Graduate Education and Generative AI

Generative AI (GenAI) tools have rapidly propagated within the last several years, opening new opportunities for innovation in scholarly work, while also raising questions about the impact of these technologies on scholarship and student learning. This site provides preliminary guidance on the ethical use of GenAI tools in the context of graduate education at UBC and all aspects of graduate student scholarship, including admission applications, coursework, comprehensive exams, major projects, and theses/dissertations.

For the purposes of this site, Generative AI (GenAI) refers to tools that use machine learning tools and algorithms to produce new substantive text, analysis, charts, images, code, audio, or video.

Throughout academia and beyond, scholars are experimenting with uses of GenAI to advance innovation and productivity. It is inevitable that the use of these technologies will increasingly be integrated into how students approach their academic work. Facility with GenAI tools is rapidly becoming a valued capability in many career trajectories for those with advanced degrees, and graduate education is an ideal context for learning about its ethical uses.

While GenAI tools have significant potential as innovative academic resources, they also can pose risks to the academic mission of graduate education and to the integrity of student work. Developing high levels of personal and professional capability in a wide range of scholarly skills (e.g., academic research and writing, idea analysis and synthesis, etc.) and attributes (e.g., ethics, integrity, intellectual curiosity, bias identification, etc.) are the hallmarks of graduate education. The use of Generative AI should never compromise or replace a student’s development in these areas.

The use of GenAI in scholarly work is an academic decision, subject to principles of academic freedom and standards of scholarly integrity. Many academic disciplines, associations and publications have produced specific guidelines for use and citation of GenAI, which should be followed by students as appropriate.

The substantive (i.e. non-editorial) use by graduate students of Generative AI tools and outputs must be done with full transparency and with the approval of the academic(s) responsible for evaluating the given work in question — course instructor, research supervisor, or other advisor as appropriate. Students must proactively obtain such approvals, and document in writing any use of GenAI in their work. Students themselves are solely responsible for ensuring that their use of GenAI and all work produced is in alignment with program-level guidelines and university policies on student Academic Misconduct and Scholarly integrity.

If GenAI was used in the research described, the drafting of, and/or the preparation of a thesis or dissertation, the Preface must include a concise description of how it was used. If generative AI was not used in any way, a clear statement that generative AI was not used for any aspects of the work must be included.

This section will be updated occasionally to reflect the ongoing evolution of Generative AI tools and technologies and guidance on their acceptable uses in Graduate Education at UBC.

Additional guidance on the use of GenAI at UBC is also available.

Frequently Asked Questions

Gen AI

What Gen AI tools and outputs does this guidance refer to?

For the purposes of this guidance, Generative Artificial Intelligence (GenAI) is the use of machine learning tools and algorithms by graduate students to generate new outputs in various forms, including text, images, audio, video, and code. Such tools include ChatGPT, Gemini, GitHub Copilot, DALL-E and many others in this rapidly evolving space.

AI-powered transcription tools, and those that provide standard editorial support on student-produced writing (e.g Grammarly for grammar checking, writing organization/flow suggestions), are considered broadly acceptable and are not the focus of this guidance. However, there are still disciplinary norms, learning considerations, and ethical and intellectual property issues related to the use of editorial and transcription AI tools. Students should discuss the use of these tools with their instructor or supervisor.

What is UBC’s general stance on the use of Gen AI by graduate students?

The use of GenAI by UBC graduate students is an academic decision best made in collaboration between a student and the UBC academic personnel that are responsible for assigning/evaluating the relevant student work (e.g. course instructor, research supervisor, supervisory committee, other advisor). It is subject to principles of academic freedom and must be approved by the responsible academics as an appropriate method.

Any use of non-editorial GenAI in student academic work is also subject to UBC’s policies on scholarly integrity and academic misconduct. Students must always be transparent about and properly cite the uses and outputs of GenAI in their academic work, must not present GenAI-created content as their own work, and will remain entirely responsible for the accuracy of all material that was created or adapted using GenAI. Violation of any of these principles may be grounds for academic/scholarly misconduct proceedings.

Within these boundaries of academic approval, transparency, and scholarly integrity, graduate students are encouraged to learn, experiment and innovate with Generative AI in their academic work.

Graduate Students

May I use Gen AI tools to assist with completing my course assignments?

UBC’s guidelines on the use of GenAI in teaching and learning establish that course instructors have the authority to determine what, if any, uses of GenAI tools are allowable for students to assist in completing their course assignments. Each course syllabus should have a statement providing parameters on the use of GenAI. If in doubt, discuss with the course instructor.

Students are entirely responsible for ensuring their submitted academic work complies with UBC’s academic integrity expectations. Unauthorized or uncited use of Generative AI tools or AI-generated content may be deemed to constitute academic misconduct, including plagiarism.

