AI Authorship for Graduate Education

Introduction

Members of the AI Use in Graduate Education Working Group identified precedents from 18 peer institutions as part of a broader effort to inform recommendations on the responsible use of generative AI in graduate education at The University of Texas at Austin. Each precedent includes a link to the policy reviewed, an assessment of the approach within the policy, an overview of the guidance provided, highlights from the precedents to consider, and an exemplary quote from each precedent.

Contents
Columbia University	UT Dallas
University of Minnesota	Illinois State University
Virginia Commonwealth University	UNC Chapel Hill
Texas A&M University	University of Maryland
University of Georgia	University of Florida
University of Colorado Anschutz	Michigan State University
University of Michigan	University of Washington
Duke University	University of Illinois
Georgia Tech	University of Louisiana: Lafayette
	Appendix

Columbia University

provost.columbia.edu → Generative AI Policy

Approach

The Office of the Senior Vice Provost maintains holistic guidance specifying how the University community, including graduate students and researchers, should adopt generative AI.

Guidance Overview

Columbia’s guidance is titled “Generative AI Policy.” This document combines formal policy language, processes, and best practices, organized into the following sections: Purpose; Policy; University Approval Applications; Generative AI and Academic Integrity (with subsections for faculty and students); Generative AI and Research; Related University Policies; Related University Resources; and Questions. The guidelines were produced by a working group of staff, students, and faculty.

Highlights to Consider

Columbia includes an open comment process whereby community members may provide input, questions, or comments on the policy. The policy includes details on the limitations of generative AI use and notes that it may be adapted in alignment with changes in generative AI technologies. While the guidance addresses generative AI use in research and classroom instruction, it does not differentiate or provide direct guidance on curricular activities such as theses, dissertations, graduate student teaching, or graduate student research.

---

UT Dallas

policy.utdallas.edu → Generative AI Use in Academic Work

Approach

Formal governance: UT Dallas has an official policy specific to generative AI use in academic work within the Handbook of Operating Procedures.

Guidance Overview

The HOP is titled “Generative AI Use in Academic Work.” It is organized into the following sections: Policy Statement; Student Responsibilities and Role of the Faculty Member; Academic Integrity; and Policies and Statutes. Subsections under Responsibilities and Role of the Faculty Member cover Permitted Use, Partial Use, and Prohibited Use. The policy refers to other related policies including the UT Dallas Acceptable Use Policy, Student Code of Conduct, and FERPA policy.

Highlights to Consider

UT Dallas emphasizes that faculty have both the freedom and the responsibility to determine how AI may be used in academic work, including graduate-level work such as theses and dissertations. The policy also requires faculty to communicate AI policy before academic work begins for a given semester. This balance is worth highlighting as it provides faculty autonomy and requires the communication of requirements to help students navigate ambiguity across courses or with different faculty. The policy acknowledges that students may need “workforce-ready” AI skills and requires students to validate and verify AI output. Misuse by students, faculty, or staff may result in disciplinary action. The policy does not address research specifically.

---

University of Minnesota

grad.umn.edu → Guidance on Generative AI

Approach

The graduate school has chosen not to make specific recommendations about generative AI use in graduate coursework. They instead have curated a list of resources and guidance.

Guidance Overview

In addition to providing links to the University’s evolving IT policy, the graduate school has offered guidance to departments and programs on the following:

• Defer to the state of the art according to the discipline’s professional societies.
• Prioritize professional development and ongoing discovery of AI tools for education.
• Determine appropriate use for in-program and with future professional environments in mind.
• Define prohibited generative AI use (from admissions through academic work) and the consequences of such use for inclusion in course syllabi.
• Update handbooks and communications with information about prohibited use, evaluating output, asking for clarification about department policy, and policies around citing AI.

Highlights to Consider

The graduate school is deferring to graduate programs to make discipline-appropriate decisions about generative AI use among students, while setting some requirements about what the policies should cover and how to document them.

