How to Use AI Responsibly in Academic Research Without Violating University Policies

Using AI in academic research is allowed in specific ways only, such as research support, but may not be used to create academic work submitted to the University. What is allowed is in accordance with your university AI policy usage, and in some instances, your supervisor’s guidance. Plagiarism isn’t necessarily the worst of the risks; it can be confidential research data, participant information or unpublished findings that are revealed. This Guide clarifies the concept of Responsible Use, Disclosure, Verification and Data Protection. To get a broader perspective on the impact of AI detection tools in academic research writing, reference the article on how AI detection tools are changing academic research writing.

Why “Check Your University’s Policy” Is the Real Answer

It is not possible to say there is a universal university AI policy. There may be variations in rules from institution, faculty, department and individual supervisor.

Also, before employing AI for academic research, consult the most up-to-date guidance appropriate to your research! Certain institutions may permit the use of generative AI for research and writing that is clearly defined, while others may have more specific guidelines.

There can also be specific requirements in a current university AI policy for acknowledging, disclosing, handling and/or assessing data. This transforms AI academic integrity into a reality more than just a matter of “allowing” or “banning” AI.

National guidance to schools should not be automatically taken to the universities. They have alternative institutional regulations. Policies are also subject to a high rate of change. Last year’s reading may not have conveyed the same expectations.

What Policies Generally Permit

In areas where using AI in academic research is allowed, typical areas of assistance can be:

• Being able to understand challenging texts: posing questions about unfamiliar ideas.
• Literature discovery: developing search terms or topics to look into; then self-checking of source information.
• Language support: proofreading sentences you have written for errors in grammar, clarity and readability.
• Script assistance: providing support and/or assistance with scripts for analysis when it is appropriate for your discipline.
• Checking for clarity: having AI summarise your own draft to determine if the argument is clear.
• Practice: creating questions to check comprehension.
• Reference checking – recognising formatting inconsistencies, then performing manual checks.

The best approach to responsible AI use research is to keep the responsibility for the intellectual contribution with the researcher. AI can be used to assist with a task, but should not be used in place of the researcher’s judgement.

What Policies Generally Prohibit

Most institutional rules clearly differentiate between assistance and authorship.

Any work you submit that is generated by AI may violate rules for AI academic integrity. That doesn’t necessarily change if it’s edited a little bit.

Other common uses that are banned are:

• Creating research data, findings or results.
• Examinees using AI during examination (unless specifically allowed).
• Not disclosing the use of AI when required.
• Giving false references or giving false information.
• The posting of information, matters or work that is confidential, personal or unpublished to an unapproved tool.

This is the same concept which comes into play when using AI in dissertation writing. If you do not clearly know exactly what the tool did and what you personally did, STOP and determine if it complies with the appropriate AI academic integrity policy before you go any further.

The Disclosure Requirement

Disclosing AI use is gaining momentum and is now commonplace for universities, journals and research organisations. The format required, however, is subject to change, and may be specified in the rules where appropriate.

A helpful disclosure is typically the name of the tool, if applicable, its purpose and the portion of the work that it impacts.

For instance: Grammatical and readability checks were done by a named AI tool in the discussion chapter. Any analysis, interpretation and argumentation are solely in the author’s own voice.

You don’t have to do something to make it AI-generated and therefore inapplicable. Documentation can be used to show compliance if disclosing AI use is allowed and properly documented.

In addition, and importantly, it is the case that generatively disclosing AI use could be required by a journal or funder in addition to having to do so by your university.

Data Protection: The Risk Nobody Warns You About

While plagiarism is a potential issue, it’s not necessarily the most significant one regarding using AI in academic research. It may be the manner in which information is handled after it’s uploaded.

If you copy/paste content into a third-party application, you may be copying/pasting it outside of the institution’s controlled environment. Where it includes participants’ data, unpublished results, commercial material or confidential research data, this transfer could pose a compliance issue.

Researchers should:

• Do not use identifiable information about participants in a general AI tool.
• Do not upload data not yet presented at the workshop or to collaborators or sponsors.
• See if an institution offers an accepted AI service.
• Ensure that information is anonymised when necessary and that the process used seems sufficient.
• Go through consent documents with participants and discuss their information as it was explained.

Institutional research data rules and applicable data protection obligations thus need to be taken into account for responsible AI use in research.

Researchers should check with their institution’s research data/information governance office before using AI in academic research to access sensitive materials.

Verification: AI Fabricates Sources

AI can generate realistic citations of fake articles. AI can generate realistic citations for non-existent articles. They can also mix some real author, journal, title and date information into a totally bogus reference.

This is crucial for ensuring the accuracy of using AI in academic research. Verify each citation separately in a known database, publisher, library catalogue or in a DOI record.

Also, do not presume that a correct summary of a real citation. Before using the results of any original paper you read, read the appropriate section of the paper.

In particular, in the field of AI in dissertation writing, ensuring that citations are correct can make or break the overall argument.

It is also applicable to figures, definitions, legislation and statements in regulations. Evidence is in the form of plausible wording.

Build an Audit Trail

An easy audit trail is a safeguard of your research process.

Save drafts and version logs, changes tracked, notes, reading logs, and reference logs. Log AI tools, dates, and why and how you used them whenever using an AI tool.

This helps to maintain AI academic integrity as you can trace the process of how you did your work and not depend on a detector.

The research has triggered worries concerning false positives in AI detection systems. When questions come up, it can therefore be more beneficial to have a documented process.

Your university AI policy could also define what the records or disclosures are that researchers should keep.

Frequently Asked Questions

• Can I use AI to write my dissertation?

Typically, it’s not for work that’s submitted. Some support can be granted, but the use of AI in dissertation writing will be subject to institutional restrictions.

• Do I have to declare AI use?

Yes, if needed by your institution. Disclosing AI use should be in a format as outlined below and should specify what the AI was used to perform.

• Is using AI for grammar checking allowed?

Many times, but not always. Confirm if editing of language is allowed in your programme at the time of the programme.

• Are AI detection tools reliable?

There are issues of accuracy identified and false positives. Do not use results of detection to determine an audit trail.

Conclusion + CTA

Read and protect research data, document AI assistance, and use AI in academic research only within the scope of the policy.