Lecture Series “Robust AI”: Kacper Sokol

The next lecture in our lecture series will be given by Dr. Kacper Sokol on the topic of “Naturalistic Explainable Artificial Intelligence for Clinical Reasoning Support Systems”.

When & where:

Wednesday, August 28 2024, 4:15pm at Bielefeld University (Room CITEC-1.204) or online. Join us via Zoom https://uni-bielefeld.zoom-x.de/j/64775735478?pwd=TFpEUVFPME5EQXFKMHZHY1ZsM2Y4Zz09

Abstract:

In this talk I will discuss the state of explainable artificial intelligence and interpretable machine learning in the context of medical applications. Among others, I will highlight the technical and socio-technical barriers to adoption of AI in healthcare and show how explainability falls short of its promises in this domain. 
I will then identify promising research directions for technical and human-centred aspects of explainability systems, also briefly touching upon the challenges of evaluating AI explainers. The discussion will be grounded in how explainability fits into real-life clinical workflows, drawing a distinction between decision- and reasoning-support systems. The talk will culminate in an introduction of an alternative perspective on AI explainability – which I call naturalistic explainable artificial intelligence – that envisages supporting epistemic and cognitive functions of humans rather than providing them with abstract explanations of data-driven decisions.

Speaker:

Kacper Sokol is a research fellow at the Medical Data Science group at ETH Zurich. He has also been working with Peter Flach at University of Bristol. Google scholar: https://scholar.google.com/citations?user=k-oiaCkAAAAJ&hl=en