---

Illinois State University

grad.illinoisstate.edu → Ethical AI

Approach

This is a non-specific set of guidelines meant to outline the Graduate School’s position. It makes reference to University Policy 1.8 (Integrity in Research, Scholarly, or Creative Activities) and has one stated requirement: disclosure.

Guidance Overview

This is a restrained approach grounded in transparency. Students are required to disclose the use of generative AI to their committee and school as well as to their audience. Students are required to cite the use of any synthetic media according to the standards of their discipline.

Highlights to Consider

This guidance is specific to the use of AI in “graduate milestones.” It notes that all use of generative AI in the thesis/dissertation/capstone/comprehensive exam process must be disclosed to the student’s committee, but does not define or provide a framework for how that should be done.

---

Virginia Commonwealth University

faculty.provost.vcu.edu → Gen AI Guidelines for Students

Approach

The Center for Teaching and Learning at VCU maintains a set of generative AI guidelines for students, with subsections specifically addressing graduate students, students conducting research, and those writing theses and dissertations.

Guidance Overview

The guidance for graduate students covers two main areas: contribution of knowledge to a field and reinforcing the ethical conduct of researchers. The first section requires graduate students to:

• Maintain a critical perspective when assessing outputs/results;
• Assess and validate information;
• Acknowledge Gen AI use;
• Experiment to optimize results from prompts;
• Remain aware of all concerns regarding intellectual property, privacy, and confidentiality;
• Avoid significant assistance by Gen AI (as opposed to editing assistance);
• Follow publishing policies.

Highlights to Consider

The guidelines are more specific and prescriptive than others, choosing to remind students about the line between editorial and substantial assistance of AI tools and recommending that students optimize and refine their prompts. The guidelines also include a section on the limitations and ethical considerations related to generative AI.

---

UNC Chapel Hill

handbook.unc.edu → Graduate School Handbook

Approach

This section of the UNC Graduate School Handbook is mostly an aggregation of various University resources, with some specific notes from the graduate school. The areas of guidance are split into students, teachers, and researchers, all linked back to a central page maintained by the provost.

Guidance Overview

The guidance reinforces existing university policy with call-outs about how they intersect with graduate work. Students are reminded that they take full responsibility for the output of AI in their work and research. Individual graduate programs are encouraged to establish their own guidelines for the use of generative AI for theses and comprehensive exams in alignment with the university honor code. Teaching assistants are subject to the same guidance as other campus instructors. When conducting research, graduate students are encouraged to have open conversations with their supervisors and committees about how AI will be used and about the policies of relevant journals and agencies.

Highlights to Consider

The guidance for TAs specifically mentions that the supervising instructor should be clear about whether they “support the TA using AI in their own grading/teaching work.” There is also an extensive list of discussion topics related to the use of generative AI in research for students and advisors.

---

Texas A&M University

grad.tamu.edu → Guidance for AI in Relation to Theses (PDF)

Approach

The issued guidance is a PDF not marked as official policy. The document directs the reader to other relevant policies (Resources on Generative AI in Research, Responsible Conduct of Research, Thesis and Dissertation Guidelines).

Guidance Overview

The document’s stated purpose is to advise awareness and explicitly does not set any policies or restrictions so as not to discourage AI use. Instead, it acknowledges the scope of impact of AI (teaching, learning, research) and relates these back to guidance issued by the VP of Research and the Center for Teaching Excellence.

Highlights to Consider

This document frames the use of generative AI for theses, dissertations, and other records of study as in line with established guidelines. Students should produce work with “independent professional effort” that is overseen by their advisors and committees. The decisions about appropriate use should be negotiated through discussions between those individuals.

---

University of Maryland

ai.umd.edu → Guidelines for Use

Approach

The policy was issued by a committee organized by the president and is titled “Guidelines for Use,” addressed to all Faculty, Staff, and Students. This document does not explicitly mention graduate students, but includes guiding principles and guidance for students, TAs, and researchers.

