In June 2023 McMaster launched an initial set of Provisional Guidelines on the Use of Generative AI in Teaching and Learning. These Provisional Guidelines were updated further in August 2024 and then used throughout the academic year.

After a further year of teaching and learning with generative AI, these Provisional Guidelines require some revision. These revisions were informed by feedback from faculty, students and staff and were written and endorsed by the Expert Panel on AI in Teaching and Learning and the AI Advisory Committee. They are intended to guide both educators and students in understanding and interacting with generative AI in teaching and learning contexts. Staff may wish to consult the Provisional Guidelines on the Use of Generative AI in Operational Excellence.  

In this update we move from “Provisional” to “Guidelines” recognizing that the Guidelines are not a policy but do intersect with existing McMaster policies and that the Guidelines will need further updating.  

Questions, comments or suggestions about these Guidelines may be directed to the Vice-Provost, Teaching and Learning or to the Special Advisor to the Provost on Generative AI at macgenai@mcmaster.ca 

Learning requires active engagement: building mental models and habits of mind, then communicating, applying, evaluating, and reflecting on those mental models and habits of mind. Learning requires work – often slow and challenging work.

Generative artificial intelligence offers a promise of efficiency and speed, a promise that can be at odds with the deliberate effort of learning. At McMaster, we suggest that generative artificial intelligence tools should be used for learning only when the educator judges that their use will aid in active, critical, and reflective engagement.  

In most disciplines, university-level learning focuses on process, rather than the end-product. In other words, it is more important that students learn the how (e.g. how to find a solution, how to evaluate sources, how to construct an argument, how to perform a lab task, etc.), as opposed to the what (i.e. specific content). Generative AI poses a significant risk to the process of learning by bypassing or obscuring the how. 

While generative AI poses these risks to learning, at McMaster we remain open to generative AI use for teaching and learning, as well as experimentation with new tools and techniques when they advance a course or program’s goals. Educators experimenting with generative AI for its benefits for learning will want to be clear on pedagogical goals and how generative AI will advance those goals.  

Beyond learning, generative AI introduces a range of concerns, including environmental impacts, disinformation, impacts on labour, questions of copyright and ownership, lack of transparency in model design and function, privacy and data collection and use. Educators and students can learn more about these issues and should discuss them in courses where appropriate. Educators can get support in talking with their students about generative AI through the MacPherson Institute, the Privacy Office, or through these resources.

Generative artificial intelligence systems are trained to produce outputs that are plausible in each context. The models learn to associate inputs to outputs by processing massive amounts of text and image data from the open web and other sources. Despite their ability to produce human-like outputs, there is no proven correspondence between the way that human beings think and learn and the way that generative AI models learn, process, and generate text, images, sounds, and other media.

Because of the way they process training data and produce their outputs, generative AI systems are prone to producing ‘hallucinations’ (false statements, impossible images, false citations and attributions and other errors). Likewise, these systems produce outputs that are biased by their training data; these biases are often subtle and thus hold the risk of being uncritically reproduced. There is no current solution for the hallucination and bias problems.  

As generative AI models produce output that cannot be guaranteed to be accurate and unbiased, university-educated domain experts will continue to be required, both to produce original work and to verify the accuracy of the work that generative models produce. McMaster must continue to train students in the core competencies of any domain of study, regardless of the ability of generative AI models to answer undergraduate-level questions in that domain.

Deciding as a learner and educator whether, when, and how to use generative artificial intelligence requires careful thought and evaluation of the benefits and risks to learning. It is ever more essential for educators to carefully consider what skills they need students to learn – that is, what process of learning – for the course and program. Educational developers at the MacPherson Institute can support this planning.  

• 1. Undergraduate and graduate course outlines should include a statement on the acceptable and unacceptable use of generative artificial intelligence in the course, with attention to the use of generative artificial intelligence for studying/learning and/or assessment. For instance, students may or may not use generative AI tools for idea generation, structuring, editing assistance, translation, summarizing/paraphrasing information, etc. (refer to the Use of Translation and Editing Tools guide for information about how to use these tools responsibly in the context of academic integrity).
  • If no syllabus statement is included, students should ask the educator for clarification on expectations, and if generative AI use is permitted, receive written confirmation before using generative AI in the course.
  • Educators should review expectations on the use of generative AI with their students in class [see: sample slide deck].
• 2. Educators incorporating generative AI into courses should consider how the course and its goals align with those of the program and with program learning outcomes. Educators should consult with the Departmental chair for the program learning outcomes and any program or departmental norms and expectations related to the use of generative AI at the program level. Programs may wish to establish norms and expectations of generative AI use at the program level. Support for this work is available by contacting macgenai@mcmaster.ca 

• 3. Use of generative artificial intelligence by students in ways not described in the course outline may be cause for a violation of the academic integrity policy. Likewise, the submission for course credit of uncited or unacknowledged work not created by the student(s) who submitted it violates academic integrity (See McMaster Policy on Academic Integrity). Reach out to the Office of Academic Integrity with any questions on process or for support. Reach out to the MacPherson Institute at mi@mcmaster.ca for support with redesigning assessments.  
• 4. Automatic AI detection systems are not reliable and are therefore not recommended for separating student- from AI-produced work. Educators who suspect work may have been inappropriately generated by artificial intelligence should follow the academic integrity process. This conversation guide may be useful for talking with students. 

