From: Anna Cavender (cavender@cs.washington.edu)
Date: Wed Dec 01 2004 - 12:02:17 PST
The evolutionary origin of complex features
Richard E. Lenski, Charles Ofria, Robert T. Pennock, Christoph Adami
One line summary:
This article describes the creation and evaluation of digital organisms:
objects that can reproduce, mutate, and compete in a digital environment
simulating the evolutionary theories of Darwin.
The two most important ideas in the paper:
Digital organisms mutate by imperfectly copying their instructions
resulting in two different child organisms. They receive rewards of
“energy” when mutations result in specific, predefined logic
instructions of varying complexity. The energy is needed for the
organism to execute instructions. In this way, only the “good” organisms
acquire enough energy to continue execution. This implicit selection
allows a positive evolution where descendants obtain more and more
complex features.
Experiments show that achievement of simple functions was often a
prerequisite to more complex features and that often backward or harmful
mutations actually yielded highly beneficial mutations in later descendants.
The authors claim that these findings confirm evolutionary theories
proposed by Darwin.
The one or two largest flaws in the paper:
Darwinian theories argue that evolution is based on survival of the
fittest; however this paper based evolution on the successful execution
of predefined logic functions in a static environment. While their
findings are quite interesting and mesh well with Darwinian Theory, the
reader is left wondering about the ecological validity of these
findings. First, it is unclear that predefined rewards occur in nature,
and second, natural evolution occurs in a changing environment.
Another flaw may be a small point, but a point I found unclear. Since an
organism consists of a list of instructions, it seems that logic
instructions (functions) could persist through mutations even though
other functions are acquired. This is confirmed in the description of a
situation where the acquisition of new instructions outweighs the loss
of existing instructions. However, I’m unsure why the authors consider
these losses deleterious if the organism will no longer receive rewards
for executing old instructions: there is no motivation to keep them around.
Two important open research questions on the topic:
As indicated by the authors, an sexual versus asexual reproduction could
be analyzed, however the genetic aspect of choosing feature inheritance
may add to the “stacking of the deck” aspect of the study.
An interesting modification to this study would be one where the
environment (or the definition of success) where to change over time.
Another area for future research would be a less random mutation scheme
where external factors (environment, previous mutation choices,
knowledge about mutations that yield a reward of more energy) could be
used to decide genetic function.
This archive was generated by hypermail 2.1.6 : Wed Dec 01 2004 - 12:02:23 PST