May I use Gen AI tools to assist with the conduct of my research?

GenAI tools present both opportunities and risks for graduate student research. As with any proposed scholarly methodology, graduate students must ensure that their research supervisor and committee approve any intended use of GenAI tools in their research process. This should be openly discussed between student, supervisor and supervisory committee members, and then established and documented clearly at the point of formal approval of the student’s research proposal, or when initially considered thereafter.

Students are entirely responsible for ensuring their research and related outputs comply with UBC policies on academic misconduct and scholarly integrity and with any guidelines provided by relevant research funding agencies and publication outlets. Ensuring the accuracy and appropriateness of generative AI results used to support scholarly activity is the user’s responsibility.

May I use Gen AI for coding related to my research?

The use of GenAI for coding (e.g. producing code for running complex analysis or tasks) is an academic decision, allowable with approvals from your research supervisor(s) or instructor(s), appropriate citation, and compliance with scholarly integrity practices and policies.

What are the integrity risks of using Gen AI in research?

There are several integrity considerations regarding the use of GenAI tools and outputs in research, including:

Hallucinations. A known risk of the use of GenAI tools in scholarship is the creation (aka “hallucination”) of fictitious results and citations. All content generated by GenAI must be manually checked by students to validate its accuracy and veracity.

Computational reproducibility. Closed-source models, including many popular GenAI tools, pose significant challenges to the reproducibility as well as validity of research conducted using the models. As these models may be updated or removed at any time, they may limit the ability of work to be replicated and undermine the applicability of work that primarily relies on closed-source models. Lack of knowledge of the training sets used to create these tools also limits any conclusions that can be drawn about their capabilities[1] and should serve as a caution sign for researchers. Instead, researchers should consider open-source models, which should make code available to re-generate the AI models as well as documentation of training datasets and protocols used.

Introduced bias. Just as lack of documentation of training datasets used to create closed-source GenAI models may limit the reproducibility and generalizability of research that uses those models, this same lack of documentation may hide a potential risk of model being biased against different demographic groups.

Learn more from UBC Research & Innovation.

May I use Gen AI tools to assist with the writing and/or editing of my thesis or dissertation?

The thesis or dissertation is the essential expression of a research graduate student’s own original scholarship, reflecting their unique conceptualization, contextualization, and analysis of their research topic and findings. Any use of GenAI tools and outputs must only augment, not replace, this crucial intellectual contribution of the student-scholar.

With the approval of their supervisor and committee, graduate students may use GenAI tools to get started with drafting thesis/dissertation content.,

Students can normally assume that the use of AI tools for editorial (i.e. grammar, flow) and transcription purposes are broadly acceptable. However, there are disciplinary norms, learning considerations, and ethical and intellectual property issues related to the use of editorial AI tools. Students should discuss the use of editorial and transcription tools with their supervisor.

Literal outputs of GenAI (partial or total) must not be included in the thesis or dissertation or emergent research publications without citations that acknowledge and make clear that it was not created by the student and describe fully how GenAI was used. This applies to both drafts submitted to faculty advisors for review, and final versions. Otherwise, the action would constitute plagiarism and be addressed according to current policies on academic misconduct.

Theses and dissertations must include a statement on the use of GenAI (see Preface section) in the research and writing process, if used.

Use of GenAI in the thesis or dissertation writing process that is not approved by the student’s supervisory committee or graduate program (even if appropriately cited) may legitimately result in changes being required to the work, the student’s work not being approved for defense, or other actions in accordance with current policies on academic misconduct.

May I use Gen AI to generate or alter images used in my thesis?

The use of GenAI to generate or alter images is an academic decision, allowable with approvals from your research supervisor, appropriate citation, and compliance with scholarly integrity practices and policies.

Particular care should be taken to ensure that research data itself is not altered through the use of GenAI when producing visual representations of that data.

How should I cite the use of Gen AI tools in my submitted work?

Use of GenAI tools and outputs in your work must always be cited, with the exception of standard editorial uses (grammar checking, organization, flow) on student-generated work. This applies to both drafts submitted for review, and final products.

You must make it clear which sections or passages were created by or with the assistance of GenAI, and identify the extent of the use. If you use material that was fully AI-generated, then it must be quoted and cited as you would the work of any other content creator. If you used GenAI to develop a first draft and then paraphrased, modified, or added to the result, then you must state that. If you used GenAI to assist with searching for sources or ideas, that must be clearly stated.

Consult with your instructor, advisor, or supervisor on how to do this in each situation, as disciplinary and stylistic differences exist. Most academic journals also have requirements about use and citation of GenAI.

