Ethical Use of Generative AI in Research: Policy Guidelines

Scope: This policy applies to all researchers at I2S, including faculty, staff, and students. It governs the ethical and responsible use of generative AI (GenAI) tools in the research process. The goal is to harness GenAI’s benefits while maintaining academic integrity, originality, and confidentiality in all research activities.

Originality and Accountability

  • All content in a document or experiment must be the author’s own creation or properly attributed. 
  • You remain fully responsible for any content that you include in your research, including AI-generated material [1].
  • It is prohibited to have an AI model generate or manipulate “experimental results” (text, images, or data) that you then present as findings. This is a firm research integrity issue: you may not use AI tools to edit photographs, figures, or any results. Never alter or “improve” research images using GenAI.
  • You must independently develop, verify, or validate any AI-generated content. If you use GenAI in any capacity (even just for editing), you are responsible for verifying the accuracy and truthfulness of the output. This includes ensuring that any writing is free of plagiarism and that all sources actually exist and are properly cited.
  • All key ideas, analysis, and conclusions should originate from the researchers’ own critical thinking and creative effort. GenAI cannot be the sole source of scholarly novelty.

In summary, use GenAI as a helper for minor tasks if needed. However, the intellectual originality, accuracy, and integrity of the work remain fully with the author(s).

Transparency in GenAI Usage

  • Full transparency is required when using GenAI in the research process. 
  • Inform all contributors and collaborators if you use a GenAI tool for any aspect of your work. 
  • Graduate students shall discuss GenAI usage with their advisors ahead of time, and research teams shall agree on what is appropriate. No one shall be “in the dark” about AI involvement in producing parts of a paper or project.
  • Follow the policies of the publication venues to properly disclose the use of AI tools. 

Responsibility and Ethics

  • All researchers must comply with applicable external policies regarding AI usage, which include the rules set by funding agencies, ethics boards, publishers, and professional societies [2].
  • Advisors and faculty have a special responsibility to mentor students on proper GenAI use: Faculty should set clear expectations with their graduate students about what kinds of AI assistance (if any) are appropriate for specific projects. Graduate students should proactively seek guidance before using AI. 
  • Maintain a critical and ethical mindset when using AI. Generative AI can introduce pitfalls like biases, errors, or inadvertent plagiarism. Researchers should approach any AI output with healthy skepticism and ethical caution. It is your ethical duty to thoroughly review and refine AI contributions. 

Confidentiality and Respect for Data

  • Protect confidential information at all times. Never input sensitive or non-public data into a generative AI tool that is not explicitly approved for such use. 
    • Do not share unpublished research details, proprietary data, or any confidential documents with these tools. 
    • Do not input personal or identifying information about other people into GenAI tools:  Personal data (PII) and protected health information (PHI) are often governed by laws like GDPR, FERPA, HIPAA, etc., and sharing such data with an AI service could violate those laws and individuals’ privacy.
  • As of the creation of this policy document, Microsoft Copilot is designated as the only AI tool approved for use with sensitive or personal identifying information and data or university business information. Always check the updated University policy.
  • If the data or text is not publicly available or is not yours to share, keep it out of prompt inputs. 
  •  Do not let an AI model write critiques or reviews of someone’s unpublished work (such as during peer review). 
  • Be mindful of intellectual property when using AI. Respect and protect the work, ideas, and data of others as if they were your own. 
  • Any use of GenAI should be done in a collegial and respectful manner. Follow the most stringent standard that applies among your collaborators or as set by your advisor.

Allowed vs. Prohibited Uses of GenAI

To help clarify the above principles, we enumerate some allowed versus not allowed uses of generative AI in typical research scenarios. When in doubt about a particular use, consult your advisor or the research office before proceeding.

✅ Allowed Uses (With Conditions)

  • Brainstorming and Ideation: You may use GenAI tools to brainstorm research ideas, questions, or approaches in the early stages of a project. However, you must critically evaluate any AI-suggested ideas yourself. 
  • Outlining and Summarizing: It is permissible to have GenAI assist in generating an outline for a paper or a summary of existing text that you have written. 
  • Editing and Proofreading (“Non-Substantive” Use):Using GenAI for minor editing, grammar correction, or translation of your own writing is allowed and is considered a non-substantive use of the tool. Conditions: The original input must be your work (AI should not be writing new scholarly content from scratch), and you must carefully review the AI’s output. 
  • Coding Assistance (Verification Required): If your research involves programming or data analysis, you may use GenAI coding assistants to help with routine coding tasks, for example, generating boilerplate code, suggesting how to fix a bug, or providing pseudocode for an algorithm. However, you must test and vet any code the AI provides. It’s your responsibility to ensure the code is correct, efficient, and appropriate for your data. Be mindful of data privacy when using cloud-based coding AIs. In summary, coding help is allowed for non-novel parts of the work (e.g., generic functions), with rigorous human oversight and testing.
  • Non-Substantive Utilities: Tasks that do not contribute substantive intellectual content are generally permissible to do with GenAI. This includes things like: converting a chunk of text into a different format or style (e.g., turning an essay into a bulleted list for your own study purposes), using AI to translate a document you wrote into another language for outreach, or having AI format references in a bibliography. Such uses do not typically require disclosure (especially if purely for internal convenience), though if in doubt, you can mention them to your advisor or in a footnote. You must still review the results (e.g., ensure a translation is accurate and doesn’t inadvertently change technical terms). 
  • Learning and Skill Development: Finally, using GenAI as a learning aid for yourself is allowed. Do not abuse this by having the AI do your homework or qualifying exams – that would violate academic integrity policies. Use it in the spirit of learning, not cheating.

