Principles
Principles
Generative Artificial Intelligence Task Force Report
Adaptation to AI at TMU will take many forms, depending on the function of any given unit or office at the university. Academic departments/schools vary greatly by discipline. Administrative offices can serve vastly different functions. As such, the way they adapt to and use AI will look different across the university. The task force, therefore, recommends that TMU adopt common principles for AI adaptation and use that will guide the development of local guidelines appropriate to the needs and functions of the diverse academic disciplines and administrative units of TMU. Those principles should reflect broader institutional values and priorities while allowing local flexibility.
These prospective principles build upon existing principles and guidelines at the university including the Generative Artificial Intelligence in Learning and Teaching at TMU: Principles and Guidelines and YSGPS Guidance on the Use of Generative Artificial Intelligence (GAI) in Graduate Studies.
Based on the work of the task force, its working groups, and its consultations with the community, we recommend the following institutional principles for adaptation to AI:
As an institution of higher education we are committed to the development of human potential. Accordingly, we prioritize uses of AI that amplify human intellect and agency over those that merely optimize for transactional efficiency. Ultimately, we value augmentation over automation, such that AI serves to unlock human potential and does not ultimately displace learning and knowledge creation.
In practice:
❌ Relying on generative AI to aggregate and summarize research papers without researchers or students reading and critically analyzing the source material
✅ Generating alt-text for images in presentations to improve accessibility
TMU supports and encourages responsible innovation with and continuous evaluation of the application of generative AI technologies and the potential benefits and opportunities. The university will lead in the development and application of these technologies across a wide range of disciplines. It will also lead in the critical evaluation and critique of AI and its effects on society.
In practice:
❌ Licensing proprietary AI solutions under terms that prohibit or restrict the university’s ability to independently audit the system for accuracy, algorithmic bias, or data privacy
✅ Create policies that enable researchers from different fields to work together using AI, while giving them access to a variety of AI systems that are secure and used responsibly
Inclusion must be foundational in AI governance and application at TMU. Our emphasis should be on fostering equitable opportunities for students, researchers and employees to engage in learning, scholarly research and creative activities, and experimentation. The principle of equity should extend to access to training on AI fluencies as well. At the same time, we must recognize that AI systems can replicate and reinforce existing societal biases and systemic inequities embedded within their training data and design. Training for all university members should raise critical awareness of these algorithmic risks and promote strategies for their mitigation. TMU will serve as a vital hub for scholars, creators, and students to study, critique, and work to address such inequities, ensuring that this technology ultimately serves the public good and supports human flourishing.
In practice:
❌ A course instructor expects students to use sophisticated AI tools in their course work but does not address access or training.
✅ An academic program ladders AI fluencies into the curriculum ensuring that no graduate of the program gets left behind
Generative AI technologies must be used ethically and responsibly at TMU. AI has inherent limitations and biases, including those that reinforce dominant perspectives. When used uncritically, AI can undermine learning, compromise academic and research integrity, and disrupt the operations and mission of the university. These risks must be considered carefully when evaluating AI tools for adoption or continued use.
In practice:
❌ Relying on AI tools to scan resumes, evaluate cover letters, and automatically rank or filter out applicants before a human recruiter ever sees them
✅ Treating AI outputs as potentially biased and requiring manual verification, especially regarding marginalized histories or niche topics.
TMU community members should disclose the use of AI technology whether in academic work, research, or administrative purposes. Proper and clear disclosure ensures clarity for all about the role of AI in generating content or insights, and facilitates verification and critical reflection.
In practice:
❌ Using an AI model to clean, sort, or identify patterns in a massive scientific dataset but omitting this methodology when publishing the research
✅ Providing clear guidelines in a course outline for students on how to disclose AI use in assignments and sharing information about the instructor’s own use of AI
Human judgement and intelligence must remain at the centre of our work. Authorship is ownership. Individual users are accountable for any outputs generated using AI. Because these technologies are based on probabilistic statistical models, their outputs are unpredictable and cannot be replicated. As such, it is critical to review and revise all outputs for accuracy, safety, and security.
In practice:
❌ Treating AI-generated data or code as stable and reproducible, leading to systematic failures when the probabilistic model yields entirely different outputs upon subsequent attempts
✅ Human review of AI-generated meeting notes prior to distribution to team members
TMU community members must exercise caution in using sensitive, private, or proprietary data with generative AI technologies. All uses must be in compliance with existing data privacy and security policies at TMU. Generative AI use must also comply with all legislated requirements and regulations including those concerning privacy, cybersecurity, copyright, employment, and human rights. Because deep learning models rely upon data collection from users, TMU community members should be mindful of the data they input into AI tools and, where appropriate, use university-approved applications.
In practice:
❌ Uploading a spreadsheet containing student names, student ID numbers, emails, and final grades into a non-university-approved AI platform to analyze academic performance trends
✅ Analyzing non-proprietary research data using institutionally supported tools with data privacy protection of the institutional license agreement