Explore challenges and techniques for evaluating machine learning models when data is not independently and identically distributed (IID) in this 48-minute talk by Indrė Žliobaitė from the Finnish Center for Artificial Intelligence FCAI. Delve into three main settings: spatially autocorrelated data, concept drift over time, and phylogenetically non-independent observations. Gain insights into knowledge discovery and generalization from machine learned models, with practical examples from macroecology and industrial process control. Learn from Žliobaitė, an associate professor at the University of Helsinki, who leads research on data science for understanding evolutionary processes in nature and society, and has contributed to evolving data learning, fairness-aware machine learning, and macroevolutionary analyses.