Artificial intelligence, especially generative AI, has quickly become part of the conversation around open scholarship. In many ways, it was perhaps inevitable. New technologies have always shaped how knowledge is produced, analysed, and shared, and AI is already doing exactly that. For research support roles, including research librarians and Open Research advocates, this shift raises several practical and ethical questions.
How do we maintain openness and transparency when the tools researchers increasingly rely on operate in ways we do not fully see or understand? Open research has always been built around the idea that research processes should be transparent and accessible. Practices such as open data, open methods, and open access publishing aim to make it easier for others to understand how knowledge is produced and to reproduce or build on that work. AI complicates this picture. Many AI systems work as what we often call “black boxes” (Pasquale, 2015). They produce answers, summaries, or text, but the steps that lead to those outputs are not always clear to us. For researchers who care deeply about transparency, that raises legitimate concerns.
At the same time, it would be hard to argue that AI tools are not useful. Anyone who has experimented with them quickly notices how much time they can save. Tasks that once took hours, drafting text, summarising articles, organising ideas, even writing code, can now be done in minutes. For researchers working in environments where time and resources are limited, these efficiencies are understandably attractive. Because of that, the real question is no longer whether researchers will use AI. Many already do. The more meaningful question is how they can use it responsibly.
At Sheffield Hallam University, a public research university in England, Open Research Champions are staff and researchers committed to promoting open research practices, advocating for transparency across disciplines, and providing guidance to colleagues navigating the rapidly evolving landscape of scholarly communication. As one of these Champions, I do not approach AI as something that should either be strongly promoted or strictly avoided. Instead, I see my role as helping researchers across disciplines think carefully about how they use AI systems. Responsible use of AI means recognising both its strengths and its limitations. Generative AI systems are known to produce outputs that may contain mistakes, biases, or information that simply is not correct. These so-called “hallucinations” are well documented (Ji et al., 2023), yet the responses often appear confident and convincing. That makes it easy to take them at face value if we are not paying attention. Because of this, a critical mindset becomes essential. AI outputs should be treated as a starting point rather than a finished answer. Researchers still need to verify information, check sources, and apply their own disciplinary knowledge and judgement. None of the core practices of good scholarship disappear simply because a new tool is available. If anything, they become even more important.
Recently, I had the opportunity to explore these issues in practice while designing a teaching session on AI for our postgraduate students. There was, predictably, both curiosity and uncertainty among participants. The goal of the session was not just to demonstrate tools, but to open a discussion about what sits behind them. I sought to explore and evaluate how generative AI systems produce their outputs and why they should not automatically be treated as reliable sources of information. More importantly, I aimed to prompt researchers and students to think critically about what transparency really looks like when AI is part of the research process.
If a researcher uses AI to support their work, whether to structure ideas, summarise literature, or draft text, should that involvement be acknowledged? It should. One practical approach is to document the use of AI as part of the research method. This might include referencing the tool that was used, describing the prompts that guided the interaction, or keeping a record of how the outputs informed the final work. Recording prompt histories, for example, may help clarify how certain ideas or drafts were generated. While this does not fully open the “black box” of the AI system itself, it does make the researcher’s process more transparent. This idea fits naturally with the broader aims of open research. Openness has never been only about sharing final outputs. It is also about making the process of knowledge production more visible. Documenting AI use becomes another way of showing how research evolves and how decisions are made along the way. It also helps others better understand the context in which the work was produced.
One of the central concerns of open education is equity and access. On the surface, AI tools can appear to level the playing field. They can assist with language editing, help structure writing, or support literature searches, which may be particularly helpful for researchers working in multilingual environments or for those with limited institutional support. Yet the situation is more complicated than that. AI models are trained on enormous datasets, and those datasets do not always represent the full diversity of global knowledge (Bender et al., 2021). Certain languages, regions, or perspectives may be underrepresented. When that happens, the outputs generated by the model can reflect those gaps and biases. If researchers rely on these outputs uncritically, there is a risk that existing inequalities in knowledge production will simply be reproduced in new ways (Noble, 2018).
This is where research librarians and open research advocates can and should play an important role. Through workshops, teaching sessions, and everyday conversations with researchers, we can help build what is often referred to as AI literacy (Long & Magerko, 2020). This goes beyond learning how to prompt an AI tool effectively. It involves understanding where these systems come from, how they are trained, what their limitations are, and how they might influence research practices. Teaching about AI also creates an opportunity to talk about broader issues that sometimes remain invisible, such as data representation, algorithmic bias, and the global distribution of technological power. By raising these questions, we encourage researchers and students to think more critically about the tools they are using and the systems behind them.
In many ways, the role of research librarians is evolving alongside these technological changes. Supporting open research increasingly means engaging with digital tools that shape how knowledge is produced and shared. AI is just the latest example of this shift. It introduces new challenges, particularly around transparency and accountability, but fortunately it also opens up conversations about how research should be conducted in this changing technological landscape. What seems most important is not losing sight of the core principles that guide open scholarship. Transparency, rigour, critical thinking, and inclusivity remain just as relevant as before. AI does not replace these values. If anything, it reminds us why they matter.
As researchers continue experimenting with AI in their work, these principles will become even more important. Tracking how AI is used, questioning its outputs, and being aware of issues such as bias and representation might seem like small actions, but they make a real difference. These steps can help researchers navigate this new landscape without losing sight of the values that make research responsible and open.
AI Use Statement
The author acknowledges the use of ChatGPT to explore potential phrasing during the drafting of this article. All intellectual content, arguments, and final iterations remain solely those of the author.
References
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots. Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623. https://doi.org/10.1145/3442188.3445922
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730
Long, D., & Magerko, B. (2020). What is AI literacy? Competencies and design considerations. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–16. https://doi.org/10.1145/3313831.3376727
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. NYU Press. https://doi.org/10.18574/nyu/9781479833641.001.0001
Pasquale, F. (2015). The secret algorithms that control money and information. Harvard University Press. http://www.jstor.org/stable/j.ctt13x0hch
About the Author
Domi Smithson is a Research Support Librarian at Sheffield Hallam University, where she serves as an Open Research Champion, supporting faculty, researchers and students across disciplines in adopting transparent, open, and responsible research practices. Her work sits at the intersection of research support, open scholarship, and emerging digital technologies.
Correspondence: d.smithson@shu.ac.uk
Cite this article
Smithson, D. (2026). Critical thinking for equity on both ends of the AI model. Society and AI. https://societyandai.org/insights/critical-thinking-equity-both-ends-ai-model/
Write for Society & AI
We welcome contributions from educators, researchers, policymakers, and practitioners examining the intersection of artificial intelligence, education, and society. All work published on Society & AI is made available free of cost and in open access, in service of education and the public good. If you have a perspective to share, send your proposal to the Editorial and Content Director at submissions@societyandai.org.
