Ultra‐Processed Research?

We cannot escape artificial intelligence (AI) in research today. Each day brings new features, apps and functions. Sign on to your computer and prepare to be startled, frustrated, annoyed and possibly pleased. But oddly, the frenzy over applications of AI in research feels a lot like a time warp to me. It calls to mind ultra-processed foods of all things. I well recall the buzz that ultra-processed foods created during my childhood and youth. The buzz-worthy brands, trendy products and must-have status might have shifted by locale, but the hype for so-called convenience or fast food was constant and remains so to this day. Now, AI holds a similar place in our collective consumer consciousness as researchers. AI has inflated academic consumerism beyond all imagining, nowhere more so than in research and scientific publishing. Many now see AI as integral to most elements of research operations from conception through grant writing onwards to implementation and dissemination. AI apps undoubtedly offer areas of narrowly defined utility in research. In contemporary terms, research is clearly and necessarily processed in the same way that foods like oats are processed by cutting, crushing or rolling to improve the balance between nutritional accessibility and palatability. Such processing showcases effective use of tools in food preparation. Unfortunately, we nurses seem to be losing our collective ability to distinguish between the appropriate and useful and the spurious and risky, especially when it comes to the set of tools we call AI. We cannot seem to discern where and when to use AI. That is where the parallel with ultra-processed food becomes revelatory. Decades ago, industrialized agriculture and food—and I use the word food advisedly here—manufacture seduced many around the world with the convenience and cachet of ultra-processed foodstuffs. We moved away from whole foods with some processing to ensure optimal nutrition and to enhance palatability. Now, we know too well the costs of industrialized agriculture and ultra-processed food manufacture and consumption to human health. Realizing the cost to our planet in terms of the triple crises of climate, pollution and biodiversity only adds to the anguish that this discovery holds. Similar costs may now apply as we contemplate what it means to move from the current state of what we might reasonably call processed research to the looming prospect of ultra-processed research. The buzz of AI in research echoes what the world endures with ultra-processed foodstuffs. Now, as we foresee the possibility of moving beyond the industrialized agriculture and food-manufacturing era, we see the harms of this period ever more clearly. Our newly balanced perspective creates vision that brings us back to foundational understandings of nutrition and pleasure in eating. Even more fundamentally, we gerontological nurses know that building knowledge and skill over the course of a lifetime—as with learning how to prepare and sometimes even to grow and harvest our own foods—ensures healthful ageing where function and well-being are maintained. Learning knowledge and skills that allow us to prepare food and, for some, to grow it as well cannot be seen as antiquated hobbies in the 21st century. Preparing and growing food are emblematic of human endeavours that build health, function and well-being. Like food preparation, most research processes are necessarily human. As with food preparation, tools improve what we do in research. But our discernment about what constitutes an effective tool, which tools to use and when to use them proves critical. For example, using a teaspoon to ladle soup into bowls quickly proves frustrating. Using a dull knife without knowing how to sharpen and safely use it may prove deeply dangerous. When it comes to tools, knowing when to save labour and when to expend it is crucial. Our application of AI in research currently lacks such discernment, leaving us open to inefficiencies, errors and ethical and legal violations. Our absent discernment when it comes to AI in reporting and publishing research, for example, is reaching a dangerous pitch. Many editors and other authors are writing about using generative AI (GAI) to write research reports. As some favourable views on this sort of application come forward, many still neglect the obvious errors in content and referencing produced by use of GAI. These errors of incorrect content and frequently fictitious references, misnamed hallucinations, require exceptional vigilance and lots of time to identify and correct. Left unchecked, these errors when published may be used by others and replicated, altering the course of scientific development. Researchers who fail to identify them waste their time and that of colleagues but risk far more. The eventual price to be paid when those errors and the researchers who fail to notice them are discovered by hawk-eyed editors and reviewers is likely to be increasingly severe. Momentum behind issuing sanctions that go beyond rejection and retraction is growing. Researchers who lose sight of why we do research—to better humanity and the world we share—in favour of expanding their curricula vitae with slipshod use of GAI beware. Concerningly, using GAI to conduct peer review is no longer a theoretical possibility dismissed as counterintuitive. It is a reality, paradoxically used by researchers who voluntarily accept invitations to conduct peer review of a research report only to void that agreement by doing what can only be called machine review. Although some of my fellow editors posit that AI in peer review may offer important savings in time and labour, I take a different stance. As a gerontological nurse, I think constantly about ageing with health and function. My professional and personal experience point to the incredible benefits of learning new skills and cultivating them over time with persistent practice, habits supported by science. My decades serving as a peer reviewer and editor reveal how, how through dedicated effort, much faster and more skilful each review becomes. I teach students the habits that I employ and share the advantages that I have enjoyed as a result. Consequently, I read comments from peer reviewers that I am confident were written not by the human being who uploaded them but by a GAI app with terrific and escalating concern. Most elementally, peer review, by its very name, mandates a process that is human. Peers are people, not machines. We all—authors, reviewers and editors—apply the use of tools that