Important: Even for all “allowed” uses above, you should keep your advisor or mentor informed. In particular, graduate students should err on the side of disclosure: tell your advisor “I used an AI tool to proofread my draft” or “I got some coding suggestions from Copilot for this analysis, and then I verified them.” This ensures there are no surprises and that your supervisor agrees those uses were appropriate. 

❌ Not Allowed Uses

  • Producing Core Research Content: You cannot use GenAI to generate core content or results that you then pass off as genuine. This includes writing whole paragraphs of your paper that contain scientific arguments, generating hypotheses or analyses that you have not vetted, or creating any data that did not actually come from a legitimate research process
  • Faking or Altering Results: It is unacceptable to use AI to fabricate experimental results, images, or datasets. All data and figures in research must come from actual experiments, observations, or legitimate simulations you conducted. You cannot use an image generator to create a graph that “looks about right” for missing data points, nor use AI to morph a microscopy image to “enhance” a result beyond what the raw data shows. Such actions constitute academic misconduct.
  • Uncredited Writing of Substantial Passages: Do not rely on GenAI to write substantial portions of your manuscript or thesis, even if you plan to edit them later. While polishing grammar is fine, having the AI draft entire sections (e.g. “write the Related Work section for me”) crosses into unethical and unacceptable territory. 
  • Fabricating Citations or Sources:do not use AI to generate references, citations, or quotes. GenAI tools will often fabricate reference details that look plausible but are entirely fake (e.g., a non-existent article by a real author, or a bogus DOI). Relying on AI for your citations is strictly forbidden. 
  • Breaching Confidentiality or Privacy: Any use of AI that could breach confidentiality is disallowed. Do not upload or share sensitive data, unpublished results, or personal data with an AI tool. This prohibition is repeated here to stress its importance: violating this not only breaks our institutional policy but can also violate laws, grant agreements, and research ethics protocols. If you are unsure whether some content is confidential or allowed to be processed by an AI, err on the side of caution and assume it is not allowed.
  • Replacing Human Judgment in Critical Decisions: GenAI should not be used in any manner that replaces critical human judgment in research decisions. For instance, you should not ask an AI “which dataset should I choose for my study?” and blindly follow it, nor “do these results support my hypothesis?” These are core research tasks that require your expertise and analytical thinking. Over-reliance on AI not only risks errors but stunts your development as a scholar. 
  • Any Use Violating External or Disciplinary Rules: If a particular use of AI is banned by your publisher, funder, IRB, or any authoritative body relevant to your work, then it is not allowed under this policy. Always stay informed about the guidelines specific to your field  When in doubt, the strictest applicable rule should guide your actions.

In summary, if a use of GenAI feels like it might be “cheating”, whether cheating the process of inquiry, undermining the honesty of your work, or risking someone’s trust – then you should avoid it. When uncertain, seek advice from your advisor, the Office of Research Integrity, or other institutional resources before proceeding. We encourage curiosity and learning about AI, but it must be done hand-in-hand with ethical vigilance.

References

  1. University of Kansas, University AI Guidelines (2025). https://ai.ku.edu/
  2. Columbia University Office of the Provost, Generative AI Policy(2025). https://provost.columbia.edu/content/office-senior-vice-provost/ai-policy
  3. University of Oxford, Policy for Using Generative AI in Research (2025). https://www.ox.ac.uk/research/support-researchers/research-practice/policy-generative-ai-research
  4. Ohio State University, AI Responsible Use Guidelines (2025). https://ai.osu.edu/resources-buckeyes/responsible-use

[1] All major research venues (NeurIPS, ICML, ICLR, AAAI, ACL, etc.) have affirmed that AI systems cannot be authors. This means the human researchers must take credit and accountability for all aspects of the work.

[2] U.S. federal funders, like NSF and NIH, currently prohibit using generative AI for writing peer reviews or handling sensitive proposal data.