• 5. If using generative artificial intelligence in courses or for teaching and learning activities, educators and students should use institutionally supported tools that have a completed Privacy and Algorithmic Impact Assessment.  McMaster has an enterprise license with Microsoft Copilot. Educators and students who use generative AI should use their McMaster single-sign on credentials to login to Microsoft’s Copilot.
• 6. In selecting third-party technology tools educators must avoid those that sell student data to companies building large language models, as well as companies that use student data to train AI models or to improve services and products; educators should review user agreements and consult with the Office of Legal Services or the Privacy Office if unsure.
• 7. Legal questions of intellectual property (such as copyright and privacy) continue to be evaluated by provincial and federal courts. Until such questions are resolved, employees should not use generative AI created content for proprietary work and all McMaster policies continue to remain in effect.

• 8. Students in co-op and experiential learning environments that use generative AI as directed by or approved by the co-op/experiential learning employer should initiate a discussion with the course instructor on how to document and reflect on this use with respect to the course learning.  
• 9. Students and educators who use generative AI should complete this module to help them review and reflect on broader societal implications of the use of generative AI, including labour, copyright, bias, and environmental impact.
• 10. Students and educators who use generative AI in any context within a course should cite or acknowledge its use drawing on McMaster Libraries’ LibGuide and follow any course specific instructions.

• 11. Generative AI tools should not be used to provide grades (letter or numeric) for student work. Generative AI tools may be used to provide feedback on student work, provided the following conditions are met:
  • When generative AI tools are used to provide feedback of student work this use must be explicitly included in the course syllabus. 
  • Students should have the ability to opt-out of AI generated feedback.
  • AI generated feedback must be checked for accuracy and bias before being returned to the student.
  • Documentation of the AI tool used for feedback is required.
  • Instructors, or teaching assistants when directed, are responsible for feedback, however it is produced, to ensure appropriateness and accuracy.

If you would like to discuss options for feedback and evaluation, please reach out to the MacPherson Institute at mi@mcmaster.ca 

• 12. If instructors use generative AI in their teaching materials, they should explain in the course outline the extent to which generative AI has been, or will be, used and should clearly cite or label such uses in their course materials. Instructors should fact-check any generative AI produced materials and evaluate these materials for bias.
• 13. Course instructors have three options for directing teaching assistant use of generative AI. These directions should be given to teaching assistants in writing.
  • Permitting teaching assistants to use generative AI for any aspect of teaching assistant work, with the exception of evaluations that include a numeric or letter grade (see #11). TAs must inform the instructor of the intended use of generative AI, and receive approval, before implementation.
  • Requiring teaching assistants to use generative AI for specified teaching tasks as outlined in the hours of work form and with training provided.?In the instance of required use: As directed by the course instructor explicitly in the hours of work form, teaching assistants will use generative AI for the specific teaching tasks. Teaching assistant may create role-based accounts for such use. Course instructors will provide teaching assistants with the necessary training to use generative AI for the specified teaching purpose(s) with this training included in the hours of work. Teaching assistants will evaluate all teaching materials/formative feedback developed with generative AI for accuracy before use with students. Any planned use of generative AI by teaching assistants will be shared with students in the course outline.?
  • Prohibiting teaching assistants from using generative AI for teaching tasks.

Over the fall and winter of 2024 the AI Expert Panel on Teaching and Learning will continue to work. Some of the activities of the Expert Panel include:

• Additional course outline language to differentiate uses of generative AI and recommendations for consideration in policies related to course outlines
• Engaging educators and students in discussion about the use of generative AI for evaluating student work
• Developing resources to support departmental chairs
• Additional resources for educators to talk with students about generative AI
• Suggested language for acknowledgement of generative AI use (in addition to language for citation)

If you have suggestions for additional resources or discussions the AI Expert Panel on Teaching and Learning could support, please reach out to macgenai@mcmaster.ca

Students are not permitted to use generative AI in this course. In alignment with McMaster academic integrity policy, it “shall be an offence knowingly to …  submit academic work for assessment that was purchased or acquired from another source”.  This includes work created by generative AI tools.  Also state in the policy is the following, “Contract Cheating is the act of “outsourcing of student work to third parties” (Lancaster & Clarke, 2016, p. 639) with or without payment.” Using Generative AI tools is a form of contract cheating.  Charges of academic dishonesty will be brought forward to the Office of Academic Integrity. 

Example One 

Students may use generative AI in this course in accordance with the guidelines outlined for each assessment, and so long as the use of generative AI is referenced and cited following citation instructions given in the syllabus. Use of generative AI outside assessment guidelines or without citation will constitute academic dishonesty. It is the student’s responsibility to be clear on the limitations for use for each assessment and to be clear on the expectations for citation and reference and to do so appropriately.  

Example Two 

Students may use generative AI for [editing/translating/outlining/brainstorming/revising/etc] their work throughout the course so long as the use of generative AI is referenced and cited following citation instructions given in the syllabus. Use of generative AI outside the stated use of [editing/translating/outling/brainstorming/revising/etc] without citation will constitute academic dishonesty. It is the student’s responsibility to be clear on the limitations for use and to be clear on the expectations for citation and reference and to do so appropriately. 

Example Three 

Students may freely use generative AI in this course so long as the use of generative AI is referenced and cited following citation instructions given in the syllabus. Use of generative AI outside assessment guidelines or without citation will constitute academic dishonesty. It is the student’s responsibility to be clear on the expectations for citation and reference and to do so appropriately.  

Students may use generative AI throughout this course in whatever way enhances their learning; no special documentation or citation is required.