Automated Ontology Evolution

Most work on automated reasoning is based on manually-built representations of knowledge. There is, however, suggestive evidence that human representations evolve both between and within episodes of reasoning. Unfortunately, how to automate the evolution of representations is a neglected problem in informatics. Challenges such as managing: the interaction of multiple agents; and the co-development of programs and their specifications, have it urgent to address this problem. Our improved understanding of representations and their relationships to reasoning techniques has made it timely. I will describe and motivate the problem and explain why it is difficult. I will outline some current approaches to this problem in research at Edinburgh and elsewhere.