AI policy

This page describes Chris Proctor's policy on the use of generative AI in PhD milestones. These policies apply to Chris's advisees and to all students on whose committees Chris serves. For reciprocal accountability, this page also includes Chris's own statement on how he uses AI in research, teaching, and service.

Policy for students

Acceptable use of AI in academic work is a contentious issue with a variety of defensible positions. Your advisor and your committee will define what constitutes acceptable AI use for your doctoral work. Chris's general view is that methodological innovation using AI is an important skill for emerging scholars; he has no blanket restrictions on student AI usage. However, as part of every milestone, students will submit an AI statement explaining their use of AI for agreement by their committee members. Any AI usage not described in the statement is prohibited, and would be considered a violation of academic integrity (and potentially also research misconduct). For Chris's advisees, academic integrity with respect to AI means adherence to an agreed AI policy. This is similar to how the IRB works: you describe what you plan to do, the plan is approved before the work begins, and changes must be approved before they take effect.

What an AI statement should contain

An AI statement should address how AI will be used in all aspects of research, including but not limited to:

• data collection and generation
• software development
• literature review
• analysis
• all phases of the writing process, including brainstorming, outlining, drafting, getting feedback, revising, editing, translation, and managing citations

Every AI statement must have a date. The statement should also describe how AI usage is (or will be) addressed in relevant IRB protocols. Explain where research data, especially participant data, will be processed, and whether consent materials tell participants that AI will be used to process their data. Committee approval of an AI statement does not substitute for IRB approval, nor for compliance with UB's policies on data risk classification and technology use.

AI statements for specific milestones

Research analysis exam

For the RAE, you do not need an AI statement because the policy is fixed: no AI (aside from incidental use, defined below) may be used, for reading, research, writing, or any other aspect of the exam. After you submit your RAE, Chris will meet with you for a discussion of your RAE. This discussion is not graded, but you are expected to be able to discuss your essay in detail.

Research component

For the RC, the AI statement should be written during committee formation, and any updates to the statement must be approved by the committee. The AI statement will be part of the RC submission. For students who are not Chris's advisees who want Chris on their RC committee, Chris will ask for an AI statement before accepting, and will not join committees until Chris agrees to an AI statement.

Dissertation proposal

For the dissertation proposal, similarly, the AI statement should be written during committee formation, to ensure that committee members support your proposed AI usage. The AI statement may be a separate section of the methods chapter, or it may be a separate document.

Dissertation defense

All versions of the AI statement governing your dissertation research (including the version from your dissertation proposal) must be submitted to the committee with the dissertation defense. If your AI use has deviated from your prior AI statement, address this with your committee well ahead of time: Chris will not participate in a dissertation defense containing last-minute changes to AI usage to which he has not agreed.

Incidental AI use

Generative AI is now built into the everyday software we use in ways we might not even notice. Incidental AI use is always permitted, and need not be described in your AI statement.

Examples of incidental use include spelling and grammar checking, autocompletion of a word or a line of code, AI-generated summaries that appear in search results, and accessibility features such as automatic captions in meetings. Examples which are not incidental include asking a chatbot to rewrite a paragraph, using the generative rewriting features of a writing assistant, using the Google search "AI mode" to develop an argument, having a coding assistant write functions or programs, and transcribing interviews.

Incidental AI use is changing rapidly; the following criteria help define incidental AI use:

• It is built into everyday software as an ordinary feature (a word processor, code editor, search engine, or video-conferencing tool), rather than being a tool you sought out for its AI capabilities.
• It makes local, not structural, changes. It affects spelling, a word, a phrase, or a line of code, not the ideas, argument, organization, analysis, or interpretation of your work.
• The user reviews changes individually and decides whether to accept them.
• It does not substitute for the intellectual work the milestone assesses. For example, reading an AI-generated summary of an article instead of reading the article is not incidental.

Use of AI with research data is never incidental. Any processing of research data by AI, including transcription, must be described in your AI statement (and likely your IRB as well). When in doubt, discuss your plans with your advisor and committee.

Processes

Writing your AI statement

Write and date your AI statement before you assemble your committee. If you are not Chris's advisee, Chris will ask to read your statement before agreeing to join your committee. Your statement comes into force once each committee member has approved it, either by signing it or by confirming by email.

Updating an AI statement

If you are considering extending how you use AI, you are encouraged to discuss your thinking with your committee. Updates to your AI statement must be approved by your committee (by signature or email confirmation) before they come into force. Updates must have a date, must contain a full statement of the new policy, and must explain what changed and why. Keep every version: all versions are part of the record of your milestone.

Chris will generally not accept revisions to your AI statement after the relevant research has been conducted, and will never accept revisions to your AI statement following questions about your academic integrity.

Deviations

If you realize that you have used AI for a milestone in a way your statement does not describe, report it promptly, in writing, to your advisor or committee chair. Describe what happened, which parts of your work are affected, and what you propose to do about it. As with the IRB, a deviation you report yourself is treated very differently from one that someone else discovers. When Chris serves on your committee, Chris will generally treat a prompt, self-reported deviation as a matter to be resolved with the committee, rather than as an academic integrity matter. Reporting a deviation does not retroactively revise your AI statement; if you want to change how you use AI going forward, submit an update as described above.

Academic integrity concerns

AI usage which is not described in your AI statement and is not incidental is a violation of academic integrity. Concerns may arise in many ways, including from the RAE discussion or from reading your work. Potential violations of academic integrity will be handled according to the policies and procedures of UB's Office of Academic Integrity.

Confidentiality

For all students on whose committees Chris serves, AI statements will not be shared publicly, but may be shared (by Chris) with GSE faculty and staff who have a legitimate educational interest. Students who would like to see examples of AI statements are welcome to ask their peers directly. It is up to you (with guidance from your committee) on whether and how you address AI usage in your dissertation.

Chris's AI statement

In order to be held reciprocally accountable, Chris has also written an AI statement for his own teaching, research, and service. Don't use this as a model for your own AI statement: this is a general policy which cuts across all aspects of Chris's work. Students' AI statements are expected to be more specific.

Research

Chris is open to using generative AI for any aspect of research, so long as it is compliant with the relevant IRB and UB's policies related to technology usage, data risk classifications, accessibility, and any other relevant policies. This includes but is not limited to the following:

• Using generative AI for audio transcription
• Writing research software
• Qualitative data analysis
• Drafting and revising writing

Chris commits to honestly disclosing the AI usage for research publications, both to co-authors (through internal discussion) and to readers (by following the policies of the publication venue). Chris will generally defer to a lead author on AI usage, while still remaining honest, transparent, and compliant with requirements. In particular, if Chris collaborates on a student's milestone work, Chris follows the student's AI statement.

Teaching

Chris uses generative AI in curriculum development and refinement, and in developing software which supports teaching, including automated checks of student work. Chris does not use AI to draft communication with students, including email correspondence and feedback on assignments. Chris does not upload student work to cloud-based generative AI services.

Service

Chris regularly uses generative AI in creating documents which will not be published in academic venues, and does not necessarily disclose such usage. This includes websites, slide presentations, reports, etc. Chris does not use generative AI in correspondence or other contexts of direct communication where the reader has a reasonable expectation that the writing is Chris's own ideas and voice.

When serving on students' committees, Chris does not upload students' milestone work to cloud-based generative AI services. When reviewing manuscripts or grant proposals, Chris follows the policies of the venue or funder, and does not upload materials under review to cloud-based generative AI services.