Wednesday, June 6, 2012

Substrate Neutral Inheritance Systems

My preoccupation with cultural robustness and evolvability has naturally caused me to think frequently about the problem of cultural inheritance.  If we take a broad definition of culture to be non-genetic, non-epigenetic information that is socially transmitted, then culture becomes a Darwin machine like any other. For those unfamiliar with the term, a Darwin machine is any iterative process in which heritable variation is generated and subsequently undergoes some sort of selection.  The components necessary for creating a Darwin machine, then, are variation, selection and inheritance.

Note that this formulation is substrate neutral.  In other words, it does not depend on what the media of variation, selection and inheritance are.  Although the specific details will change, the same general logic applies whether we are talking about genetic systems, certain kinds of evolutionarily inspired computer algorithms or, presumably, culture.  This makes the notion of a Darwin machine potentially quite powerful in its ability to describe and understand many different kinds of systems from within a common framework.  Unfortunately, though, there remains much work to do in order to realize this potential.  In particular, to become more broadly applicable the individual components that make up Darwin machines must also be framed in substrate neutral terms.

To take but one of the necessary ingredients into consideration, how could inheritance systems be framed in substrate neutral terms?  To begin to imagine this I think it makes sense to start by considering the necessary components that make up the system we understand best, the genetic inheritance system.  So thoroughly studied has this system been that a central dogma has arisen around it that describes the flow of information within cells.  This flow begins with information storage in the form of DNA.  Almost every cell within a body contains a complete copy of all the genes necessary to build and maintain the organism.     However, not every gene is expressed at every moment.  Indeed, timing is critical for many cellular processes and so cells have evolved mechanisms to determine which genes should be turned on and off at specific times in response to environmental signals.  Once a gene is turned on, it is transcribed into an RNA intermediate which is, most often, translated into a protein product that performs some important function for the cell, be it structural, regulatory or metabolic.

At some level, carving this system up into essential components is a bit arbitrary.  Nevertheless, I think some general themes do emerge.  For instance, it seems an important component is some kind of information storage system.  Obviously, for inheritance to occur there must be some way of transmitting this information to new entities.  At the same time, there needs to be an information retrieval system and a crucial part of this information retrieval system would seem to be way of "reading" the environment to "know" what information is required and when it is needed.

Are these the minimal necessary components?  Would they be sufficient to establish an inheritance system?  Are all of them needed?  Are there other components missing from this list?  I would very much like to see more discussion along these lines and I welcome any insights others might have.

I recently had someone tell me that there was no need for new, substrate neutral terms.  He felt that genetic inheritance provided a sufficiently broad framework and that it could easily be expanded to accommodate other kinds of systems.  I think this is a mistake for at least two reasons.  First, I worry that using genetic terminology predisposes researchers to only look for the familiar.  While there are undoubtedly similarities in cultural and genetic inheritance systems, there are just as assuredly significant differences as well.  Foregrounding future work with genetically laden conceptual frameworks makes it likely that at least some of these differences could be completely overlooked.

Second, I believe that taking a step back to look at inheritance from a substrate neutral perspective produces new questions and opens space in which to explore them.  For instance, if it is true that an inheritance system needs some form of information storage then an obvious difference between genetic and cultural evolution is that the former contains information in a relatively concentrated from while the latter uses a distributed system.  This observation leads naturally to the question of how concentrated vs. distributed systems might respond differently to natural selection.  It similarly leads to questions about how information retrieval is achieved in a distributed system in a way that allows entities to respond adaptively to their environment.  I don't feel these same kinds of questions emerge as naturally when using analogical reasoning to expand a genetic framework.

10 comments:

  1. I call this "informational genetics".

    The last major stab at alternative substrate-neutral terminology for inheritance was Dawkins with his replicators. Unfortunately, this was widely misunderstood - and has caused much confusion.

    Before that, were Semon's mnemes, from his 1921 book "Die Mneme". Oh, and Evan Louis Sheehan (2006) proposed using the term "meme" for this basic concept, in "The Mocking Memes: A Basis for Automated Intelligence" (not standard usage of the term).

    ReplyDelete
    Replies
    1. I'm not sure I'd call that the last major stab, though most efforts I've read since Dawkins rely in one way or another on his terminology. One recent attempt at reframing these issues is Hodgson and Knudsen's "Darwin's Conjecture." They take some interesting strides and many of them strike me as steps forward. Nevertheless, their theoretical work still seems rather labored to me, as though they're having to work too hard to make all the pieces fit together.

      A more promising trajectory, in my opinion, is launched in Claes Andersson's "Splitting the replicator: Generalized Darwinism and the place of culture in nature," an article that appeared last year in the Journal of Economic Behavior and Organization.

