- 09:20 – 09:30 Welcome and General Introduction (Organizers)
- 09:30 – 10:30 Keynote: Michèle Barbier (Inria and Independent Ethics Expert for the European Commission) – Beyond Optimization: Preserving Human Dignity in the Era of Generative AI, Virtual Twins, and Responsible Innovation in healthcare (abstract below)
- 10:30 – 11:00 Morning Coffee Break
- 11:00 – 11:20 Assessment of LLM-Based Simulated Virtual Patients in Interdisciplinary Clinical Simulations (Sofia Martinelli)
- 11:20 – 11:50 Is it time for ethical boards in top-tier machine-learning conferences? A case-study on an Interspeech challenge about children suicidability. (Vincent P. Martin)
- 11:50 – 12:00 The Invisible Patients: A Systematic Review of Quiet Harms and Intersectional Bias in Clinical Large Language Models (Hana Tamiru)
- 12:00 – 12:30 False Consensus in Multi-Agent Clinical AI: Ethical Risks of Collusion and Verifier-Based Defences (Adeela Bashir)
- 12:30 – 14:00 Lunch Break
- 14:00 – 14:30 Talking as a Drug – Administering LLM Conversations for Medical Treatment in Elder Care (Tim Hallyburton)
- 14:30 – 14:35 Tensions Between Biomedical Ethics Review and Responsible AI Research at a University Hospital in South America (Sofia Martinelli)
- 14:35 – 15:05 Assessing Validity and Bias in Qualitative Clinical Analysis: LLM-Simulated Personas vs. Human Experts on the Borderline Experience (Marcin Moskalewicz)
- 15:05 – 15:30 Sharing experience: AI to support medical treatment – ethical aspects beyond the technical implementation (Roland Roller)
- 15:30 – 16:00 Afternoon Coffee Break
- 16:00 – 16:30 Next Ethicaia – Call for Action (Organizers and Attendees)
Keynote Abstract:
The dual rise of Virtual Human Twins (VHTs) and Generative AI in digital medicine demands a radical ethical vigilance to preserve human dignity and clinical accountability. While VHTs offer powerful predictive simulations, Generative AI introduces a new frontier of risk: the potential for “hallucinations” in clinical advice, the generation of synthetic patient data, and the erosion of the human dimension in patient interactions. The primary risk remains the dilution of responsibility. In a system where a Generative AI model suggests a treatment plan or a VHT simulates a surgical outcome, we must clearly define whether liability rests with the developer, the clinician, or the institution when a decision causes harm. Beyond legal liability, medicine cannot become a purely optimization-driven business; it must remain a human practice grounded in compassion, presence, and moral responsibility—qualities that algorithms, whether predictive or generative, fundamentally lack.
To operationalize these principles, strict governance frameworks must evolve to address the unique nature of Generative AI. Aligning with the EU AI Act (1) and utilizing tools like the Assessment List for Trustworthy AI (ALTAI) (2) is essential, but we must go further. We need specific protocols for validating the training data of large language models to prevent the propagation of historical biases in healthcare. The STANDING Together consensus recommendations (3) provide the critical methodological backbone for this data validation, ensuring that datasets are diverse, transparent, and auditable. Together, these regulatory and technical standards allow us to translate ethics into actionable protocols, ensuring that Generative AI is used to augment human intelligence rather than replace clinical judgment, and that it does not compromise patient privacy.
We must also guard against the slide toward elitist medicine or a “healthcare divide” where advanced generative tools benefit only the privileged. The risk of “doing without understanding” is acute with Generative AI; clinicians must be trained to critically evaluate AI outputs rather than accepting them as absolute truth. Drawing on the philosophical warnings of Hans Jonas and Hannah Arendt, we must ensure that technological capability never outpaces our moral comprehension. There is an urgent need for interdisciplinary collaboration among clinicians, data scientists, and policymakers to foster education in ethical AI, focusing on transparency, explainability, and the “human-in-the-loop.”. In conclusion, as the power of VHTs and Generative AI grows, our ethical frameworks must evolve in tandem to ensure this technology serves humanity. The goal is not to let technology define medicine, but to use these powerful tools to preserve the essential human elements of care—empathy, ethics, and responsibility—while navigating the complex, evolving regulatory landscape of global digital health.