Re: virus: Memes and Genes
Thu, 17 Oct 1996 00:57:20 -0500 (CDT)

On Wed, 16 Oct 1996, Reed Konsler wrote:

> >From:
> >Date: Wed, 16 Oct 96 09:26:19 BST
> >Strangely enough, I was discussing this with some friends the other day, during
> >a conversation debating Foetal Research. Apparently, it has become possible
> >for geneticists to test the unborn child for social and physical traits, which
> >could be "abnormal"! This includes things such as Alcoholism, and Criminality.
> >Therefore, does it not seem reasonable that as these things are controlled by
> >genes, then memes which can infect the mind are also more prevalent in
> >different
> >genetic codes? For example, one with a Criminal gene is likely to be more
> >susceptable to "criminal" memes, and less so to "law abiding" memes!
> I've yet to see such a study which I found compelling. The problem is
> controlling for developmental and environmental factors, which one cannot
> do given the statistically insignificant sets of people as yet studied.
> There is also a correlation vs. causation problem. Just because some
> genetic marker tends to correlate with a behavior doesn't mean one causes
> the other. The mechanism which links the two may be socio-economic in
> nature and may in fact be an artifact of the selection criteria. For
> instance: how do you measure criminality? Lets say you decide to do so
> based upon convictions in court. As a percentage, African-Americans are
> more likely to be convicted of crime than caucasians. Is this an
> indication of gentically differentiated criminality? Another plausible
> interpretation is that our court system has inherent racial bias.

Agreed. The above study description sounds like highly oversimplified
pseudoscience. I need more details, as well.

> So, we have two competing theories. Attempts to disprove "selection by
> inherent bias" have, instead, corroborated this theory in other social
> systems such as applying for loans, purchasing cars or houses, and
> receiving medical treatment. The reason studies have not been done in the
> courts is that our legal institutions will not tolerate such inspection and
> have the capability of blocking it. It is simplest, however, to assume
> that the courts function in a similar manner, or else we must justify
> experimentally why they should be different.
> >On what grounds do you base your assertion?
> Cultural evolution has been occuring in earnest over only the last 10 to
> 100 thousand years. This is an insignificant period, genetically. We are
> all of one genetic mold, so to speak. In the time of recorded history
> culture has been changing so rapidly that what makes a "fit" human brain is
> dramitically different with each generation.

I find "dramatically different" a strong claim. We're so dependent on
communication [especially verbal] that the pressures there are plausibly
fairly consistent. Certain phonemes are dramatically easier to
distinguish than others; these form the key parts of classifying verbal
language. Any genes that impair these would be selected against, once
cultural evolution started, fairly consistently.

Hmm. Let call a generation at 20 years [overestimate]. According to the
above, cultural evolution has been running for at least 5000 generations.

That is a VERY significant period, genetically!

Assume that our population is large enough to avoid random genetic
drift. Let's describe the allele we are interested in as A, and all the
others as a, and furthermore that the locus does not have marked
cross-interaction effects with other systems. [The last constitute
internal selective pressures, and are not likely near equilibrium in the
absence of outside pressures.]

Let's say that at any given proportion x for A, we have the following
result in the next generation. [Yes, there's an artificial quantization
effect here.]

A_(n+1) = (1+k)A_n/((1+k)A_n+a_n), a_(n+1) = a_n/((1+k)A_n+a_n). [You can
check that this gives normalized probablities for the next generation
n+1, given the generation n. I am using _ to denote subscripting.

So...what does this mean? Let's try some numerical calculations. I'll
try out a 'low-incidence' case. [say we just started intensive cultural
evolution, which suddenly introduces a selective pressure not experienced

Let k be 0.01 for the following instance. I'm going to assume 8-digit
precision [not really measurable, but these effects are small]

[This numerical technique IS evaluating a differential equation
numerically! This is why I call natural selection derivable from
Ordinary Differential equations]

Generation A_n a_n Normalization for n+1
0 0.05 0.95 1.0005
1 0.050474763 0.949525237 ~1.00050475
2 0.050953792 0.949046208 ~1.00050954
3 0.051437121 0.948562879 ~1.00051438 [slow A_n]
4 0.051924783 0.948075217 ~1.00051925 [slow A_n]
5 0.052416813 0.947583187 ~1.00052417

This suggests that in slightly over 20 generations, we're going to see
A_n go from 5% to 6%. And we've got *5000* generations? Furthermore,
the procedure seems to be speeding up in this regime as time goes on.

So, to get to 50%, it should not take much more than 22*45 = 990
generations out of 5000.

Let's try this:
Generation A_n a_n Normalization for n+1
0 0.5 0.5 1.005
1 0.50248756 0.49751244 ~1.00502488 [slow A_n]
2 0.50497500 0.49502500 1.00504975

Oh. 5 generations to clear 51%. It seems we get to halfway point
MUCH faster than 990 generations.

I think the possibility of such a dramatic shift in less than 1000
generations i.e. 2000 years is evidence that 10,000 years is a
significant time span for natural selection, from supposedly weak
selective forces.

/ Towards the conversion of data into information....
/ Kenneth Boyd