Using AI with Integrity: Our Commitment to Responsible Research
At the University of Bath, leading research is being conducted on the topic of Artificial Intelligence (AI), including how to build intelligent systems (for an example see here). The focus of this blog is not on such research conducted on AI, but instead how we as researchers might use AI technologies to support research. For some disciplines, AI has long been embedded into research practices; however, for others, it may have the potential to transform and reshape how research is designed, conducted, communicated, and used.
Recent developments, especially in Large Language Models, have extended the possibilities of AI from scientific applications to every stage of a research process (see this recent Conversation article). This includes identifying potential collaborators for interdisciplinary research (e.g., Scinapse), concept mapping (e.g., Kumu); conducting searches (e.g., Consensus), reviews (e.g., Elicit), and analysis of the literature (e.g., Paper Digest); or using ‘deep research’ AI agents to analyze data and write reports. A recent tool introduced to me by a member of my own research group is NotebookLM that enables you to convert published academic articles into podcasts – perfect for listening on the go! For simplicity, I note that I refer to AI generally throughout this blog; however, I acknowledge it is a contested term with a wider range of technologies with different technical characteristics and purposes (e.g., Generative AI, computer vision, reinforcement learning etc).
Whilst AI developments can provide many opportunities for academic research and scholarship, they can also present complex challenges. It is important that we as a research community improve our literacy in this area so that we can strike a balance between benefitting from the possibilities that the use of AI tools provide, whilst recognising and managing the ethical, legal, and reputational risks associated with their usage. It is essential that if we employ AI tools, it is done so with integrity and transparency. We also encourage researchers to be mindful of the environmental impact of using AI tools and reflect on their reasons for using AI to justify its usage (e.g., is the AI tool doing something a human couldn’t do, or providing meaningful efficiencies?)
Our University position on the usage of AI in research is aligned with the UK Research Integrity Office (UKRIO) guidance.
Research Integrity Risks Associated with AI Usage
We support responsible and creative use of AI in research, provided it aligns with the following principles:
1. Respecting creativity and critical thinking
AI can assist, but must not displace, independent thinking or scholarly originality. This is especially important for early-career researchers and PGRs developing foundational skills required for rigorous and independent research. For example, Mariachiara Barzotto (School of Management) talks about the potential loss of agency and basic skills in this reel. Researchers are encouraged to reflect on their current level of knowledge around responsible usage of AI within their research and consider further training where appropriate. For example, the University of Bath Research Integrity Training (which has a section on responsible use of AI) or open access resources available at MIT or University of Glasgow.
2. Maintaining a robust and reliable research record
AI can proliferate poor-quality research and, in some cases, completely fabricate sources (an example of an “AI hallucination”). Researchers must adopt rigorous oversight to ensure that any outputs from AI (e.g., citations, analysis or visualisations) are verified and accurate. The following resources may support in this regard: PubPeer, RedacTek, Retraction Watch Database, and Google Lens. As researchers, we will need to improve our skills in using these tools effectively (“prompt engineering”) so that we know what we are asking AI and also the limits of what we get back from it. The importance of researcher oversight is supported by UKRI whose policy on the governance of good research practice requires funding applicants to ensure that anything submitted is not falsified, plagiarised, fabricated, misrepresented, or confidential and used without consent.
3. Ensuring human accountability and full transparency and disclosure
AI cannot be a substitute for academic judgment and final responsibility for content, interpretation, and decisions always remain with the researcher. As per our University guidance for authorship, it is unacceptable to list AI tools as an author. Researchers must declare the use of AI in ethics applications, submissions (e.g., doctoral thesis), funding applications (see research funders’ joint statement and UKRI guidance), and published outputs (see e.g., Nature, Taylor and Francis, and JAMA AI guidance; see also this article for an example of when AI technologies are integrated into writing without declaration). An included declaration should name the tools used, describe how they were used and the role they played in the final output, and note any limitations and measures taken to ensure human oversight. In relation to ethics applications, we have a new section on the use of AI in our Ethics@Bath system, whereby researchers should identify and document any potential ethical risks and harms associated with AI usage. Both the TREGAI checklist and UKRIO’s checklist can provide useful resources for supporting ongoing ethical reflections and planning. Additionally, UKRIO offer guidance on information to be included in participant information sheets to ensure transparency and appropriate consent.