Visit UBC Library for more information on citation requirements and styles.

How can I work with faculty to determine appropriate use of Gen AI tools in my academic work?

Decisions about use of GenAI in academic work should come through discussion between student and supervisor or instructor that results in a common understanding and agreement on how any use of GenAI supports the learning, development and scholarship objectives of the student and the graduate program.

Here are some topics and conversational prompts for students and faculty to explore together:

Graduate program learning objectives. What are the stated learning objectives of the graduate program? How might a student’s use of GenAI support or detract from those objectives? Are there any program-specific GenAI guidelines we need to follow?
Student objectives. What are the student’s own unique learning objectives, scholarship goals and career interests? How might the student’s use of GenAI support or detract from those objectives and goals?
Planning GenAI use. What does the student envision for how GenAI could contribute to their development as a scholar, and to their overall scholarship? What concrete work would it be capable of effectively and ethically doing for the scholarly project? What tools are proposed and why?
Ethics and anti-bias. What concerns (e.g. privacy, copyright, data sovereignty/intellectual property, biased outputs) might be at play in the specifically proposed work/use of GenAI?
Learning tradeoffs. Of those tasks that GenAI would potentially be able to take off the student’s plate, what might the student not have the opportunity to practice/learn? What value do those things have toward becoming a practitioner in the field/student’s professional goals?
Disciplinary context. What is each person’s understanding of how the use of GenAI fits within both the conventions and the vanguard of the discipline? How would the student’s adoption of these tools position them within the field? Are there reputational risks involved with using GenAI?
Impact of scholarship. In what ways might GenAI enhance the reach or impact of the student’s scholarship and capacity to contribute to the field?
Collaborators. Has there been discussion and agreement with all collaborators about how GenAI can/will be used in the work?
GenAI dependency. In what ways might the use of GenAI in this work contribute to building a dependency on GenAI to do any work in the field? What could be the implications of that potential dependency for the student and for the field overall?
Transparency in GenAI use. How comfortable would the student be in publicly sharing all of the specific ways and areas they propose to use GenAI, especially with a future employer? If uncomfortable with any of this transparency, why is that? What would be the risks/drawbacks of NOT being transparent about its use?
Documentation/citation. How do we document the use of GenAI (e.g., keeping track of prompts, what software/tools were used and when)? What is the appropriate style for citation? If the work might be published, what guidelines are in place from the publisher?
Impacts of contributing to GenAI tools/models. Is any of the research data sensitive in a way that might present a bias, or a conflict, if it was incorporated into future large language models?

Once a student and their instructor/supervisor have decided on parameters and approach for using GenAI, this should be recorded in writing, to maintain clarity of common understanding, and revisited as needed.

Download prompts as PDF

How can I ensure my use of Gen AI in my work complies with UBC policy on Academic/Scholarly integrity?

Integrity is the foundation on which the credibility and potential contribution of all academic research relies. Learn more about this essential attribute and related UBC policies.

Taking these steps will help ensure you maintain scholarly integrity in any use of GenAI in your work:

Discuss and confirm with your instructor, advisor, supervisor, and/or graduate program what use of GenAI is allowable in your context and follow your agreement with care. Course syllabi should include this information.
Ensure any collaborators on your work are aware of and agree to any use of GenAI tools or outputs.
Consult with your instructor, advisor, or supervisor on how you are expected to cite your use of GenAI.
Clearly identify your use of GenAI by footnoting, quoting, or explaining. How you do this will depend on whether you’re using it in an assignment, examination, final project, or thesis.

Make sure you understand what constitutes academic or scholarly misconduct and how GenAI fits in. Continually test your planned approach against integrity standards.

What intellectual property (IP) considerations should I be aware of with Gen AI?

Many GenAI tools will use the prompts you use, or other data you enter, to further train the algorithm/model it uses. If so, you may be handing over the intellectual property (IP) associated with the entered data to the model and/or future models generated from the same database. This issue is important to consider not only when using GenAI tools to build on your own scholarly output, but also if you prompt the model using content produced by others (e.g course syllabus, teaching materials, books, papers, or work of your collaborators). Entering copyrighted information into a tool may constitute copyright infringement.

Furthermore, the data that has been used to train a GenAI model may have been used without permission, and it may be difficult for you to discern the sources of such data. As such, you may be implicated in using others’ IP without permission.

Material that is generated by the model may also be subject to intellectual property constraints associated with the terms of service of the model, which could impact your ability to use this in further work.

Finally, there may be restrictions on where you can publish or use work generated by GenAI tools, which could be detailed in the author publication guidelines for a scholarly journal or publisher.