Guidance Overview

This guidance covers the use of GenAI tools in any aspect of university business and is broadly defined by a set of guiding principles: human oversight, access, privacy, transparency, and accountability. Use policy for teaching and learning is encouraged at the course level, with instructors setting their own boundaries and communicating them to students. Theses, dissertations, and comprehensive exams are explicitly called out as research activities and subject to the guidance for research and scholarship. Guidance for researchers stresses attribution, the policies of other institutions (journals, agencies, professional societies), data governance and protections, and intellectual property rights.

Highlights to Consider

The set of guiding principles for campus-wide use provides good grounding for the rest of the recommendations. Having complete and comprehensive guidelines has the benefit of providing highly contextualized information in one place; however, this means that the guidance can be somewhat general.

---

University of Georgia

grad.uga.edu → Generative AI in Theses and Dissertations

Approach

This is a policy published by the graduate school in January 2024. It issues guidance for theses and dissertations and restates policies related to producing independent or group coursework.

Guidance Overview

This guidance explicitly prohibits the use of generative AI for theses and dissertations unless authorized by the student’s advisory committee. The guidance related to coursework does not mention generative AI at all and categorizes technology as a source of unauthorized help.

Highlights to Consider

This is the shortest and most prohibitive generative AI policy reviewed, but it still allows for discretion by faculty advisors.

---

University of Florida

ai.ufl.edu → Guidance for Researchers

Approach

The University of Florida maintains a central resource in both document and web formats that provides guidance on best practices for instructors, graduate students, students, researchers, and HR professionals. While there are some firm policies related to human subjects research and IRB, the bulk of the guidance is in the form of best practices rather than formal policy. A committee of 16 individuals authored the guidance.

Guidance Overview

Florida provides comprehensive guidelines covering: research on AI; using AI in research; human subjects research; intellectual property; collaboration and partnerships; education and training; institutional resources; reporting requirements; and monitoring and evaluation. In all sections except Human Subjects Research and Reporting Requirements, the guidance is in the form of best practices using “should” rather than “must.” Florida also has a more comprehensive document titled “GenAIBestPractices” with detailed recommendations for instructors of graduate education specifically (see page 12).

Highlights to Consider

This guide clearly documents limitations, problems, and hazards of using AI and emphasizes the importance of AI literacy. It also has specific guidance for instructors of “graduate education” related to transparency, responsibility, learning, and integrity. While centralizing all best practices in one place is beneficial, the document may be difficult to navigate for those seeking guidance specific to graduate students. The web-facing version is more navigable.

---

University of Colorado Anschutz

graduateschool.cuanschutz.edu → Generative AI Guidance

Approach

Provides graduate school–specific guidance in the context of the medical campus. Includes guidelines and tips rather than policies. Guidance was developed by the Medical Campus, revised by the Graduate School of CU, and is posted on a navigable website.

Guidance Overview

Includes seven principles for graduate students covering: data privacy; limitations of AI; biases of AI; dangers of using AI to replace evidence-based practices; the importance of critical judgment of AI output; the need for transparency and honesty about use; and warnings about academic integrity. Additional tips are provided from Risk and Compliance as well as the Office of Information Technology.

Highlights to Consider

This document may not have been updated since 2023. It provides general principles but not specific policies related to research, graduate study, or graduate teaching. A section on additional resources is included.

---

Michigan State University

ai.msu.edu → Guidelines

Approach

This set of guidelines covers most uses on campus and supersedes all previous guidance about AI adoption.

Guidance Overview

The major components of this document are: an overview with a bulleted summary of stances and a list of ethical principles; permissible uses for teaching and learning, research, scholarship and artistic endeavors, and administrative work; and implementation, procurement, and resources. None of this content speaks directly to graduate work specifically, except for a call-out in the teaching and learning section for graduate TAs. Students are encouraged to follow their instructors’ guidelines.

Highlights to Consider

This is an expansive but generic set of guidelines that returns often to the issue of data security and governance (mostly for research) and the application of critical thinking in the process of learning.

