The current project seeks to explore the cognitive, computational and neural mechanisms governing lifelong learning. It will elucidate how both humans and artificial systems adapt to novel challenges while retaining knowledge from prior experiences. We hypothesize that effective lifelong learning entails the seamless integration of novel information with existing knowledge, a challenge that has been largely overlooked in both cognitive and computer science. This project will investigate the role of stimulus, task, and contextual diversity and overlap for efficient learning and retention of information. We compare several neural network models with human performance and neurophysiological signatures (EEG, fMRI). Additionally, we consider how such diversity interacts with active learning, that is, a learning setup in which the agent can decide what task to carry out next. This will allow constructing optimal training schemes (curricula) for humans and machines.