AI is increasingly shaping many aspects of healthcare, from predicting disease trajectories to identifying patterns within large health datasets and personalising treatment plans. As these technologies become more influential, a fundamental question arises: who gets to shape the research behind them? While patient and public involvement (PPI) has long been recognised as an important part of healthcare research, applying it to AI and big data projects presents a unique set of challenges. Recent research suggests that meaningful involvement requires more than simply inviting patients into the room. It requires new ways of working that bridge technical complexity, differing priorities, and questions of trust.
PPI in Healthcare Research
Public and Patient Involvement (PPI) refers to the active collaboration between healthcare researchers and individuals with lived experiences, patients, carers, service users, and members of the public, throughout the research process (HSE: Patient and Public Involvement in Research (PPI)). This partnership encompasses the planning, design, execution, and dissemination phases of research, ensuring that studies are conducted with or by the public, rather than to, about, or for them. The perspective of the Irish Platform for Patients Organisations, Science & Industry (IPPOSI) is that there should be no debate about whether to involve patients or not – there should be a discussion about the best way to involve patients (IPPOSI: Patient & Public Involvement).
Healthcare research support and funding increasingly emphasise the importance of planning for PPI from the earliest stages when planning their research (HRB: Public and patient involvement in research) while academic institutions such as RCSI champion patient and public partnership as central to healthcare development for students, thereby establishing PPI principles for future healthcare leaders (RCSI: PPI in research and engaged research).
Why Public Involvement Matters for AI
Healthcare AI systems are often developed using large datasets such as electronic health records, raising significant concerns around privacy, fairness, transparency, and bias. Researchers have increasingly argued that patients and members of the public should not be viewed as passive recipients of these technologies, but as active stakeholders in their development and governance.
Public involvement can help ensure that AI research remains focused on outcomes that matter to patients. It can also provide valuable scrutiny of assumptions that researchers may overlook, particularly when technologies are designed around available data rather than lived experience. As AI systems become more embedded within healthcare services, involving those affected by them may be essential for building trust and social legitimacy.
When Patient Priorities Meet Data Reality
A recent study from the AI-Multiply Consortium (Thompson et al., 2025) explored public and patient involvement within a large UK research programme using AI and routine health data to investigate multiple long-term conditions. The researchers identified an important challenge that emerged repeatedly throughout the project: patients and researchers often prioritised different outcomes.
Public contributors consistently highlighted outcomes linked to everyday life, including quality of life, independence, wellbeing, and periods of wellness. Researchers, meanwhile, often focused on measures that were readily available within electronic health records, such as hospital admissions, medication use, or mortality. This difference reflected a practical limitation of many healthcare datasets. Some of the outcomes that matter most to patients are simply not recorded in routine health records, making them difficult to study using existing AI methods. As a result, public contributors sometimes found that their suggestions could not easily be translated into research questions.
The study suggests that an important aspect of involving patients in AI research is to help them understand both the possibilities and limitations of the available data from the outset.
Learning on Both Sides
Another key finding was that meaningful involvement required learning and adaptation from both researchers and public contributors. Many contributors initially felt uncertain about engaging with concepts such as machine learning, data linkage, or predictive modelling. Researchers, meanwhile, often questioned how public perspectives could contribute to highly technical projects.
Over time, these concerns began to fade. Introductory training sessions helped contributors build confidence, while researchers increasingly recognised the value of patient perspectives. Some public contributors were even integrated into technical meetings that would traditionally be considered inaccessible to non-specialists, including data engineering discussions and research planning sessions. These collaborations often challenged researchers to think differently about their assumptions, communication styles, and research priorities. The study found that trust, familiarity, and repeated interaction were central to creating a genuinely collaborative environment.
The Basics Still Matter
Although AI and big data present new challenges, the researchers reached an interesting conclusion: many of the barriers they encountered were not unique to AI at all. Contributors repeatedly highlighted familiar issues such as unclear expectations, insufficient feedback, limited discussion time, and uncertainty about how their input had influenced decisions.
These findings echo longstanding guidance on good public involvement. Clear communication, adequate preparation, opportunities for meaningful participation, and transparent feedback remain just as important in AI research as they are in other areas of healthcare research. The challenge can be less about creating entirely new frameworks and more about consistently applying existing best practice within complex, interdisciplinary projects.
Towards More Inclusive AI Research
As healthcare AI continues to evolve, public involvement is likely to become increasingly important. Responsible AI encompasses ensuring that the people affected by these technologies have strong opportunities to influence how they are developed, evaluated, and implemented. The experience of the AI-Multiply Consortium suggests that meaningful involvement is possible even in highly technical research environments, provided adequate support, leadership, and opportunities for collaboration are in place.
HRB: Public and patient involvement in research. Retrieved 09/09/2026 from https://www.hrb.ie/funding/responsible-research-assessment/public-and-patient-involvement-in-research/
HSE: Patient and Public Involvement in Research (PPI). Retrieved 09/09/2026 from https://hseresearch.ie/patient-and-public-involvement-in-research/
IPPOSI: Patient & Public Involvement. Retrieved 09/09/2026 from https://ipposi.ie/work/patient-public-involvement/
RCSI: PPI in research and engaged research. Retrieved 09/09/2026 from https://www.rcsi.com/society/engage/ppi-research
Thompson, A., Bartle, V., Remfry, E. A., Reynolds, D. J., Barnes, M. R., Reynolds, N. J., & Hanratty, B. (2025). Public and patient involvement in artificial intelligence and big data healthcare research: an exploration of issues and challenges within the AI‐multiply project. Health Expectations, 28(6), e70490.
Photo: Alan Warburton / https://betterimagesofai.org / © BBC / https://creativecommons.org/licenses/by/4.0/