---

University of Michigan (Graduate School)

rackham.umich.edu → GenAI Guidance for Applicants

Approach

For the purposes of this precedent document, focus has been placed on guidance from the graduate school specific to applicants.

Guidance Overview

This policy lays out what students can do with generative AI while preparing their applications (research schools and programs, check grammar on essays, translate words and phrases) and what they cannot do (generate outlines or drafts of essays, alter their writing voice, translate large portions of their work).

Highlights to Consider

This is a direct approach with concrete advice for the various scenarios an applicant might face.

---

University of Washington

grad.uw.edu → Effective and Responsible Use of AI in Research

Approach

As noted at the top of the page, this guidance is lifted directly from Georgia Tech. UW adds references to specific institutional policies and expands on ethical questions about equity. For a full overview, see the Georgia Tech entry below.

---

Duke University

myresearchpath.duke.edu → Using Generative AI Tools in Research

Approach

Duke provides specific guidance for researchers emphasizing the importance of responsible adoption. This is a set of do’s and don’ts for using generative AI tools in research and a list of responsible principles, rather than a formal policy document. It is hosted by Duke’s research office and was drafted by a Digital Humanities librarian with feedback from the Director of Research Integrity, Copyright and Policy Librarian, STEM Librarian, and a faculty member.

Guidance Overview

This guidance includes a summary, a list of responsible use Do’s and Don’ts, and nine Principles of Responsible Conduct of Research with AI, framed as best practices. Each Principle opens with a guiding question, for example: “Human researchers are responsible for the outputs we generate with AI tools. Are you prepared to stand behind your AI-assisted research?” The guidance does not differentiate or provide specific details for graduate students, but does provide additional resources, including resources on AI and research from the NIH and NSF.

Highlights to Consider

This resource was researched and written by a librarian and reviewed by key stakeholders rather than through a committee. The reference section was helpful for grounding the curated best practices. The guidance is comprehensive, covering important considerations such as environmental impact and labor concerns related to AI. The use of guiding questions to lead off each Principle was unique in comparison with the other precedents. This document did not cover graduate student research.

---

University of Illinois

genai.illinois.edu → Best Practices for Generative AI in Research

Approach

The document is the output of a working group suggesting best practices (notably not guidance or policy) for using generative AI in research. There are no specific mentions of graduate education products like theses or dissertations.

Guidance Overview

The best practices are organized into the following categories:

• Evaluating and choosing AI tools wisely
• Authorship and the context for dissemination
• Balancing GenAI use with your contributions as an author
• Disclosing AI use
• Prompt engineering
• Organizational best practices

Highlights to Consider

This best practices document includes specific examples of publication guidance from three journals in different disciplines, enforcing the importance of seeking these rules when planning and conducting research.

---

Georgia Tech

grad.gatech.edu → Guidance for Effective and Responsible Use of AI in Research (PDF)

Approach

The document consists of practical questions about how generative AI can be used for research. It is an aid and guide rather than a set of rules or a list of discouraged behavior.

Guidance Overview

This is less of a policy and more of a collection of viewpoints about the role that generative AI can play in research and the experience of scholarship for a graduate student. The questions posed to the experts consulted for this work were also posed to ChatGPT, and the output is included in the appendix. In addition to talking more generally about the strengths and weaknesses of generative AI, there is an overview of how AI might be used ethically in the various stages of research and writing publications, with advice about how to use and evaluate output.

Highlights to Consider

There is a section with sample rules about authorship from various professional societies, journals, and funding agencies. This kind of practical and grounded advice demonstrates that the graduate school’s recommendations are aligned with outlets for scholarly work.

---

University of Louisiana: Lafayette

louisiana.edu → Guidelines for Use of GAI in Graduate Research (PDF)

Approach

UL Lafayette has comprehensive guidance from the Graduate School specific to using generative AI in graduate research. These guidelines were developed based on the foundation of previous guidance on Generative AI in the Classroom developed by their Academic Affairs Division, which covers undergraduate and graduate coursework. UL’s approach maintains a separate set of guidelines for graduate research and scholarship. The research guidance was created by the Graduate School and received endorsement by academic college deans and other stakeholders.

