#memes #memetics #Pigliucci #Dawkins #darwinism
[updated 07 jan 2010 9:30 am CET]
The main part of the post is in [bad] english after a somehow long introduction in french, commenting Massimo Pigliucci’s essay on memetics, published as part of a trilogy about Dawkins, published in 2007 in Skeptical Inquirer1, 2, 3. Go below the fold for the meat. I would be happy to hear from the man himself.
Jean-Michel Abrassart, dans son dernier post au sujet des mèmes et de la mémétique, fait référence à un texte de Massimo Pigliucci1 qui a trait à la question. C’est le deuxième d’une trilogie, où Pigliucci semble vouloir régler le compte de Richard Dawkins.
Dans la première partie2, il questionne l’attitude de Dawkins quand il affirme que l’hypothèse-dieu est une question scientifique. Pigliucci préfère que ce soit une question philosophique. Chacun prêche pour sa paroisse ? Le fait est que croire à une instance surnaturelle malgré l’absence de données empiriques qui pourraient soutenir son existence est irrationnel. Réfuter l’irrationnel n’est pas une exclusivité philosophique, c’est même une spécificité scientifique. Dans un cadre où les scienligieux essaient d’accaparer la science pour faire croire qu’elle supporte leurs croyances (ou tout au moins à la Staune : « autorise à penser, sans le prouver » que [insérer l’insanité de votre choix, e.g.]) l’hypothèse-dieu gagne à être examinée aussi en tant que question scientifique, surtout telle qu’elle est formulée par la trinité de la Templeton Foundation, Francis Collins (le créationniste comme Staune est créationniste), Ken Miller le créationniste façon catholique, et Conway Morris, le dernier à avoir fait son coming-out. Templeton Foundation qui pour préserver une image clean a su changer sa mission de rapprochement de la Science et de la Religion aux Grandes Questions.
J’ai envie de rester aux questions de Dawkins, en considérant l’hypothèse-dieu comme une question scientifique, si et tant qu’on souhaite l’aborder en termes de recherche d’évidence empirique qui pourrait la soutenir, la faire rentrer dans le champ du rationnel; non pas que je pense qu’on en trouvera un jour 😉
Ca n’empêche pas le philosophe qu’est Pigliucci de poser la même question en tant que question philosophique et conclure la même chose en absence d’évidence empirique.
Dans la troisième partie3, c’est le concept du gène égoïste que Pigliucci essaie de mettre en cause, avec l’argument que la sélection naturelle agit à d’autres niveaux que les gènes eux-mêmes. Il a parfaitement raison.
Ce qu’il n’évoque pas dans son argumentaire c’est qu’aucun de ces autres niveau n’existerait en l’absence de l’information génétique et que la sélection naturelle ne serait pas opérante en l’absence de la particularité de réplicateur du matériel génétique qui porte cette information et qui permet sa transmission, plus ou moins altérée (par mutation et recombinaison). On peut tourner les choses dans tous les sens, point d’évolution biologique en l’absence du réplicateur, qui réplique des gènes, aucun des autres niveaux auxquels la sélection naturelle agit, niveaux qui sont déterminés quand même par les gènes et leur environnement, qu’il s’agisse des autres gènes avec lesquels ils sont associés, ou non.
Pas-mal-adaptationniste et avec une vision géno-centrée du monde biologique, tant que l’hérédité est génétique (ou épigénétique), je me place facilement du côté de Dawkins encore une fois, aucun des arguments de Pigliucci n’invalidant le concept du gène égoïste.
On pourra discuter le premier et troisième essai de Pigliucci par ailleurs, mes commentaires ci-dessus ne servant qu’à offrir le contexte de « The trouble with memetics », qui vient encadré des deux autres.
L’attaque est décliné en trois parties :
Le concept du selfish gene tient pas mal la route, quel que soit le niveau auquel la sélection naturelle agisse. Qu’en est-il du concept du meme, qui plus tard a donné naissance au concept memetics. Quels sont les arguments de Pigliucci. (et là je passe à l’anglais, pour faciliter les éventuelles échanges avec Pigliucci)
The introductory paragraph present the main Dawkins’ point, universal darwinism, nicely exposed:
Darwinian principles are universal, and they apply to whatever other system features the basic characteristics of heritable variation in attributes affecting fitness.
The concluding paragraph present Pigliucci’s positions concerning
[…]Dawkins’s claim about universal Darwinism is probably correct […]
Memes, on the other hand, have a long way to go before becoming a sufficiently fecund concept for scientists to work with.
The former being supported by the work done by computer scientists, the latter by arguments presented in between the first and the last paragraphs.
I will not question one or the other of Pigliucci’s conclusions, I agree with both.
What I would like to discuss is how the argument is set up to conclude about the memes, if it presents weak points and, most importantly, what can we learn while trying to reply to the remarks made; which begs the question why Pigliucci, who is a bright guy, and a bright (this is a second agreement with Dawkins after the universal darwinism stuff), didn’t answered a few questions he asks by himself.
Let’s go back to universal darwinian principles applying to whatever system features the basic characteristics of heritable variation in attributes affecting fitness and examine if culture (and thus cultural information) can be considered as such a system. I will stick to the definition considered by-nowstandard (2007) as presented in the first sentence of the text:
« an element of a culture or system of behavior that may be considered to be passed from one individual to another by nongenetic means, especially imitation. »
And for the first part of the discussion I’ll consider only those elements of a « culture » or « system of behavior » that affect the fitness of their owners, elements for which it is obvious that the will be prone to selection (positive or negative) and give them the short name: memes.
