Using AI as a graduate researcher

Artificial intelligence (AI) is changing the ways in which we think, work and conduct research.

The University does not prohibit the use of AI and other digital assistance tools (DATs) and encourages candidates to explore whether appropriate use of such tools may be of benefit in their research. However, such use cannot be in breach of the policies and expectations set out below and elsewhere.

Technologies and applications are advancing very rapidly and thus we do not identify specific tools or uses of them. If you have specific questions that are not addressed below, please take them to your supervisor or other supports in your department.

What are digital assistant tools (DATs)?

‘Digital assistance tool’ (DAT) is a broad term that covers AI-enabled tools, as well as other software used to support research-related work. DATs include generative artificial intelligence (GenAI) tools, machine translation tools, machine learning and pattern discovery systems, and editorial assistants.

Widely used general GenAI tools include ChatGPT, Claude, Gemini, SparkAI, Microsoft CoPilot, GitHub CoPilot, MidJourney, DallE, Elicit, and many others.

These tools can be used for tasks such as:

Creating descriptions and research plans
Responding to questions
Coding
Analysing data and generating graphs and diagrams
Generating hypotheses
Creating literature reviews and review papers
Summarising literature and other similar kinds of material
Generating or modifying text, audio, images and other forms of content.

Only some of these uses are acceptable when conducting research

There are limits on AI use in a university setting. Many publishers prohibit AI-generated images, for example, and in a thesis, the text must be written by the candidate.

AI and responsible research

While these tools are impressive, they have limitations that can lead to misleading or incorrect results. Graduate researchers should take a cautious and considered approach in their use.

Critically, these tools do not remove or mitigate candidates’ ethical responsibilities. Misuse can harm research integrity and create risks for intellectual property, data security and accuracy. There are also serious threats to privacy, confidentiality and authorship.

The University recognises and supports that GenAI, and other DATs are part of the contemporary research environment and their use by graduate researchers is not prohibited. For example, researchers may use AI to summarise information, create draft transcriptions, support coding, or assist with proofreading their text.

Consider the following when determining whether to use AI in your research.

Text that is produced or altered by GenAI

If a text or presentation has been produced or significantly changed by GenAI or a similar digital tool, this use must be cited as a quote in any publication, proposal, review or thesis. The tool and date of use should be named in a footnote. This is a minimum requirement only; in some contexts, including graduate research theses, more stringent requirements may apply.

Clear citation of AI use is also required in contexts such as emails between graduate researchers and supervisors. This helps supervisors assess their candidate’s understanding of the research topic and communication skills, that is, it helps them to distinguish between human-authored and AI-authored materials.

Using GenAI within a thesis

In a thesis, editorial use of digital assistance tools must be acknowledged and be limited to the level of assistance permitted in the Graduate Research Training Policy (MPF1321).

You may use GenAI for editorial assistance in a thesis, but this use is limited specifically to grammar, clarity, and spelling.

According to University advice on the responsible use of DATs, if you generate text or alter text without noting the use of GenAI, it may be considered a breach of the Australian Code for the Responsible Conduct of Research, 2018 (Code).

To comply with the Code, you must maintain a record of any digital assistance use. This should include:

The programs or applications used and the dates of use
Material such as logs of interaction, prompts, material that has been uploaded, and outputs that have been used.

The use of DATs in your thesis needs to be acknowledged, but excessive use of such assistance may affect the degree to which the work is considered the author’s own. This may also trigger plagiarism concerns.

This means that DATs cannot be used for initial drafting, substantial redrafting, or for producing an English translation of text that has been written in another language.

For further guidance, view the process for preparing a graduate research thesis.

AI tools cannot be listed as authors. Authors must be able to consent to their authorship, take responsibility for their contributions and support the accuracy and integrity of the research output.

Use of generative AI or other digital assistance tools in theses and publications must be consistent with research integrity guidelines. These standards are outlined in:

DATs and confidential information

SparkAI and Copilot are platforms licenced by the University and are considered secure spaces, unlike typical free-to-use public AI tools.

Depending on your research project, you may use AI tools that are not licenced by the University. In this instance, you should avoid entering confidential information or unprotected intellectual property into the AI platform.

The use of these AI tools should always comply with privacy and confidentiality obligations and be carried out in consultation with your supervisors.

Tools that are freely available online rarely have appropriate levels of control of privacy and confidentiality, even if they offer the option to not retain information. Data should only be uploaded to tools where contractual terms of use establish that confidentiality will be maintained.

If you are unsure, please only use tools that are provided by the University.

Considerations when using AI for academic research

Consider the following when deciding whether you should use GenAI for academic research purposes.

The technology can be unreliable

The summaries that GenAI generates can appear knowledgeable and plausible. Nonetheless the content can be inconsistent, wrong, incomplete, misleading, or absurd.

Similar issues may arise when DATs are used to translate text into or from other languages.

Always check outputs for correctness, accuracy, completeness, relevance, and bias. Your work as a researcher must always meet academic, ethical, and professional standards. Keep in mind that your research is your responsibility; and if breaches are discovered years or decades in the future, you may nonetheless still be held accountable.

Use of the technology can undermine your skills development

As a graduate researcher, you are expected to be able to undertake all stages of research, from the development of concepts to the communication of results.

You will develop a set of skills beyond those of research – including learning to write, network and communicate your research findings to relevant audiences. These are skills that can continue to support you beyond the development of a thesis, throughout your future career.

Writing skills are far more than the capacity to describe ideas in text. Writing, along with oral presentations like the viva, sharpens the ability to clarify, organise, structure and shape persuasive arguments for specific audiences. Learning how to express ideas and concepts clearly in writing is crucial for developing these concepts into research contributions. Likewise, you need to learn how to structure your arguments and write clear, detailed descriptions.

Development of these skills is an important part of becoming a researcher.

GenAI can write initial drafts, translate them unreliably into English from another language, and assemble and summarise material unreliably for a literature review. Using GenAI in this manner can prevent you from acquiring fundamental research skills that are critical to your career: doing the work yourself is how you build your capacities.

This applies not just to theses but to the use of GenAI in creating emails, literature reviews, proposals, and progress reports. It is through the assessment, review and critique of your individual writing that your supervisor can support your development and growth as a researcher.

Data governance

University guidelines covering the use of data in research are designed to protect the interests of participants and maintain good research practice (see also the research data services directory for further guidance on managing your data).

AI must be used with care, in consultation with your supervisors, to ensure that you are fulfilling your legal and ethical obligations.

Intellectual property

Be aware that generative AI tools may retain the material you enter by default, even if you ask them not to.

Confidential information should not be shared. If a prompt submitted to an online DAT concerns an innovation of yours, even temporary retention of the prompt by the DAT may mean that the IP is no longer owned by you or by the University.

You should only enter confidential, personal, propriety or otherwise protected information into tools that provide appropriate safety measures for that information.

This includes material such as papers that are copyright to a publisher. These cannot be entered into a DAT if it is possible that their content will be retained.

Further guidance and support

These rapidly developing tools are transforming how we work, study and research, and there are contexts in which they can be an amplifier of skills. However, they must be used with caution.

See more information on the University’s guidance and responses to commonly asked questions about writing with AI and our statement on the responsible use of digital assistance tools in research.

If you have any questions about the use of generative AI tools in research or have any suggestions for further resources about GenAI and research, please contact Graduate Research (Melbourne Research and Enterprise) at: gr-chancellery@unimelb.edu.au