Building Knowledge Together: Citizen Science, Communication, and Trust | 5 - 6 October 2026, Berlin

1. Context: AI in critical care research

Artificial intelligence (AI) is being increasingly integrated into clinical research and care, including in resource-limited settings. I explore how these developments challenge existing models of participation, communication and trust, drawing on practice-based work at OUCRU Vietnam. I argue that AI-enabled clinical research in critical care creates a form of constrained participation, in which patients contribute to knowledge production but cannot meaningfully engage with or understand it.

OUCRU is based at the Hospital for Tropical Diseases (HTD), a tertiary hospital for infectious diseases in southern Vietnam. Its intensive care unit (ICU) operates under significant constraints, including high patient loads and limited monitoring equipment. Tetanus, for instance, has a mortality rate of 30 to 40 per cent globally and requires continuous monitoring and rapid intervention, placing substantial pressure on clinical staff. In response, OUCRU has developed an affordable, wearable monitoring device that continuously captures patients' vital data, and uses an algorithm to detect early signs of deterioration and alert clinicians.

1. Constrained participation and trust

Patients enrolled in this research contribute continuous streams of vital data that inform both clinical decisions and knowledge production. Their contribution constitutes a form of participation in scientific research, albeit one that is passive and mediated by clinical care. This raises important questions about what counts as participation in such data-intensive research.

In this setting, participation is true but highly constrained: patients are severely ill, often unconscious, and consent is typically provided by family members. Clinicians and researchers, who are responsible for communicating with patients and families, are not always equipped to explain the "black box", how algorithmic systems function or what their limitations are. This challenge is intensified by Vietnamese culture where patients and families defer strongly to medical authority and have limited reference points for understanding AI. Trust, in this case, is shaped by necessity and structural dependence rather than informed engagement.

1. AI disclosure

Unlike auxiliary diagnostic tools, AI technologies in ICUs do not simply generate data but actively shape clinical decisions. This has direct implications for patients' autonomy and dignity (Montanari Vergallo et al., 2025). Thus, the use of AI must be adequately communicated, even when full understanding is proven challenging.

The question is how much information to communicate, and how. One approach is to match the obligation to inform to the level of risk on patient safety, where higher risk systems require more extensive explanation. However, simply providing more information is not sufficient. Transparency alone does not build trust; information must be intelligible and accessible to those who need to act on it (O’Neill, 2018). In this setting, the gap between what the technology does and what patients and families can meaningfully understand remains significant.

1. The role of science communication

I would argue that science communication has a critical role in addressing this gap. At OUCRU, communication has traditionally focused on translating complex research findings to make them accessible and support public trust in science. However, in the context of AI in clinical research, communication becomes a condition for ethical practice rather than a final step.

The challenge is to develop forms of explanation that are accurate, accessible, and usable within the realities of critical practices. This requires moving upstream, from disseminating results to shaping how research is explained and understood at the point of participation.

1. Practice contribution

I propose a practice-based inquiry into how AI is currently communicated to patients and families across several OUCRU clinical studies, with the aim to develop practical guidance for clearer and more responsible communication during consent and care discussions.

This includes identifying what information is necessary, how it can be explained in plain language, and developing practical tools such as communication scripts, checklists, and training for healthcare workers that can be integrated into clinical practice.

By examining a setting where participation is constrained but stakes are high, this work contributes to a broader discussion on citizen science, communication, and trust. It argues that as AI becomes more prominent in research and care, new communication practices are needed to support understanding and maintain trust throughout the research process and beyond.