facilitate the human labour of research and its dissemination. In the first half of the 20th century, typewriters were the most common tool used in peer review. How wonderful it must have been for those who had handwritten their reviews to make the change to a typewriter or, even better, to enjoy the professional services of a secretary to use that typewriter for them. Later in the century, early word processors must have felt revolutionary to those who recalled their use of manual typewriters. Those tools truly improved research operations while supporting human knowledge and skill. Now, we count computers, word processing apps and search engines among our labour-saving toolkits. They help us process our portion of research operations more efficiently and effectively. Those tools save some time, expand our frame of reference, allow us to double-check aspects of the research and reduce labour without altering our knowledge and skill acquisition. Critically, these tools do not replace our human perspective on the research report we are reviewing. Thus, the tools we accept use of in peer review today do not replace human effort; they simply streamline it. Using AI to generate peer review is a far different sort of tool use. It transforms peer review into machine review, moving that review from processed into the domain of ultra processed. Replacing peer review with machine review by using AI is profoundly problematic. It is also redundant. In fact, scientific publishing is replete with machine reviews. We use AI-mediated review tools—what can accurately be called machine review—to screen every manuscript submitted to a journal for potential plagiarism and other text duplication. Similar Al-driven processes can specifically screen reference lists for retracted and fraudulent citations. Contemporary science would collapse without machine review creating processing that now ensures robust publishing practices. The structure and processes lent by these and other tools we humans use to improve scientific publishing are invaluable. These tools cannot, however, replace the people—authors, editors and peer reviewers—that keep science, research and publishing true to purpose. Replacing peer review with another machine process defeats the purpose of science. Science is a human endeavour, developed to satisfy our curiosity and to make our shared world a better place to live and—crucially for we gerontological nurses—grow old. Nursing too is a human endeavour, anchored by a deep understanding of our social existence as a species. Gerontological nursing goes a step further with our shared view of ageing as part of human growth and development fostered by learning and skill acquisition over the course of a lifetime in a social milieu. Science, nursing and gerontological nursing, taken together, form an argument for reinvestment in peer review as an intrinsically human process and against replacing peer review in part or in whole with AI. Nothing can substitute for human intelligence, discernment and debate among authors, peer reviewers and editors. As a gerontological nurse, always watchful for ways to help others grow older with health, function and well-being, I underscore the value of doing for ourselves with as few tools as practicable. Doing for ourselves is as true in peer review as it is in preparing meals with whole foods. We know now definitively that diets laden with ultra-processed foods result in a host of effects on health, spanning malnutrition and altered body weight to elevated risks of many consequential chronic health conditions. Analogously, the use of GAI in writing and peer review may pose risks that look eerily similar in some ways. In conducting peer review, we need to critically reflect on our own writing, incorporating and expanding our knowledge of research as we do the review, to then complete and convey that review through our written comments. Using GAI to do peer review strips away possibilities for a rigorous cycle of critical practice and reflection in which we will build skill in writing and reviewing over time. Thus, using AI in peer review chances shrinking the intellectual ‘muscles’ we would otherwise build doing peer review in the ‘old fashioned’ way. More broadly, skills like writing and reading, both inherent in peer review, are widely acknowledged as vital to brain health and healthful ageing. As gerontological nurses, we should advocate for ways to use and develop reading and writing skill in meaningful ways, not push for technological ‘solutions’ to avoid using these fundamental human skills. To do otherwise would be akin to saying to someone while in our clinical roles ‘that's ok, no need to eat whole fruits, just go ahead and dissolve this artificially colored, synthetically fruit flavored, sugar-sweetened, vitamin enriched solution in water instead’. AI technology and tools must only be used on a limited basis if at all in connection with peer review. A GenAI (sic) tool can only be used by a peer reviewer to improve the grammar and paragraph structure of the written feedback in a peer review report. This use must be transparently declared upon submission of the peer review report to the manuscript's handling editor. Independent of this limited use case, peer reviewers must never upload manuscripts (or any parts of manuscripts including figures and tables) into any AI technology or tool. AI technology and tools may use input data for training or other purposes, which would violate the confidentiality of the peer review process, privacy of authors and reviewers, and the copyright of the manuscript under review. Moreover, the peer review process is a human endeavor. Responsibility and accountability for submitting a peer review report, in line with a journal's editorial policies and peer review model, sits with those individuals who have accepted an invitation from a journal to undertake the peer review of a submitted manuscript. This process must never be delegated to AI technology and tools. I and the other members of the IJOPN editorial team welcome, as always, opportunities for conversations with our authors, peer reviewers and readers. We know that AI is a ‘hot’ topic in research and in care for older people. Share your thoughts on the use of AI in gerontological nursing research, practice and education as well as in services for older people via the IJOPN LinkedIn page (https: //www. linkedin. com/company/international-journal-of-older-people-nursingg? trk=publicₚostfeed-actor-name). The author declares no conflicts of interest. The author has nothing to report.