ABSTRACT
G eneralization-reasoning from the properties of observed entities to those of entities as yet unobserved-is at the heart of many aspects of human cognition. As several of the chapters in this book indicate, generalization plays a key role in language learning, where learners need to make judgments about the linguistic properties of utterances (such as their meaning or grammaticality) using their previous experience with a language. It also underlies our ability to form and use categories, allowing us to identify which objects are likely to belong to a category based on a few examples, and is central to learning about causal relationships, where we predict how one event will inuence another by drawing on past instances of those events. Language learning, categorization, and causal induction are three of the most widely studied topics in cognitive science, and allow us to communicate about, organize, and intervene on our environment. They are also three examples of problems for which human performance exceeds that of automated systems, setting the standard to which articial intelligence and machine learning research aspires. This raises a natural question: What makes people so good at generalization?
