Studying Creativity via Computational Modelling

One way to investigate creativity is to build artificial agents which might be capable of creative output -- perhaps basing them on theories of human cognition and creativity -- and see how well they work. This talk will describe recent and ongoing work on a range of projects in the Computational Creativity Lab at Queen Mary University of London, explain how they relate to general models of human cognition, and discuss what insights they can give us. We will examine general statistical models of sequential and hierarchical learning, discuss how they can be connected to higher-level conceptual structures, and show how they can be applied to model language and music while generating interesting, novel outputs. Matthew Purver is Reader in Computational Linguistics at Queen Mary University of London