We want to implement a variety of computer programs capable of generating unbounded novelty via emergent evolution. We also want a principled, structured way of analysing our novelty-generation programs and improving them over time.
We have developed a definition of embodiment in terms of a conceptual framework of hierarchical phenomena, mechanisms, and worlds, for use in analysing and building novelty-generation programs. We present these concepts and demonstrate them on two existing artificial chemistry systems: Stringmol and GraphMol. Our two demonstration systems behave in very different ways. We describe these differences, explaining why they arise in terms of our framework. The systems better able to generate novelty have more embodied implementations.
@inproceedings(NellisStepney2014, author = "Adam Nellis and Susan Stepney", title = "Computational novelty: Phenomena, mechanisms, worlds", pages = "506-513", doi = "10.7551/978-0-262-32621-6-ch081", crossref = "ALife14" ) @proceedings(ALife14, title = "ALife XIV, New York, NY, USA, July 2014", booktitle = "ALife XIV, New York, NY, USA, July 2014", publisher = "MIT Press", year = 2014 )