Guidance Overview

Guidance requires students to engage in “Full Disclosure,” specifying the tools employed, their purpose, and the scope of their application (e.g., for brainstorming, organization, data analysis, revision). Documentation and placement of disclosure must be prominently included in the manuscript, either in the introduction, methods section, or a dedicated GAI disclosure statement following the acknowledgements section. Appendices that document the use and include work product (e.g., text of prompts and queries, prompt evolution, post-generation editing) may also be required. An example of a disclosure statement is included for students. The guidance also indicates that graduate programs have been “charged” with establishing and maintaining requirements for capstone projects, theses, dissertations, and synthesis projects, and specifies prohibited practices such as “generating substantive sections of a manuscript without acknowledgment.”

Highlights to Consider

This guidance firmly requires “disclosure” of AI use. While transparency is a foundational principle, the term “disclosure” may imply something was done that must be “disclosed” (Schell acknowledging a discussion with Sharon Strover who helped advance her thinking on this). It is worth considering whether to use the term “disclosure” or instead use something more neutral such as “acknowledge” or “attribute.” The document focuses heavily on graduate research related to milestones but does not dive deeply into researcher responsibilities. Our interest in helping students engage with PIs is worth building out in ways that many of these precedent documents do not. This guidance does state: “Faculty Engagement: Graduate students must engage with their faculty, chairs, and committee members early in the research process to ensure mutual understanding and alignment on the acceptable use of GAI tools.” Adapting this language to specifically address the PI–student relationship is something UT should consider. Dean endorsement was a unique feature of this guidance. The working committee should consider whether graduate associate deans within the CSUs should review final recommendations through an open comment period.

---

Appendix

---

1. Generative AI in Higher Education: Seeing ChatGPT Through Universities’ Policies, Resources, and Guidelines
   • One precedent submitted was a research article by this title. This paper reviewed policies at the top 100 US Universities (according to US News 2024 Rankings) regarding the use of generative AI. The paper focuses primarily on use cases related to teaching and learning, but did analyze policies related to AI in research — specifically, acknowledging the use of AI-generated content in research (12.5% of the policies) and the confidentiality and security of data in research settings (26.9%). The paper indicated: “Professional research writing and publication in academic may need more guidelines from institutions” (p. 60). In addition, the paper emphasized that “more explicit policies and guidelines are necessary to raise researchers’ awareness of which information is considered sensitive and personal and to address the appropriate boundaries for using GenAI in research in higher education” (p. 11). This paper establishes a need for more clarity for use of AI in graduate research related to curricular milestones and academic research and writing.
2.  Anthropic Diligence Statement
   • Another precedent to consider is the Anthropic “Diligence Statement.” This statement is derived from prompting an AI tool to provide a summary explaining with transparency how you used AI to develop output or a product. It may be helpful to develop a prompt that students can use to develop a transparency statement for themselves.
3. How Instructors Regulate AI in College: Evidence from 31,000 Course Syllabi
   • This paper analyzes 31,000 syllabi at a large selective public research university with 50,000 students in Texas over a 5-year period. Some findings pertaining to graduate education are that graduate syllabi may be less likely to have a policy pertaining to AI. The paper defines three mechanisms through which AI affects student task practice, each prompting a different instructor response:
     • Task Displacement: AI performs tasks instead of students, eliminating practice opportunities and risking skill erosion. Instructors respond by restricting AI for these tasks.
     • Task Augmentation: AI supports student performance of existing tasks without displacing essential cognitive effort. Instructors respond by permitting or encouraging AI for these tasks.
     • Task Reinstatement: AI enables entirely new task types not previously feasible — such as prompt formulation, output verification, and evaluation of AI-generated content. Instructors respond by introducing new AI-based assignments.