      Delete
    2. As it sounds as though we have already discussed, I propose using standard genetics terminology - i.e. we already have a science of inheritance, it's called "genetics".

      If you want other substrate-neutral terms, heredity, memory, copying, persistence, selection and elimination all look pretty useful. However, really we should be borrowing most from evolutionary biology, epidemiology and genetics.

      I looked at the Claes Andersson abstract. It looks like a bit of a straw man attack to me. Nobody ever thought that copying is the only way to retain things over time in the first place. Attacking this proposition seems like a waste of energy.

      Delete
  2. I am sceptical whether the level of information "distribution" differs much between cultural and genetic systems. Both are massively decentralised and highly redundant in most respects, surely. In what sense are DNA genes "relatively concentrated" - when compared to memes?

    ReplyDelete
    Replies
    1. I'm not sure in what sense a genome would be considered massively decentralized. Could you clarify what you mean by this?

      As I'm thinking of these terms (which admittedly need refinement) I am considering primarily the distribution of information across those elements that store it. Genes are generally stored on chromosomes which are, themselves, discrete entities. Moreover, most of the cells within an organism's body contain the totality of its genetic material even though this information is packaged and accessed differently. Cultural information, by contrast, is in some sense stored in the brains of individuals. No single individual brain contains all of the cultural information necessary to reconstitute the society in which it is embedded. It is in this sense that I consider cultural knowledge to be stored diffusely relative to genes.

      Delete
  3. Brains? I suppose so - in the past - though meme pools are about as decentralised as gene pools.

    Today a DNA molecule could still be described as being a "more concentrated" source of information than, say, an USB drive - because it packs more information into a small space. However, this seems kind-of irrelevant to the resulting dynamics. Also, pretty soon the USB drives are likely to shrink down to the size of cells - and then then this difference will disappear.

    ReplyDelete
    Replies
    1. I suppose what I am getting at is the relationship between information and its carriers. And individual organism is formed de novo from the fusion of gametes (in the case of most eukaryotes). All of the cells that subsequently differentiate from that fusion contain the same genetic information and, in principle at least and to a limited degree in practice, it is possible to derive cells of one type from cells of another type. These collections of different cell types work together to form the complete organism which becomes a carrier for the particular subset of the gene pool it happened to inherit. It is in this sense that I regard genetic information to be relatively concentrated.

      This strikes me as different from the relationship between cultural information and its carriers. Whereas the individual is formed from its genetic information and can only potentially pass the information that formed it to future offspring, cultural information is not similarly moored. An individual carrying particular cultural information is unlikely to pass all of it to any particular offspring. Indeed, an individual passes different parts of his or her cultural repertoire to different individuals, most of whom are probably not genetically related. Moreover, the individual is not limited only to the information he or she receives at one moment in time, but rather becomes a flexible repository of constantly changing information.

      Inasmuch as information depends on its carriers, then, it strikes me that these two systems might react differently to similar selection pressures. At this point, I'd classify this idea as more suspicion than workable theory and I would not be at all surprised to learn that I'm way off base. Nevertheless, I don't believe I would have conceived of this distinction at all as an avenue for future inquiry had I grounded my thinking strictly within a genetic framework.

      Delete
  4. So: it *sounds* as though you need the concept of a "cultural creature". I.e. genes produce organic creatures, while memes produce cultural creatures. The resulting cultural symbionts aren't human beings, they are all manner of other things - including artifacts and behaviour patterns.

    Organic symbionts are passed on to non-offspring, just as cultural symbionts are. Your gut bacteria, your diseases, your pets, and so forth, can be passed on to all kinds of unrelated individuals. Organic contagions and cultural contagions both obey the rules of "generalised epidemiology". This is, in fact, a deep and important similartity between the organic and cultural realms - not a difference between them.

    ReplyDelete
  5. Yes and no. I really do like this terminology. It's quite evocative and I can see how it could be tremendously useful. At the same time, though, it still doesn't seem to quite capture the relationship between the information and its carriers. Yes, symbionts may be passed on from one host to another. However, the information needed to create, maintain and reproduce the symbiont is still largely self-contained within the genetic material of the symbiont itself and not with its host. Clearly, the picture is far more complicated that that, I realize, so I need to give this further thought. The differences I am perceiving may well be more apparent than real.

    ReplyDelete
  6. You are suggesting that an AIDS virus may be somehow less involved with its host that a meme is with its host? I'm having a hard time in thinking of senses in which this is true. Even if so, I figure you should be thinking of this as a non-essential feature of mutualistic and symbiotic relationships. Surely you can't have universal Darwinism without mutualism and symbiosis. It would be a joke version of the idea.

    ReplyDelete