A European project is attempting to rethink how technologies for cognitive health are developed: no premature patient exposure, direct involvement of clinicians, and constant attention to limits, risks, and social implications.
By 2050, Europe will count 152 million people over the age of 65. Neurodegenerative diseases are already rising, and dementia – currently affecting more than 10 million Europeans – is often diagnosed when meaningful action is no longer possible. Artificial intelligence is transforming fields such as oncology and cardiology, yet research on cognitive decline is progressing more slowly. The reasons are not solely biological: they stem largely from how technological tools are designed and tested, and from their compatibility with unresolved ethical and social dilemmas.
This is the context in which COMFORTage emerges, a four-year Horizon Europe project involving 39 partners across 12 countries. Its purpose is not to build the most accurate predictive algorithm. Instead, it seeks to invert the dominant logic of AI development in healthcare, bridging the structural gap between laboratory research, clinical practice, and patients’ lived experiences.
TAKEAWAYS
The innovative core of COMFORTage lies in this reversed validation process. Engineers do not hand a tool to clinicians for testing. Instead, they observe how physicians work with patients following their usual protocols. Only afterwards do clinicians re-examine the same cases using the algorithms, comparing the machine’s output with their own expert judgement.
“We have brought the hospital inside the research and development laboratory, not the opposite,” explains Lorena Volpini, Ethics Lead at CyberEthicsLab. “Clinicians are not beta-testers of a finished product; they are co-developers from day one.”
The goal is to reduce the risk of designing AI systems that are detached from the complexity of medical practice.
Alongside this approach, the project gathers a substantial volume of data: historical clinical records, authorised biobanks, and real-time information from wearables and home sensors (sleep, physical activity, body composition, daily habits). Partners are also conducting clinical studies on modifiable risk factors, lesser-explored correlations (such as swallowing or oral health), and possible lifestyle interventions to delay disease onset. Some countries participate with living labs conducting community-based experimentation.
But the most critical dimension is ethical. Predicting an incurable condition is not a neutral act.
“We are medicalising ageing,” Volpini warns. “Without a carefully designed approach, the risk is psychologically devastating healthy individuals, enabling discrimination by insurers or employers, and shifting attention from social determinants to personal responsibility.”
To turn these concerns into operational requirements, COMFORTage adopts a multi-layered ethics-by-design framework. CyberEthicsLab supports the entire process:
– ensuring consent forms are genuinely understandable, even when patients never interact directly with AI;
– translating broad principles (transparency, human oversight, non-discrimination) into precise technical specifications: mandatory alerts, traceable decision pathways, and the ability for clinicians to halt the algorithm at any point;
– producing policy briefs when the issue is social or regulatory rather than technical, raising questions that no code can resolve: Who should communicate an incurable risk? Where do the data go? Who will have access, and for what purposes?
Another concern is the risk of reducing the person with dementia to a mere “user” of a device. Turning a complex experience into interaction metrics risks overlooking the lived reality of patients, caregivers, and families.
“Technology can help, but it does not solve loneliness, poverty, chronic stress, or pollution,” Volpini notes. For this reason, the project aims to involve patient and caregiver associations in Greece, Spain, and Italy – even within the constraints imposed by clinical ethics. Their contribution highlights structural limits of technology and invites a broader reflection on the very paradigm of prevention.
Why focus almost exclusively on lifestyle factors when many risks stem from environmental or socioeconomic conditions individuals cannot change alone?
COMFORTage will conclude in 2027. It is unlikely to produce a market-ready solution in the short term. The intention is not to accelerate but to demonstrate that an alternative model of innovation is possible: slower, more resource-intensive, but potentially more robust and more acceptable to clinicians, patients, and society as a whole.
“It requires ongoing translation between worlds that usually do not speak to each other,” Volpini acknowledges. Yet the alternative is well known: sophisticated tools that no one uses, scandals that erode public trust, technically advanced but clinically irrelevant solutions.
In an environment dominated by the mantra “move fast and break things”, COMFORTage consciously chooses not to rush. It slows down to preserve what matters most: people’s trust and the responsibility owed to those who may one day face an irreversible diagnosis.
The question the project leaves open is not simply whether we can build an algorithm, but something more fundamental: should we build it? If so, how? And above all, with whom?
These are the questions that increasingly shape innovation in healthcare – and will determine both its legitimacy and its real value.