We have the elements composing a system featuring the basic characteristics of transmissible variations in attributes affecting fitness. Is it such a system subject to the universality of darwinian principles? The main difference from the gene analogy is the way information flows: restricted to follow the genetic material vertically or horizontally in the case of genes, without physical strains and only horizontally in the case of memes (this is challenged below). From a gene centric point of view (even limited to those selfish genetic elements Pigliucci concede to Dawkins in his third essay4) there is at least one nice analogy between genes and memes as information subject to variation determining attributes affecting fitness. This may concern a limited set of genetic elements and limited transmission mode (HGT), but the analogy stands.
I’ll assume that we have the elements to consider the system subject to darwinian principles and hopefully I will consolidate this stance below. Under this assumption the study of memes, memetics & evolutionary memetology5, should be considered as much interesting as genetics & evolutionary biology. (2) Independently of any analogy between genes and memes. Independently of any analogy between genetics and memetics. Independently of who first proposed the term meme.
But Pigliucci finds that « there is quite a bit that is rather unconvincing about the whole idea.«
The first objection comes about the clarity of the definition of memes. Certainly the dictionary’s definition is unsatisfactory for one who would like a functional definition to work with. There is no distinction between what encodes the meme, the effectors allowing a meme to be expressed, the phenotype corresponding to a meme. As Pigliucci states, this isn’t a fatal problem, but let’s work it out before going further, because it may cause things to look worse than they are.
Thus the sensors used to observe the phenotype may be various, the memories types and encodings may be various, the effectors used to exploit them to reproduce the phenotype may be various, as long as the phenotype is reproduced. The phenotype is the replicator, the attribute affecting fitness, eventually subject to variation.
Sensors and effectors are the « interactors », the phenotype is the replicator. Easy enough to make the distinction. And, that keeps a nice analogy between genes and memes, which as I support isn’t important, but adds some familiarity to the subject, at least for biologists.
So, not only we don’t have a fatal problem, we don’t have a problem at all, and nothing that could point to a disanalogy between genes and memes, yet.
For the second objection I’ll cite directly :
The second problem with memes is that nobody seems to know what their physical basis is. Genes are-roughly speaking-pieces of nucleic acids (DNA or RNA), with known physical-chemical characteristics. But memes can be instantiated equally well inside someone’s mind (where presumably they correspond to specific patterns of neuronal firings), on a computer’s hard drive, in a book, or on an iPod. While it is true that, for decades after Gregor Mendel proposed the idea of genes, biologists didn’t know what they were made of either, the likelihood of pinpointing a physical makeup for memes is less likely because they seem to be a sort of « diffuse » entity that can have many physical incarnations. Again, this is disanalogous with genes.
If the argument was presented at 1976 there would be little to say to counter it. This was published in 2007. Genes, genetic information encoding for the production of biological effectors, with or without their expression regulating sequences, with or without the primary sequence of the mRNAs, miRNAs and proteins, with or without data corresponding to secondary and tertiary structures of these molecules (at least at some crystalized state) there are plenty my computer hard drive, books and my iPod. Some of them are recombinants made in silico, tested for conformity of expression of the protein in silico, one of them synthesized from scratch based on information gathered from NCBI. And I can get more of this stuff. And this kind of stuff, genetic information, was widely available before the publication of The trouble with memetics in the September/o-October 2007 issue of the Skeptical Inquirer. At the same moment that « The Diploid Genome Sequence of an Individual Human [J. Craig Venter] » was published.
What part of it could escape the vigilance of a bright scientist and philosopher? None.
A meme’s instantiation in any of these supports may be as much active or inert as a gene’s instantiations in the same supports, as far as the system don’t dispose of the necessary effectors to express them. Not only I don’t distinguish the faintest of disanalogies here, but I clearly see homologies between the two situations, with a major difference: memes may be effective independently of biological objects if the carrying systems dispose of the appropriate effectors to reproduce the corresponding phenotype.
With a hopefully cleared expression of what is the meme’s code (memory), that a meme is expressed as a phenotype via effectors (of biological or not origin) and with the restriction I made concerning memes to which we may ascribe a function (affecting the carrier’s fitness), let’s examine the next point that bothers Pigliucci, « meme complex, » or « memeplex. »
Bacterial metabolic pathways often come in operons, sets of contiguous genes encoding for proteins implicated in a series of catalytic activity allowing the production of a product of some use for the carrier. The metabolic pathway may be seen as a single function from those people who may be satisfied by a black-box.
Similarly, sets of contiguous memes, encoding for phenotypes implicated in a series of social activities of some use for the carrier, come in sets. Let’s call that a memetic pathway which can be seen as a single function from those people who may be satisfied by a black-box.
We can follow the same path, with increasing numbers of genes or memes, through what would be episomes carrying several operons, chromosomes, individuals, groups, populations, etc. And there would be no need to spend time to establish analogies between specific biological and memetic features.
A memeplex is composed by memes necessary for a particular function. The interest of the researched have nothing to do with it, except in the particular case where the researcher is interested by this particular function.
In the case of chimps using short strong wood sticks to open holes in the walls of a termite mound, then long, thinner, flexible, chewed at one end, sticks to fish the termites through the hole they just made, we may talk about a simple memeplex, just two memes (or three if you consider stick-fishing and chewed stick fishing as two different memes, both being observed), serving a quite important function, foraging.
For more complex functions, such as being a roman catholic, more memes are necessary, they are transferred in a more systematic manner, catechism, and there are even summary forms for quick display of the phenotype in a more explicit manner than just declaring being roman catholic, the recitation of the credo. