Guidance to PGRs on the use of GenAI

This document on the use of Generative Artificial Intelligence (GenAI) for postgraduate researchers (PGRs) draws on existing guidance for researchers as well as guidance for taught students. Please also refer to the University’s Good Research and Innovation Practice guidance.

PGRs, as researchers, produce original knowledge for an assessment (thesis and viva) that leads to the PhD, and as such, you must be mindful of the principles of research integrity and academic integrity, and your use of generative AI must align with the University’s expectations for responsible research and academic integrity.

The University recognises that Generative AI holds the promise of revolutionising how research is conducted, from productivity aids in writing and presenting results, through to the generation of results themselves. The University of Sheffield is committed to ensuring that you are supported to develop AI literacy without compromising on academic rigour and integrity. Alongside other Russell Group Universities (see their principles on the use of GenAI tools in education here), we want to ensure that Generative AI tools are used critically, effectively, ethically, sustainably and transparently in your research, and that research students and their supervisors are supported to do this. 

Many of the tools we are already familiar with and use have built-in AI elements and this will increase rapidly over the next few years. However, the ubiquity of GenAI tools, and the sometimes rich and highly convincing content they produce, can lead to a blurring of what might be a legitimate use of technological assistance, and  what might constitute academic dishonesty or research misconduct (cheating), where it is no longer your own work but the work of GenAI. In using GenAI tools, you must be mindful of principles of academic integrity. Remember that while GenAI might support learning, it is not essential to achieving a good result. While AI tools may augment your personal insights, analysis, thinking and judgments, they should not replace them. The final content you produce must be your own, that you can defend in a viva.

GenAI and your PhD

The conduct of research must comply with the principles and regulations in relation to research integrity, and all material, wholly or partially generated, or modified using an AI tool must be declared. Passing off GenAI outputs as one’s own counts as misconduct, and failure to disclose can lead to sanctions. 

To meet the criteria for the award of a doctorate, you must satisfy the examiners that your thesis: 

•  Is original work that forms an addition to knowledge
• Shows evidence of systematic study and of the ability to relate the results of such study to the general body of knowledge in the subject
• Is worthy of publication either in full or in an abridged form

After submission, examiners assess whether the thesis meets these criteria, informed by discussions in the oral examination. You should be able to demonstrate, via the thesis and oral examination, that you can:

• Critically appraise what is and what is not known in your subject area
• Formulate appropriate questions to probe what is not known
• Choose and, as necessary, devise appropriate techniques to address such questions
• Explain to others why these questions are worth asking and why these techniques are the right ones to use to answer them in a realistic and timely manner
• Employ such techniques rigorously and viably, to produce robust and reliable answers to the questions posed, while remaining open-minded to unexpected or unintended outcomes

Where GenAI has been used to an extent that it compromises your authorship of the thesis, you are likely to find yourself unable to satisfactorily answer such questions. Moreover, even where your authorship has not been compromised, you might still struggle to answer such questions to your examiners’ satisfaction if you have followed poor practice in your use of generative AI tools – for example, if you have relied exclusively on information provided by the tool without reference to other sources.

Using GenAI appropriately 

Used properly and strategically, GenAI can augment our creativity and productivity. Software you currently use may already have grammar and spell checks built in. Additionally, GenAI can:

• Help generate ideas and frameworks
• Organise rough typed notes
• Paraphrase and summarise sources as a study aid
• Reformat text, tables etc.
• Suggest code completions or assist with debugging for programming work
• Translate text between languages
• Summarise meetings, video or audio transcripts
• Help create effective presentations

The key is developing skills to harness GenAI effectively as part of your own learning process, while maintaining full academic integrity through proper attribution and transparency. Maintaining academic integrity requires human direction over how generative AI is leveraged - copying and pasting from a GenAI tool is tantamount to plagiarism. Furthermore, while outputs from GenAI may seem convincing and exhaustive, they need thorough checking as GenAI can generate factually incorrect or non-existent sources and information, and will tend to reflect the biases in the data the models were trained on. 

