Bio: Giovanni was trained as mathematician and a philosopher: after a MSc in Philosophy of Science at LSE he went on to complete a PhD in Mathematical Logic at the University of Amsterdam. He subsequently joined Pacmed as a Data Scientist, working on predictive models for chronic diseases. His current main interests are explainable and fair AI, in particular in the healthcare domain.Back to speakers list
Explainable medical AI: pain points and painkillers
...why should I get surgery? Black-box AI is particularly problematic in healthcare, where software could make suggestions that impact the well-being of patients. Starting from user experience on real-life examples, we single out the most asked questions and discuss desirable types of explanations for users without deep knowledge of machine learning. We showcase how different methods and packages can come together in a Python interpretability toolbox and describe at which stages of development they should be used. These considerations generalize to the development of any AI-powered product.