EvoEco is an experimental piece of software for the creation of abstract images using simple virtual drawing creatures. Its an ecosystemic approach to developing new forms of the interactive genetic algorithm or IGA. In this system agents roam a canvas modifying the pixels they encounter. The way agents interact in this drawing ecosystem is subject to evolution using a modified version of the IGA. This IGA offered three different versions of search: random, creativity-search and phenotypic distance search. The creativity search (which we call “creativity lite” tries to consistently offer novelty in the image feature space.
As part of the research we blind tested these different search methods in a web-based applet run on the internet and in a gallery as part of the Design After Nature exhibition to try and establish which method produced the most interesting suggestions to users of the IGA. Our results seem to statistically support the use of the “creativity-lite” search as a good method to interactively explore a large generative space of phenotypes.
Here is a flickr set of images that users of the system evolved over the period September 2009 – January 2010.
And a video created by Taras:
EvoEco was created by Taras Kowaliw, Alan Dorin, Jon McCormack at the Centre for Electronic Media Art at Monash University, along with ninety-six anonymous artists. Taras was the author of the software.
Our IEEE Transactions on Evolutionary Computing paper on this work is available here.
For more information see the project webpage.