How do we generalize

Humans and animals are able to generalize information from previous experience so that they can behave appropriately in novel situations -- but what neural mechanisms give rise to this amazing ability? This Generative Adversarial Collaboration (GAC) started out with the aim of arbitrating between two alternative accounts: (a) memories for past experiences are integrated in the brain to allow for such generalization, versus (b) generalization is computed on-the-fly using separately stored (unintegrated) memories. The members of this GAC come from a range of academic backgrounds and across the course of this collaboration we found that -- although we often use different terminology and techniques, and although some of our specific papers may provide evidence one way or the other -- in reality we are generally in agreement that both of these accounts (as well as others) might be valid. We believe that different mechanisms might be weighted according to specific conditions or task factors, or even implemented in parallel in the brain. Here, we introduce some of the main mechanisms that have been proposed to underlie generalization, we discuss the issues currently hindering better synthesis of generalization research, and we propose the way forward in tackling some of these issues. Finally, we introduce some of our own research questions that have arisen over the course of this GAC, that we think would benefit from future collaborative efforts.