Norms and rules around the use of GenAI tools in research are still being developed and are likely to vary widely depending on disciplinary context. Issues relating to confidentiality, copyright, data protection, and other elements of research integrity will vary in importance based on the nature of individual projects. Rather than prescribe or proscribe GenAI tools, we provide some guidelines (see also some overarching principles in the guidance for researchers) and aligned Frequently Asked Questions (FAQs) below that will help you consider whether you are using GenAI appropriately and judiciously.  

Some practical guidelines for PGRs:

• Maintain a critical approach to AI-generated content: Always verify information and/or code generated by AI, recognising its potential for bias and inaccuracies. You should be able to explain and justify any use of generative AI in your research. Also check whether your funding body, if applicable, has a policy on generative AI use.
• Be transparent, and document any AI usage: Keep detailed records of the specific Generative AI tools used, including their names, versions, dates of use, and how they were employed in the research. Although it is inherent in this technology that reproducibility will be adversely affected (outputs are randomised to simulate a convincingly human response) it is important to be as transparent as possible to enable others to scrutinise your work effectively. There may be good reasons for using GenAI tools in your research but if you cannot explain what you did and why it was a responsible use of GenAI, then that will be a concern.
• Understand data privacy implications: Be cautious about the information input into AI tools, especially when dealing with sensitive research data or personal data/information. Checking the Terms and Conditions that are associated with such tools is the best way to understand how information input into them may be used and what you are committing to. Always be sure that you have the authority to commit to such terms before using the tool. If you upload private, sensitive, confidential or embargoed data to a generative AI tool you are likely to breach privacy and confidentiality rules (e.g. student survey data/ participant data) or intellectual property rules, or breach a contractual or licensing agreement.
• Check the terms and conditions of any Generative AI tool that you use to understand the rights that you grant to the providers to re-use your work. With most Generative AI tools, uploading your work grants the providers the right to use it as they see fit, and this may include sharing it with other people. Your work will normally be incorporated into the training data and used to generate new outputs, at least in part. If you have really novel ideas or thoughts, you might not want to share these via Generative AI tools.
• Respect copyright and intellectual property: Ensure that the use of AI-generated content does not infringe on copyright or intellectual property rights. As with all research, sources should be cited and permission should be sought when using copyright protected material. AI tools should not be considered ‘sources’ in themselves, at best they are aggregators and summarisers of other sources which can often make appropriate attribution impractical if not impossible.
• Stay informed about evolving guidelines: The field of AI is constantly developing. Keep abreast of the latest innovations, guidelines, best practices, and ethical discussions surrounding Generative AI in research. If, for example, you are using a GenAI to save time and/or money, it is worth considering the hidden cost of these benefits in terms of the possible environmental damage or exploitative working conditions involved in creating some of these tools.
• Seek the guidance of your supervisor on your plans and approaches to using Generative AI in your research. They may signpost you to more specialist advice if required. 

Frequently Asked Questions

1. What is Artificial Intelligence (AI)?

Artificial Intelligence, or AI, refers broadly to any technology or system that can reason and/or adapt, sometimes on its own, to achieve certain goals. Machines are able to mimic cognitive functions to perform tasks which usually require human intelligence. AI-based systems can be based purely in the virtual world with things like voice assistants (e.g. Siri or Google Home), image analysis tools, search engines, and technology that can recognise speech and faces. AI can also be embedded in hardware, or objects in the physical world around us such as advanced robots, self-driving cars, drones, or devices connected to the Internet of Things.

2. What is Generative Artificial Intelligence (GenAI)?

Gen AI are AI applications or tools trained on vast amounts of diverse types of data, allowing them to create and / or improve upon text, images, computer code, video and audio. Some of these tools are integrated with the software you routinely use (e.g. for word processing there is Copilot in Microsoft Word). 

Recent breakthroughs in Generative AI (GenAI) have implications for how we approach content creation. Popular tools such as Gemini, ChatGPT, Midjourney and DALL-E are able to:

• Generate text
• Generate images
• Translate text and enhance spelling, grammar and punctuation
• Generate, fix or rewrite programming code
• Summarise documents
• Ask and/or answer questions
• Interface with a user in a way which mimics human interaction

3. Is there a preferred GenAI tool that the University supports the use of?

Preferred Generative AI tools and the way in which they are used are likely to differ between academic disciplines, so support will primarily be provided by your School and supervisors, where you will be given an opportunity to explore and engage with these tools. A critical issue in all instances is how we might continue to uphold academic integrity while we harness the benefits that GenAI offers.

