Before AI: Building knowledge with people at the centre
Author: Liliana Escada Ribeiro1,2, FIlipe Pereira3, Rafael Bernardes4
1. Universidade Católica Portuguesa
2. Unidade Local de Saúde de Coimbra
3. Nursing School - University of Porto
4. Nursing School, University of Lisbon
What knowledge feeds AI?
Artificial intelligence (AI) is rapidly transforming healthcare. However, when we talk about AI, the discussion continues to focus on algorithms, overlooking an essential question: What knowledge are these systems using? And does that knowledge reflect what really matters to people?
The answers generated by AI depend not only on the quality of data but also on the knowledge that gives those data meaning, how that knowledge is represented and the evidence that supports it. After all, we know that data alone are not knowledge. As described in a reflection on AI in nursing, this journey begins with data, moves through information and only then becomes knowledge.
Why evidence is not enough
When I started my PhD in Nursing, this question stayed with me. At first, I thought that one of the main challenges would be to synthesise the best available scientific evidence. As my research proceeded, I realised that the real challenge was different. Producing or synthesising evidence was not enough. I also needed to consider how nursing knowledge could be represented without losing its disciplinary meaning, so that it could be used to support clinical decision-making.
Evidence, on its own, is not enough. It can only inform and improve practice when it is transformed into knowledge that can be understood, shared and used in practice by nurses, information systems and, in the future, AI applications.
This led me to another question: How can nursing knowledge based on the best available evidence be represented so that it can be used in practice today and by AI in the future?
But if we expect AI to support person-centred care, how do we ensure that the knowledge supporting it includes not only scientific evidence but also people’s perspectives, needs and goals?
From evidence to AI
Between the best available evidence and its use in information systems, and eventually AI, there is an essential step: representing nursing knowledge.
This means organising knowledge in a structured way while preserving how nurses document care, assess, make decisions and plan care. Only then can that knowledge be understood, shared and used.
This is where clinical terminologies and clinical data models play an important role. They help organise knowledge so that it can be understood, shared and used across clinical practice. A study conducted across four hospitals showed how standardised nursing terminologies supported the implementation of an AI-powered tool for fall prevention.
This is a challenge we are already facing in clinical practice. Increasingly, people are turning to AI with questions about their health, sometimes before speaking to a healthcare professional. Although many of the answers may sound convincing, they are not always complete or clinically appropriate. An article on personalised patient education highlights this challenge and the importance of ensuring that health information is accurate and based on scientific evidence.
An AI-generated answer is therefore no guarantee of reliable knowledge. If the knowledge on which it draws is incomplete, poorly represented or loses its clinical meaning, the answers generated by AI may reflect the same limitations.
The challenge is not simply to develop increasingly sophisticated algorithms. We must also ensure that the knowledge supporting information systems and AI is rigorous, grounded in the best available evidence, preserves the disciplinary meaning of nursing and reflects what truly matters to the people we care for.
Putting people at the centre of knowledge
As my research proceeded, I realised that representing nursing knowledge was only part of the challenge. Just as important was ensuring that this knowledge reflected what really matters to the people we care for.
This made me realise that the knowledge we use to care for people cannot be based only on evidence and nurses’ professional experience. It must also include what people themselves experience and what matters to them about their health.
When we talk about AI in healthcare, we talk about algorithms, models and innovation. But we risk forgetting that the quality of AI begins long before the technology itself. It begins with how we build the knowledge that will support it.
A recent article on relational AI in nursing care reminds us that, even as technology advances, we must continue to recognise each person as unique and understand what their experience means to them. Caring is not simply about responding to data or diagnoses. It is also about understanding each person´s experience.
Before we can have person-centred AI, we need to build person-centred knowledge.
If we want person-centred information systems and future AI applications, people themselves need to be involved in building the knowledge on which those systems are based.
This shift in perspective has come to guide my PhD research. My aim is to contribute to the development of a nursing clinical data model that represents knowledge grounded in the best available evidence while also incorporating what matters to people.
Perhaps the real challenge of AI is not to develop smarter algorithms, but to ensure that people are present in the construction of the knowledge that will feed information systems and, eventually, AI itself. Only then can we truly talk about evidence-based, person-centred technology.
Lessons learned
Throughout this journey, I have learned that:
- The greatest challenge was not finding the best available scientific evidence but understanding how to transform it into knowledge that can be used in practice
- Putting people at the centre also means involving them in building knowledge
Key messages
- Evidence can only improve care when it is transformed into knowledge that can be used in practice
- The quality of AI will always depend on the quality of the knowledge that supports it
- If we want person-centred AI, we need to involve people in building the knowledge that will support it
References
Cho, I., Cho, J., Hong, J. H., Choe, W. S., & Shin, H. (2023). Utilizing standardized nursing terminologies in implementing an AI-powered fall-prevention tool to improve patient outcomes: A multihospital study. Journal of the American Medical Informatics Association, 30(11), 1826–1836. https://doi.org/10.1093/jamia/ocad145
McGrow, K. (2025). Artificial intelligence in nursing: A journey from data to wisdom. Nursing, 55(4), 16–24. https://doi.org/10.1097/NSG.0000000000000165
Neves, H., Pereira Sousa, J., Seguro, J. M., Bernardes, R. A., Caldeira, S. M. A., Shapira, R., Oliveira, I. J., Coelho, A., & Parola, V. (2026). Travelbee and relational AI: A triadic model for nursing. Nursing Outlook, 74, 102878. https://doi.org/10.1016/j.outlook.2026.102878
Queirós, C., & Neves, H. (2026). Personalised patient education in orthopaedic and trauma nursing: The promise and challenges of artificial intelligence. International Journal of Orthopaedic and Trauma Nursing, 62, 101302. https://doi.org/10.1016/j.ijotn.2026.101302
To link to this article - DOI: https://doi.org/10.70253/BKIV1474
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