4. Confirming legal, ethical and data compliance
Ensure you are compliant with the University’s cybersecurity guidance and are accessing Microsoft Co-Pilot (the University’s only approved GenAI tool) from your University managed device. The green shield should be present at the top of the page, which confirms that any chats and data uploaded are kept with the University’s system; thus, protecting Intellectual Property and conforming to our data protection obligations. Even within Microsoft Co-Pilot, researchers should not input sensitive, personal, confidential, or copywrited data into AI tools, without appropriate safeguards and consent. Indeed, participants would need to provide explicit consent for such data to be inputted, and researchers would need to ensure clear GDPR compliance and that there is no risk of de-anonymisation of any data. Where researchers are needing to use alternative AI tools to achieve their goals, they should do so in a safe and responsible manner (e.g., inputting public datasets only).
It is also important to ensure you are compliant with funders’ and publishers’ conditions relating to AI usage, privacy, intellectual property, copyright, and data protection and security. Researchers are encouraged to transparently document in their data management plan how AI usage will be tracked and recorded. Edith Cowan University have a useful resource that researchers could explore when creating an AI data management template. When peer reviewing for Journals or funders, if you are thinking of using AI you should first consider ownership of what you are inputting, confidentiality agreements signed, and if usage will compromise subsequent publishing possibilities (and authors’ IP). Always seek advice from the publisher or funder prior to using AI to understand what constitutes acceptable use. For example, UKRI clearly states that reviewers and panelists must not use generative AI as part of their assessment activities and any release of confidential material will constitute a confidentiality and integrity breach. Consequences of upheld research misconduct allegations from the use of generative AI during a UKRI application or assessment can result in grant rejection, future restriction on submissions and assessments, termination of other funding, and reclaiming of unspent monies (for more, see UKRI guidance).
What We’re Doing at Bath
To support our community in using AI responsibly, we have implemented several key measures:
- As part of our new University-wide Research Integrity Training, all researchers can engage with guidance on the responsible use of AI.
- We have updated our Ethics@Bath system to specifically ask researchers about whether and how AI has been used in completing the application form, and whether and how it will be used in the proposed research.
- Our Skills Centre has a lot ofguidance on the use of AI tools for our work, including a specific module and an introduction to Microsoft Copilot.
- The Doctoral College has guidance on the usage of AI in doctoral work.
- The Library hasguidance about how to reference AI appropriately.
- For more on the University’s position on AI in relation to learning and teaching, please visit the Centre for Learning and Teaching. They also provide some GenAI staff development opportunities.
- To build on this blog, we are currently developing University-wide policy on the use of AI in research, in line with UKRIO and funder expectations. Following this, we will deliver an Ethics@Bath Open House session on appropriate use of AI in research to communicate the policy (in the meantime, see this seminar). This link provides an example of how an “AI for research” policy has been developed elsewhere in the sector.
Looking Ahead
The confidence of the public in our University research depends on trust: that our findings are credible, our methods transparent, and our claims accountable. If we allow AI to erode that trust, by enabling undisclosed “black‑box” generated content, unverified outputs, or providing weak oversight, we risk damaging both our academic credibility and public legitimacy. Conversely, by embracing AI with integrity, we can not only safeguard the core values of rigour, integrity, and trust that underpin excellent research but lead the field in responsible innovation. We can position ourselves as a University that is not merely reacting, but proactively shaping how the evolving role of AI is responsibly woven into the fabric of our research.
We invite colleagues from across the University to join us in constructive dialogue as we co-design policy on the use of AI in research and incorporate it into University practices. Please get in touch (research-governance@bath.ac.uk) if you would like to be involved!
*Disclosure: GenAI (Microsoft CoPilot) was used to improve the structuring and readability of this article, given the author’s limited experience in blog writing! All original sources cited have been read in full and incorporated into the blog by the author.
