virus: Replication and Fitness in memes
Sat, 1 Mar 1997 15:23:33 -0800 (PST)

Several threads here have dealt with the question of replication
of memes in other minds, and what consitutes their fitness. I
think it might help to clarify the issue to more precisely define
the "meme" in terms of the "gene" from which the analogy was born.


What is a gene? The word predates DNA, and was used to mean
"whatever it is that makes tall pea plants tend to produce tall
children; whatever it is that is passed from parent to child
that makes them similar." Notice that this definition is in
terms of /traits/, i.e., the phenotypic effects of genes, not
their particular mechanism. We happen to understand better now
the process of DNA replication that underlies the phenomenon,
but it would be a mistake to change the definition to simply be
a particular piece of DNA. We share much of our DNA with chimps,
but the same sequence of nucleotides in the same locus on the
same chromosomes of a chimp is not the /same gene/ found in a
human. The nucleotide sequence that causes human fathers to
produce human sons with certain human traits will actually be in
darwinian competition with the nucleotide sequence that causes
chimp fathers to pass certain chimp traits to their chimp sons,
even if those nucleotide sequences happen to be identical.

Even if inheritance happens in DNA, selection happens among the
phenotypes--i.e., the effects that a gene has on the world
(mostly, but not exclusively, the effects it has upon the beast
which carries it). Thus, the "gene replication" in the darwinian
process is not merely the copying of DNA--it is the copying of
the /effects/ of that DNA into children, because it is those
effects that compete in the world and cause selection. It is
important to realize, then, that in the mathematical model of the
darwinian process, a gene is /implemented/ in some physical
medium, but it is not the medium itself; it is the inheritable
trait--the phenotype--that participates in evolution.

For darwinian evolution to take place, there must be several
factors: (1) Something that replicates itself, potentially into
the indefinite future, (2) Relative advantages/disadvantages of
certain replicators--or their effects--compared to alternative
ones in their ability to replicate, (3) Occasional random changes
that create new replicators.

Do "memes" meet these criteria? I think so, because we can
define "meme" to be "transmissible phenotypic effect of a mental
process" just as we define "gene" as "transmissible phenotypic
effect of a biological process" above. In other words, whatever
real physical manifestation of mental pattern might reside in the
brain isn't relevant: whether it is dendritic connections, or
Calvin's spatio-temporal firing pattern, or something else, what
matters to the process of darwinian evolution is does it have a
phenotypic effect that can be duplicated? Clearly, that answer
is yes: a behavior such as throwing a rock may have a completely
different internal mental process in my mind than it does in the
mind of a child I teach to throw, but the end result of throwing
the rock is essentially the same. I have transmitted the /effect/
of the mental process that causes me the throw a rock in a certain
way, to another person, who probably has some different internal
process, but nonetheless winds up throwing the rock.

I don't really even care what the details of my mental process are
(well, personally I do for my own reasons, but evolution doesn't).
All I care is that I can in fact teach my children to throw rocks,
and thereby increase their fitness to survive in the world, have
children of their own, and teach their children to throw rocks.
The behavior will certainly mutate--likely much faster than a gene
would--and the traits and behaviors thus transmitted will certainly
compete, so I think the criteria for darwinian evolution are met.
If is pointless to argue about whether the internal process itself
can or cannot be transmitted. That's irrelevant: the /effect/ of
that process is what will compete, and that /can/ be transmitted.

While such behaviors as throwing may have caused the brain to
evolve this memetic faculty, that's not the only thing a brain
can do. It can also imagine, abstract, predict. These processes
too, since they result in physical behaviors, can be transmitted.
An idea like "don't throw the soft crumbly rocks, throw the hard
heavy ones" requires abstractions and classifications like "hard"
and "better". Since these abstractions modify the behavior of the
one being taught as they to the teacher, the /effects/ of those
abstractions are being replicated.


So which memes are likely to find themselves in future generations?
Certainly, ones that improve the physical survival of the bodies
that carry them will stick around. This includes behaviors that
are suited to the lifestyle and environment of the host, and also
abstractions that help accurate prediction of events ("truth", in
a more pragmatic sense than my earlier definition). But one must
not ignore the possibility of some that will subvert the process of
replication to advance themselves at the expense of their hosts,
just as "outlaw" genes like segmentation distorters do in bioloical
evolution. Similarly, since the evironment in which these carriers
find themselves is shaped largely by other meme-carriers, just as
biological evolution is shaped more by other organisms than by the
inorganic environment, it is likely that opportunistic memesets will
arise to take advantage of other memesets, until evolutionarily
stable strategies less vulnerable to exploitation emerge. Just as
biological ESSs are unlikely to be "optimal" outcomes from the a
priori reasoning point of view, it is likely that memetic ESSs will
not be optimal either.

For example, the "optimal" trading strategy--that is, the one which
would result in the highest overall payoff if everyone adopted it--
is for everyone to be scrupulously fair and trusting. No costs are
expended on defense, and everyone gains the full benefit of trade.
But if such a strategy were the norm in a population, a few criminals
could clean up using predatory strategies. Therefore, a less
optimal and less vulnerable strategy will be more stable: say, one
that involved general honesty and trust, but with retaliation against
defectors, or one that spent some costs up front on identification
to ensure accountability.

In short, the overwhelming, incontrovertable evidence of darwinian
evolution in biological processes, and the clear parallels to its
fundamental prerequisites in mental processes, can't be rationally
dismissed. Just as we may judge our behaviors such as promiscuity
in men by objective standards and still understand what evolutionary
processes may have encouraged them, so too must we judge ideas on
their objective merit, but also pay close attention to their memetic
fitness. Sometimes those will match, but sometimes they will not,
and we must deal with that reality.

Lee Daniel Crocker <>