Nicola Gnecco's research aims to develop statistical and machine-learning methods for reliable learning under changing conditions, drawing on causal inference and extreme value theory to improve robustness and understand the limits of extrapolation.