Jesse

Interesting article here:

http://www.stats.org/record.jsp?type=news&ID=435

Pinker's actual talk on this subject, which is a bit longer but worth reading, can be found here:

http://www.bioethics.gov/transcripts.../session3.html

AtomSmasher

He seems to presume we will avoid attempting eugenic modifications because of our currently poor understanding of complex gene interactions. I believe our understanding of genetics is still primative and once we have a good, thorough map of genes/protein functions and an excellent model of interactions we will be able to predict outcomes and select or create preferred genotypes. Right now we are severely limited by our poorly detailed map of the genome. We have mapped a great many gene sequences yet only know what a fraction of them do. Also the computational resources needed to model the complex interactions of genes/proteins and their effect on large scale systems is not available. Nobody said this ability would happen soon. Assuming our understanding of genetics increases as well as our ability to model and compute the outcome of alterations, I see no reason we could not successfully engineer desired traits within 20 or 30 years. I might be optimistic.

The stats.org article insinuates that it would be futile or even dangerous to attempt to select for traits such as intelligence -- pointing to the increased likelyhood of autism/asperger's, schizophrenia and other mental disorders in the children of parents with high intelligence. Indeed this is true, but if we better understood the genetics behind these conditions we could probably avoid them. The article seems to indicate it would be futile to try, perhaps it is, but I think this field of knowledge is in its infancy. Our knowledge of the possible applications for radiological elements was poor when the first atomic bomb was detonated, if we had presumed to understand it's true potential at that moment we would never have created nuclear power. No doubt we might have banned it's use outright, much like cloning research.

Jan Haugland

Very interesting article, Jesse, and it supports what I've learned about this subject so far.

It seems the only reliable way to ensure the genetic quality of our offspring is the old-fashioned way: finding a great mate!


- Jan

...who rants and raves every day at Secular Blasphemy

Jesse

AtomSmasher:
He seems to presume we will avoid attempting eugenic modifications because of our currently poor understanding of complex gene interactions. I believe our understanding of genetics is still primative and once we have a good, thorough map of genes/protein functions and an excellent model of interactions we will be able to predict outcomes and select or create preferred genotypes.

I think the behavior of complex self-organizing systems filled with all kinds of crazy feedback loops is something which human minds will never be very good at "understanding" in the conventional sense. Even with a fairly simple neural net, for example, it would probably be very difficult to have work out any clear understanding of how it is able to solve a particular problem or recognize a particular pattern. And the sum of all the interactions between genes in all the different cells of an embryo growing over time would be a problem many orders of magnitude more complicated. Of course, even if we cannot reason out the changes which we'd need to make to an organism's genotype to get a particular phenotype, it's possible that we'll develop powerful enough computers to be able to create pretty accurate detailed simulations of the development process, in which case we could use something like a genetic algorithm to locate a genotype that will lead to a simulated phenotype with certain desired traits, hopefully much faster than actual selective breeding in the real world (which is basically what such a genetic algorithm would be simulating). However, if the traits you're interested in are complex mental traits, it might be that the simulation would have to be so detailed that the simulated organism would be a full-fledged A.I. with a mind of its own, in which case you'd be stuck with all the same moral issues that come up when you consider eugenics.

jayh

You are assuming that additional information will help significantly. This is not necessarily so, in fact if the genetic interactions have signifiant components of chaos (in the mathematical sense) as is quite likely, and many results are affected by many interacting genes which cannot individually be controlled, there may be nothing we can do in a predictive manner. (we understand all the physics of a ball bouncing in a roullette wheel, but prediction is still impossible)

ps418

What you say is correct, but only for traits with low narrow heritability. However, some traits of interest to hypothetical genetic engineers (e.g. cognitive ability, some personality dimensions, lots of health related traits such as blood cholesterol) are known to have signficant narrow heritability, and relatively little evidence for significant gene x gene or gene x environment interactions. There is reason to think that for these traits there are in fact many individual genes with consistent phenotypic effects that could be selected for. In cases of traits with high broad heritability and low narrow heritability, however, it is correct that prediction would be very much harder, though still not a priori impossible (provided it is possible to understand the relevant gene x gene interactions).

Patrick

LostGirl

I still agree that traits like particular intelligence (ie; mathematical or philosophical), musical ability, or personality traits are complex and emergent. Like the roulette wheel, we may be able to identify the genes that are critical for the development of the trait via knockout effects, but understanding the delicate interaction between multiple alleles at multiple loci would take extremely sophisticated computer modelling. And, as Jesse said, . I also question the ability of any computer algorithm to accurately mimic all of the environmental variation involved in the womb, birth, and early childhood, all of which contribute critically to neural development.

The traits where there is little allelic diversity in the population and expression is controlled by the interaction of only a few genes with high narrow-sense heritability, prenatal/preconceptual selection for certain traits *might* be possible. These will probably have mostly to do with physiologic (esp. metabolic) and developmental issues, and not complex, emergent traits like intelligence and personality.


Finally, one last thing to note: the hierarchy of relatedness/heredity should include the noticeable differences between twins raised together and twins raised apart. If cloned embryos for IVF ever take hold, and more than one woman gestates genetically identical embryos, then we can also evaluate environmental influence in the womb. This is crucial to consider when looking at twin studies. They are often raised with very, very similar environmental influence, especially in the womb.

Jesse

jayh:
You are assuming that additional information will help significantly. This is not necessarily so, in fact if the genetic interactions have signifiant components of chaos (in the mathematical sense) as is quite likely, and many results are affected by many interacting genes which cannot individually be controlled, there may be nothing we can do in a predictive manner. (we understand all the physics of a ball bouncing in a roullette wheel, but prediction is still impossible)

It seems to me that gene interactions during the development process cannot be too chaotic, otherwise identical twins wouldn't look so identical! Or am I misunderstanding your argument here?

jayh

That is something different (you are thinking of random, not chaotic). It's not complex to make a copy, but it is exceedingly complex, probably unpredictably complex in many cases to assemble a new combination of genes and predict the the outcome. That's the chaotic component.

Jesse

jayh:
That is something different (you are thinking of random, not chaotic).

Actually I was thinking of the fact that chaotic systems exhibit sensitive dependence on initial conditions--the "butterfly effect." But maybe you are not arguing that the developmental process, in which cells divide and complex feedback loops switch genes in different spatial regions on and off leading to cell differentiation and spatial pattern formation, is itself chaotic. Presumably this developmental process is sufficiently chaos-free to insure that if you start with the same genes you will get nearly identical phenotypes, as shown by identical twins. But wouldn't this also suggest that similar but not identical genotypes would lead to very similar phenotypes, so you don't have sensitive dependence on initial conditions in this sense either? Perhaps in most cases small genetic changes lead to small changes in phenotype but certain critical changes can lead to very large changes in phenotype, more like catastrophe theory than chaos theory.

jayh:
It's not complex to make a copy, but it is exceedingly complex, probably unpredictably complex in many cases to assemble a new combination of genes and predict the the outcome. That's the chaotic component.

But what exactly is the system you're identifying as chaotic here, if not the developmental process which determines the genotype/phenotype relationship? What would an example of sensitive dependence on initial conditions be in terms of what you're arguing?