This guidance establishes minimum, division-wide standards for the use of generative AI in coursework, examinations, research and scholarly writing. A faculty committee will review this guidance and may revise it. Individual programs, courses, and mentors or committees may adopt additional or more specific requirements, provided they meet or exceed the standards set out below.
Purpose and Scope
This guidance sets expectations for the responsible use of generative AI (GAI) by individuals within BSD PhD and Master's programs across coursework, examinations, and research and scholarly writing, including assignments, examinations, theses, dissertations, and capstone/independent projects. It applies to all students, faculty, instructors, and staff within these programs, and supplements, without replacing, the University of Chicago Generative AI Guidance and the Policy on Academic Honesty and Plagiarism.
Definitions
Generative AI (GAI) means any tool or feature that drafts new text, images, code, or audio, or makes substantive automated edits, in response to a prompt (for example ChatGPT, Copilot, Claude, or Gemini). Basic spelling or grammar checkers are not GAI under this guidance.
Data Protection and Approved Tools
Use University-approved enterprise tools for University data; use only IRB- or data-agreement-approved tools (ex. HIPAA Claude) for human-subjects or other regulated research data. Never enter personally identifiable information, FERPA-protected student data, or unpublished or restricted research data into a GAI tool not approved for that use. Public tools may be used only with non-sensitive, public information.
| Data tier | What it covers (examples) | GAI permitted | Never permitted |
| Public | Published literature, de-identified concepts, general scientific knowledge | Any tool, including public/consumer AI; user verifies outputs | No restriction (citation rules still apply) |
| Internal | Non-public University information not otherwise regulated (internal communications, draft materials, unpublished research) | Approved enterprise AI tools only (institutional account; Approved enterprise tools do not permit vendor use of University data to train or improve their models. | Public or consumer AI tools |
| Restricted | Human subjects’ data (IRB-governed), FERPA records, PII, exam content, regulated research, data-use agreements | Only a tool specifically approved for that data category (HIPAA Claude). | Public/consumer AI tools; any tool not reviewed for that category |
When in doubt, treat data as the higher tier and confirm before entering it.
Academic Integrity and Assessment
As a general rule, students should not rely on generative AI for graded coursework or examinations, unless their instructor, program, or committee has approved that use in advance. Judicious use of AI tools for independent study, such as testing comprehension or working through practice problems, remains an acceptable and encouraged use. Ultimately, students remain responsible for everything they submit, whether an error, an invented reference, or a flawed conclusion originated with them or with the tool they used.
Thesis, Dissertation, and Independent Project Writing
Use of AI in a thesis, dissertation, capstone project, or independent project is subject to the mentor's, graduate program’s or, where applicable, committee's discretion, who may permit, limit, or prohibit it. Regardless of what is approved, the work's core research, analysis, and conclusions must remain the student's own; AI may support that contribution but may not substitute for it. Absent specific guidance, AI use beyond basic language editing requires prior authorization with a clearly defined scope, and any approved use must be disclosed in the final document and its role. Undisclosed or unauthorized AI use in a thesis, dissertation, or independent project is treated as a violation of academic integrity, consistent with the standards described above.
Research and Scholarship
AI tools may not be credited as authors on any BSD-affiliated publication or scholarly product. Where AI contributes to a manuscript, presentation, or other research output, that contribution must be disclosed in a manner consistent with the requirements of the target journal or venue. If a human-subjects research protocol also involves the use of AI tools, that use must be disclosed to the IRB.
Review and Questions
A BSD faculty committee will be convened to review and revise this guidance, which may change substantially as a result of that review. Suspected violations are addressed through existing University and divisional academic integrity procedures. Questions may be directed to the Dean of Students, Chaevia Clendinen.
Example Syllabus Statements
Instructors are encouraged to include one of the following statements in their syllabus, or to adapt one of their own:
No use permitted
“This course does not permit the use of generative AI at any stage of coursework, including assignments, exams, and other graded work. All submitted work should reflect the student's own independent effort.”
Use permitted only with prior authorization
“Students may not use generative AI in this course unless the instructor has approved a specific use [such as to brainstorm essay topics, generate practice quiz questions, debug code,…] in advance.
If no such approval has been granted, coursework, exams, and assignments should be completed without AI assistance of any kind. This includes, but is not limited to, [drafting or rewriting essays, summarizing assigned readings, solving problem sets, writing or correcting code, generating citations or bibliographies,…]. When in doubt, students should ask the instructor before proceeding.”
Use permitted with defined parameters, with disclosure
“In this course, AI may be used as a tool for [ gathering background information, defining terms, getting assistance with coding syntax,…] but it may not be used for [drafting of text, generating code, obtaining problem solutions…]. No entry of assignment problems into an AI prompt is allowed and conversely copying or paraphrasing of text from AI outputs into assignment responses is not allowed.”