syllabus statements are crucial for setting consistent standards, mitigating misuse, and fostering the ethical adoption of GenAI.
If one is allowing this type of use, here are some ideas for opening general statements about the use of AI tools (adopted from the University of Toronto):
- Students are encouraged to make use of technology, including generative artificial intelligence tools, to contribute to their understanding of course materials, under the circumstances outlined below.
- To achieve favorable results with generative AI, it is essential to invest time in building knowledge in the target subject and refining prompts, as this enables students to produce more accurate output while validating its accuracy and relevance to the topic at hand.
You may also consider adding statements about misuse, such as the following:
- Material drawn from ChatGPT or other AI tools must be acknowledged; representing as one’s own an idea, or expression of an idea, that was AI-generated will be considered an academic offense.
- Only some uses of ChatGPT or other AI tools are permitted. Prohibited uses and/or not sufficiently acknowledging use will be deemed misconduct under Concordia’s Academic Code of Conduct. Students who engage in these behaviours may be charged under Articles 18 (general cheating/plagiarism/dishonest behavior) and 19a (plagiarism) of the Code.
- In this class, writing assignments and submitting outputs that contain incorrect information related to class concepts, inappropriate responses to assignment prompts, or details that you are unable to explain or discuss in detail is considered a misuse of GenAI.
Clearly identify constraints in statements, such as the approach taken by the University of Toronto syllabus language document (adapted, and expanded):
- Students may use artificial intelligence tools for generating ideas, creating an outline for an assignment, or polishing language, but the final submitted assignment must be the student’s own work.
- Students may not use artificial intelligence tools for completing exams, writing research papers, or completing other course assignments, including creative assignments, posts to discussion forums, or smaller writing assignments. However, these tools may be useful when gathering information from across sources, assimilating it for understanding, improving writing, and even in the early stages of developing what they will produce for the assignment.
- Students may not use artificial intelligence tools for major assignments in this course, but students may use generative AI tools for smaller assignments.
- Students may use the following, and only these, generative artificial intelligence tools in completing their assignments for this course: .... No other generative AI technologies are allowed to be used for assessments in this course. If you have any questions about the use of AI applications for course work, please speak with the instructor.
If GenAI use is allowed, students must be instructed on proper acknowledgement. These statements are helpful additional statements from the University of Toronto:
- Students must submit, as an appendix with their assignments, any content produced by an artificial intelligence tool, and the prompt used to generate the content.
- Students may choose to use generative AI tools as they work through the assignments in this course; this use must be documented in an appendix for each assignment. The documentation should include what tool(s) were used, how they were used, and how the results from the AI were incorporated into the submitted work.
- Any content produced by an artificial intelligence tool must be cited appropriately. The MLA and APA are now providing information on citing generative AI.
- Generative AI may be used to draft some of the writing/phrasing, but cannot simply be cut-and-pasted, and can constitute no more than 25% of the text. You must also specify what work is your own and what comes from AI (e.g., by color-coding AI produced text or ideas).
- Acknowledge how GenAI inputs were created by documenting their workflow as well as prompts used.
Prohibiting GenAI Use
While some instructors may opt to prohibit GenAI entirely, this approach can be difficult to enforce in practice. As an alternative, we recommend focusing policies on setting clear thresholds and expectations around responsible use. However, for instructors who still find a need to prohibit GenAI use (e.g., for placement exams), here are sample syllabus statements:
- The use of generative AI tools is prohibited for all assignments in this course/exam. Their use in this course will constitute a violation of the Academic Code of Conduct.
- The use of AI tools like ChatGPT is prohibited for all items on this placement course/exam.
- Prohibited uses of ChatGPT or other AI tools will be deemed misconduct under Concordia’s Academic Code of Conduct. Students who engage in these behaviours may be charged under Articles 18 (general cheating/plagiarism/dishonest behavior) and 19a (plagiarism) of the Code.
Restrictive policies are very challenging to monitor and enforce effectively. They also limit opportunities to actively engage students in learning about responsible and ethical AI use. We also strongly discourage the use of GenAI detectors, given their unreliability. We instead encourage focusing on transparent syllabi statements that empower students to harness these technologies as learning aids while upholding academic integrity.
One of the biggest concerns of faculty continues to be the misuse of GenAI, whereby students use the tool to do their work, and consequently do not engage in the required learning. One response may be to inquire about the possibility of AI detectors, but it should be noted that online detectors, like GPTZero are known to be unreliable – commonly producing both false positive and false negative results. There are additional privacy concerns surrounding the use of such detectors, making them an overall unsatisfactory response.
It should be noted that to date, code violations using evidence drawn from detection tools have been challenged. While it is up to each individual professor to determine if they want to use detectors, we do not recommend their use at this time.
Identifying and evaluating potential misuse of GenAI requires a balanced approach. To this end, the checklist in table 1 aims to provide instructors with a three-level scoring system for detecting potential misuse. The levels range from potential evidence to strong evidence, with each factor weighted by point values.
While not definitive and not to be used as evidence for misuse, this checklist can aid in identifying assignments that merit further investigation, or discussion with students. It is recommended that instructors review this list with students at the beginning of the term so students can better understand boundaries for GenAI use in their coursework.
