Frank Ursin¹, Monika Taddicken², Tim Kacprowski³

¹ Institute for Ethics, History and Philosophy of Medicine, Hannover Medical School

² Institute of Communication Studies, Technical University of Braunschweig

³Department of Data Science in Biomedicine, Technical University of Braunschweig

Artificial intelligence is transforming research in medicine and the life sciences at a pace that poses challenges for ethical assessments and the methods used to conduct them. The search for new pharmaceutical compounds and the prediction of individual disease risks using polygenic risk scores are made considerably easier by AI. Agent-based AI systems such as Google’s Co-Scientist can also support researchers in their substantive work by highlighting complex interrelationships or suggesting hypotheses. How do researchers assume responsibility when publishing AI-based results – from questions of authorship and the disclosure of their own contribution in the face of new AI assistants, to weighing up the societal consequences of their research? Alongside these new possibilities, old challenges are also coming back into focus, as AI can be used to develop not only medicines but also poisons: how do we mitigate dual-use risks when the same AI methods can be both beneficial and harmful?

The BMFTR project KILEWI uses horizon scanning to identify early signals, blind spots and controversial expectations regarding the use of AI, and to develop content suitable for discussion in both academic and public contexts. In this talk, we will discuss the time-related methodological challenges of this approach: anticipation is largely fuelled by perception – by the observations and interpretations of researchers, which are shaped by hype cycles, funding logics and publication dynamics. As a result, classic challenges of technology assessment are entering a phase of accelerated change: whilst concepts, data, models and methods are changing rapidly, ethical, legal and social assessments remain sluggish, fragmented or overly simplistic.

Methodologically, we therefore combine a rapid scoping review with guided expert interviews (biology, medicine, biotechnology) in order to ultimately use the insights gained to conduct a horizon scanning exercise on the ethical and social challenges of AI use in life sciences research. In this presentation, we will outline the methodological challenges of this approach.