Object replacement and object composition in a creative cognitive system. Towards a computational solver of the Alternative Uses Test

In creative problem solving, humans perform object replacement and object composition to improvise tools in order to carry out tasks in everyday situations. In this paper, an approach to perform Object Replacement and Object Composition (OROC) inside a Creative Cognitive framework (CreaCogs) is proposed. Multi-feature correspondence is used to define similarity between objects in an everyday object domain. This enables the cognitive system OROC to perform creative replacement of objects and creative object composition. The generative properties of OROC are analysed and proof-of-concept experiments with OROC are reported. An evaluation of the results is carried out by human judges and compared to human performance in the Alternative Uses Test.