The University of Sheffield provides access to  Google Gemini. You must access Gemini through Sheffield credentials to maintain data security compliance. Alternative tools may be used for specific applications but require supervisor approval and compliance verification.

4. Are students allowed to use generative AI tools when writing their thesis?

Yes, students are permitted to make use of generative AI tools in their thesis writing processes. You should declare your use of GenAI tools and take full responsibility for the content of your submitted thesis. You will not be penalised for your use of generative AI tools, provided that the use is consistent with this guidance and properly declared. 

Research assistance activities such as code development, technical writing, and literature analysis are generally permitted with appropriate declaration. More substantial intellectual contributions, such as drawing research conclusions or developing theoretical frameworks, require careful validation and supervisor discussion to ensure the researcher maintains primary intellectual ownership and understanding at a doctoral level. Should your use of GenAI tools in your thesis writing exceed the circumstances permitted by this guidance, you will risk breaching the Academic Misconduct Policy.

Academic misconduct aside, you should always be cautious about your use of generative AI tools; you should discuss potential issues and concerns prior to any use of generative AI tools; you should recognise that some uses of generative AI tools may not be advisable, and may make the thesis more difficult to defend in the oral examination.

5. Do I have to acknowledge the use of GenAI in my thesis?

The use of AI tools as part of the research process – to analyse and draw insights from data, for example – would, like any other element of research design, need to be clearly outlined in a ‘Methods’ section, subject to ethical review where appropriate, and conducted in accordance with principles of research integrity as well as disciplinary norms. Examiners will scrutinise such aspects of the research design as they would any other element of a project.

Acknowledging the use of Generative AI can take the following steps:

• a pre-defined statement that indicates whether you have used Generative AI tools.
• a reflective component detailing how Generative AI has been used and the experience of engaging with it.
• sharing prompts used, the outputs produced, or the modifications made because of using the tools.
• using appropriate referencing.

6. Can I use GenAI as a source in my thesis?

No. GenAI outputs should never be used as a source in your thesis and you should never cite from the output of a GenAI tool. Content generated by AI tools is not peer-reviewed, cannot be independently verified, and is usually non-recoverable and non-reproducible.

7. Can I use GenAI for a literature review in my thesis?

While AI tools can sometimes help identify relevant literature, you must be aware of significant risks:

• Fabricated citations: AI tools frequently generate plausible-looking but entirely fictitious references. Every source suggested by AI must be independently verified to exist and to actually contain the claims attributed to it.
• Confirmation bias: AI tools can be used to find sources that support a predetermined conclusion without critical evaluation. This undermines the integrity of your research. You remain responsible for ensuring your literature review is balanced, rigorous, and critical.
• Lack of scholarly judgment: AI cannot assess source quality, relevance, or credibility in the way a domain expert can. You must critically evaluate every source yourself.

If you use AI to help identify potential sources, you must:

1. Verify that each source actually exists and is accessible
2. Read the full source yourself (not rely on AI summaries)
3. Confirm the source actually supports the argument you're making
4. Evaluate the source's quality, credibility, and relevance
5. Declare this use of AI in your thesis

Remember: Citing sources that don't exist, or that don't support your claims, constitutes academic misconduct. This remains true even if an AI tool suggested those sources.

For more details on what GenAI can or cannot be used for in your thesis, please see https://sheffield.ac.uk/study-skills/digital/generative-ai/assessment.

8. Can I upload articles for analysis on to a GenAI tool?

Inputting or uploading articles / books or other material into a generative AI tool may contravene the terms of use set by the publishers or other rights holders. Equally, generative AI may produce outputs that ignore intellectual property rights by producing content that does not include appropriate acknowledgement or breaching copyright.

9. Can I use GenAI to analyse data that I have collected?

Researchers must ensure that confidential, proprietary, or personally identifiable information is never uploaded to any GenAI platform. When working with sensitive datasets, consider using anonymised samples or publicly available alternatives for AI-assisted analyses.

The GenAI Guidance Tool will help you consider your use of GenAI in line with the principles described above. It might be useful to use the tool separately for each GenAI tool you use, as the implications for each might be different.