Table 1: Checklist for detecting potential misuse of GenAI
|Potential evidence||Writing is overly broad or generic||1|
|Potential evidence||Departure from student's usual style||1|
|Potential evidence||Lacks specificity related to class content||1|
|Moderate evidence||Incorrect information related to class concepts||2|
|Moderate evidence||Inappropriate responses to assignment prompts||2|
|Moderate evidence||Overly polished writing beyond student's abilities||2|
|Strong evidence||Student unable to explain or discuss work in detail||3|
|Strong evidence||Text that includes fabricated references||3|
|Strong evidence||Student admits to GenAI misuse||3|
Note: Please consider the ranges: Minimal evidence (0-2 points), moderate evidence (3-5 points), and strong evidence (6+ points).
Exemplary use checklist
In addition to mitigating misuse, it is essential to define and encourage exemplary use of GenAI. Accordingly, the checklist in table 2 ranges from developing skills to mastery, with each factor weighted by point values. This checklist can aid in identifying student work that demonstrates proficient to advanced integration of AI as a resource.
It is similarly recommended that instructors review this rubric with students at the beginning of the term so that students understand expectations for GenAI use in their coursework. The rubric provides a balanced method for recognizing students who use GenAI to enhance their learning and original thought.
Table 2: Checklist for detecting exemplary use of GenAI
|Developing||Writing reflects student's voice and style||1|
|Developing||Specific details related to class concepts||1|
|Developing||Student can explain in own words||1|
|Competent||Accurate information related to course content||2|
|Competent||Appropriate responses to prompts||2|
|Competent||Writing quality matches student's abilities||2|
|Mastery||Student cites AI as resource appropriately||3|
|Mastery||Student understands and can discuss work||3|
|Mastery||Writing shows original thought and effort||3|
Note: Please consider the ranges: Developing (0-3 points), Competent (4-7 points), and Mastery (8+ points).
While integrating AI tools into the classroom presents opportunities, it requires thoughtful guidance to uphold ethics and academic integrity. These best practices provide concrete recommendations across key areas including faculty support, student training, and promoting ethical AI use. The best practices aim to foster responsible AI adoption that enhances pedagogy without compromising rigor.
Provide comprehensive support for faculty members about the integration of GenAI tools in their courses. Address the benefits, challenges, and ethical considerations associated with AI use. Offer workshops, resources, and articles to help instructors make informed decisions on how to effectively incorporate AI while maintaining academic integrity.
Faculty need to equip students with the knowledge and skills needed to responsibly engage with GenAI tools. Develop modules that emphasize the ethical use of AI, including proper citation of AI-generated content. Incorporate interactive sessions where students can practice using AI tools effectively, acknowledging their usage, engaging in best practices, and avoiding potential pitfalls related to academic misconduct.
Promote ethical GenAI use
Foster discussions on ethical AI use within the student community. Engage them in conversations about potential biases in AI tools, the importance of acknowledging AI assistance, and the implications of misusing AI-generated content. Encourage students to reflect on the ethical considerations of relying on AI and making responsible choices.
- Ethical AI for Teaching and Learning, from Cornell University
- ENAI Recommendations on the ethical use of Artificial Intelligence in Education
- Wait — Is ChatGPT Even Legal? By Fenwick McKelvey
- AI Data Laundering: How Academic and Nonprofit Researchers Shield Tech Companies from Accountability by Andy Baio
Promote reflective practices
Incorporate a preliminary summary of intended AI tool use before assignments and a reflective piece afterward. Encourage students to assess the role and impact of AI tools on their learning process and assignment outcomes. This time also allows students to have discussions about prompts and other productive uses of GenAI in their work. This reflective practice fosters a deeper understanding of AI's influence on their educational journey.
FAQ for faculty
Given that GenAI can perform functions that would affect the outcome of almost any university assignment, one response is to consider banning its use. While this may be a reasonable course of action in certain circumstances, such as establishing a standardized score for a specific skill, this should not be viewed as a comprehensive response to GenAI in higher education. Instead, we recommend determining how to include AI in the classroom and some best practices.
We do not currently recommend the use of GenAI detectors due to their tendency to produce unreliable results. There is an additional issue with concerns about detectors unfairly flagging the text of second language speakers as being produced by GenAI. For these reasons and more, we do not recommend or support the use of GenAI plagiarism detectors. We note, however, that this is a quickly evolving field and that this may change in time.
The university does not currently have an official policy regarding the use of GenAI. One reason is that GenAI has not officially been adopted by the university, meaning that its use cannot be required in coursework. A second, more crucial reason, is that researchers and practitioners alike are trying to determine the best way to integrate AI in the university classroom. To learn more about how to navigate this evolving development while maintaining academic integrity, please refer to International Center for Academic Integrity or to the European Network for Academic Integrity.
Given that GenAI can perform a range of functions that would affect the outcome of almost any university assignment, re-considering assessments is an important direction., These efforts should not only focus on grading but also actively involve students in work that facilitates assessments and neutralizes the potential for the misuse of GenAI:
- Process oriented approaches: Students document the differences across drafts of a paper, highlighting what they did and the effect it had on their work.
- Pre- and post-assignment reflections: Students make statements about GenAI use in an assignment and then follow up, which can also serve as a basis for discussions about best practices, prompts, etc.
- Spontaneous, live discussions: Students have opportunities to competently, present, pose, and answer questions in a live setting, where using GenAI is not possible.
As our understanding of GenAI develops along with new tools and techniques, we recognize the need to continuously reassess best practices for responsible classroom integration. These guidelines will therefore be reviewed at least once per academic term to ensure they remain current amidst this rapidly evolving situation. We aim to provide guidance that balances GenAI's potential and risks in order to enhance pedagogy while promoting ethical use and academic integrity at Concordia.