Abstract
Evolutionary artificial life systems have demonstrated many exciting
behaviors. However, there is a general consensus that these systems are
missing some element of the consistent evolutionary innovation that we
see in nature. Many have sought to create more “open-ended”
evolutionary systems in which no stagnation occurs, but have been
stymied by the difficulty of quantifying progress towards such a
nebulous concept. Here, we propose an alternate framework for thinking
about these problems. By measuring obstacles to continued innovation, we
can move towards a mechanistic understanding of what drives various
evolutionary dynamics. We propose that this framework will allow for
more rigorous hypothesis testing and clearer applications of these
concepts to evolutionary computation.