It is advised that users carefully review the Terms of Service for any tool they are considering using, to understand how their own IP will be used by the tool, and any constraints on the use of content generated by the tool.

How can I ensure that any use of Gen AI in my work is ethical, and does not perpetuate harmful bias?

Understand potential bias. Generative AI tools are powerful and can be helpful in providing basic contextual information, summarizing, or outlining ideas. However, they lack “intelligence” to differentiate truthful information from false and can provide significantly biased outputs. GenAI tools often enforce a tone of neutrality which in itself is biasing and can mislead. These tools are also often trained on data sets that reflect social and cultural biases and can reproduce systemic inequalities in outputs. Even when a tool has protections in place to identify discriminatory content, these systems can be imperfect, and outputs can be problematic and reinforce systemic inequities. When using GenAI tools, take steps to understand sources of training data and assess outputs to identify and document known biases. Such reflections should be included in your methodology explanations.

Evaluate for incomplete or incorrect data. When using GenAI tools for a research question keep in mind that the tool is only as good as the data that it was trained on. Historical or non-digital data is generally missing from the training model which leads to incomplete, incorrect, or inconclusive outputs. GenAI tools that are trained on an unknown and uncitable dataset are unlikely to be useable in an ethical or comprehensive way for research. For some research questions, consider the use of locally run GenAI tools with controlled datasets as opposed to large open tools.

Re-think informed consent. If you intend to input data from research participants into GenAI tools, you will require ethics approval for this, and participants would need to be trained in basic digital literacy on the use and function of GenAI tools to be able to provide informed consent. For example, participants must understand whether their data can be removed from systems, whether their identity may be identifiable in a system, and what kinds of outputs may be created from the data they provide. Without sufficient understanding of core principles in the use of GenAI, informed consent cannot be provided, and the data cannot be ethically used.

When using GenAI to generate text or images, be aware that the outputs are created from AI training data, which may not have been provided with informed consent.

Seek expert guidance. Instructors and/or research supervisors may also provide additional ethical considerations, beyond those provided by UBC policies, that could restrict specific uses of GenAI in classwork or research work. Direct consultation with instructors and/or research supervisors about ethical uses of GenAI is encouraged. UBC Librarians can also support students with digital literacy, including best practices in use of GenAI for research and how to critically evaluate false or misleading outputs from GenAI.

Consider the environmental costs of AI. There are multiple direct and indirect ways in which the use of AI may contribute to environmental harm[1]. Many GenAI tools rely on huge server clusters that consume massive quantities of energy, producing significant carbon emissions. Many also rely on large supplies of water for cooling.[2] GenAI may also be used to benefit and advance industries that contribute to environmental harm. On the other end, AI technologies also have potential to help environmental sustainability and repair. Responsible citizen-scholars should become aware of these factors as they consider their use of GenAI.

[1] https://www.scientificamerican.com/article/ais-climate-impact-goes-beyon…

[2] https://oecd.ai/en/wonk/how-much-water-does-ai-consume

What considerations regarding Indigenous data sovereignty should graduate students be aware of?

All graduate students at UBC can benefit from learning about Indigenous data sovereignty, regardless of whether they are engaged in scholarship which directly involves Indigenous people and concerns.

UBC’s Indigenous Research Support Initiative and UBC’s Xwi7xwa Library provide specialized guidance on many considerations of scholarship and data sovereignty and Indigenous communities.

Some primary considerations related to GenAI include:

Indigenous data ownership and consent. Indigenous community members have a right to the ownership of their data and a right to privacy. Indigenous data sovereignty requires participatory research, informed consent in the research process and use of data, and data collection processes that align with community permissions, protocols and governance.

As well, provisions must be made for what happens when a community withdraws permission for data. A GenAI tool must make it possible to opt out of the use of data and there must be a clear process for the removal of the data from the GenAI tool.

GenAI tools also enable manipulation of data and cultural outputs that may undo protections put into place during an ethical research process such as the de-identification of data.

Some Indigenous data and its use, including in GenAI tools, is restricted to the community that it came from. For example, a locally run large language model may be appropriate to use if developed in partnership with the community and restricted to their use. An open GenAI tool that does not have restrictions on use would not be appropriate to use.

Cultural context of knowledge. The use of a black box system such as a GenAI tool for cultural knowledge is problematic because of the separation of data from its context. The loss of traditional teachings that surround a design or output can change the meaning of an output. For example, creating an image of a design in the style of a community removes cultural context, community meanings, and traditional teachings from the output. It is important to be able to articulate traditional context and protect traditions behind this type of output.

[1] Suresh, H., & Guttag, J. (2021, October). A framework for understanding sources of harm throughout the machine learning life cycle. In Proceedings of the 1st ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (pp. 1-9).