HealthDARES - Day 2 Recap
Day 2 of the course delved into critical areas of AI ethics, including emerging codes of practice, toolkits for AI governance, and data governance.
Day 2 of HealthDARES dived deeper into the ethical complexities of AI in healthcare. Presentations and discussions throughout the day covered key debates around bias, explainability, and fairness, with an emphasis and focus on practical tools and codes of practice to address challenges and ensure the equitable distribution of AI benefits.
Key Highlights:
- Ethical Considerations: Alessandro Blasimme (ETH Zurich) opened the day with a presentation on the key debates related to AI and the use of big data in healthcare. His talked covered the importance of clinical validity, explainability, fairness, trustworthiness and oversight.
- WHO Guidelines: Rohit Malpani from the World Health Organization (WHO) followed up with a presentation of the WHO's guidelines for AI governance in healthcare. His talk provided guidance on how Large Language Models (LLMs) can be governed, and how to promote equitable access to AI.
- Practical Approaches: The day then moved to explore practical approaches to ethical AI govern ance. Mikale Jensen (D-mærket/D-seal) presented Denmark's "D-seal" strategy and explained how labels can help foster organizational accountability and responsible data use. Sara Kijewski (ETH Zuruch) then gave a comprehensive overview of the plethora of AI ethics instruments, outlining the current trends and challenges inherent to the toolbox.
- Data Governance: The afternoon session focused on data governance challenges in healthcare. Eric Sutherland (OECD) shared the OECD’s key principles and guidelines to facilitate data access and sharing, and responsible AI. His talk further addressed the question of how global collarboation can advance AI innovation in health. Subsequently, Robin Wirz from the Digital Trial Innovation Platform (DTIP) introduced the DTIP’s data governance for clinical trials, alongside a reflection on the current challenges, unresolved issues and areas of complexity in this context.
- Industry Perspectives: Agata Ferretti from the AI Alliance then provided insights from the private sector on challenges and opportunities for open and trusted AI models and datasets. She introduced who the AI Alliance is and what exactly they are doing to promote responsible AI.
The day concluded with group discussions around a case of AI-assisted prostate cancer treatment. Groups’ talked about dilemmas of transparency and patient autonomy, trust, and the physician and patient relationship amongst other topics.
