It's Not About the AI
By Michael Doneman, PhD candidate, School of Media and Communication, RMIT University
Back in October 2025, I read the newly released AAG discussion paper Artificial Intelligence (AI) and ageing research: Ethical considerations and implications, suggesting that generative AI could become a research collaborator, under human direction. I still think that's true. But after substantial engagement through fieldwork, it no longer feels like the most interesting part of the story. The collaborations that landed most productively didn’t happen between a person and a machine. They happened among people around the machine.
The fieldwork was Datascapes, my doctoral project at RMIT University in partnership with U3A Network Victoria. Across four cycles of participatory action research, older adults used AI to build courses, draft concept papers and governance documents, teach one another, and argue about what the technology was actually good for. I'm now thinking about this and writing up our findings.
The AAG paper’s cautionary approach is directed primarily at ageing research: at how AI may weaken critical thinking, reproduce or reinforce ageism and cultural harm, and place particular pressure on early-career researchers still developing methodological judgement. My field notes offer plenty of examples of those risks, and my experience with Datascapes adds a related question: what happens when older people are not simply the subjects or intended beneficiaries of research, but participants in deciding how AI should be used?
Throughout the Datascapes program, ethical judgement rarely lived in any one person or policy. Early in the first cycle, Helen told me, “I don’t want AI to give me answers. I want it to help me ask better questions.” Twelve months later Terry, who began by clicking first and worrying later, was teaching AI classes to other U3A members. “It’s not about the AI,” he said. “It’s about the question.” What began as Helen’s caution had become part of the group’s shared practice.
But nobody stayed the expert for long - AI has a habit of rearranging the furniture overnight. Interfaces changed, subscriptions shifted, and yesterday's expertise became today's workaround. The durable achievement was the group's capacity to stay in play, showing unfinished work, comparing outputs, refusing premises and trying again. The person needing help one session, often solved in the next.
Safety was essential. Scams, insecure health data and surveillance are real risks. Yet, much of the time the people I worked with were evaluating, challenging and sometimes refusing the technology to which they were supposedly vulnerable. Prefabricated checklists alone don’t build ethical capacity in use. In our group, it grew through conscious communal repetition: checking before believing, arguing before using, and learning collaboratively.
For ageing research, the implication is straightforward: formal safeguards and support for early-career researchers should be paired with peer environments where older people, practitioners and researchers can experiment together. Co-design is not a substitute for governance, but a way in which ethical judgement becomes practiced and shared. Policies frozen in time won’t do it. One-off training won’t do it. Human oversight, wherever I saw it working best, was collective.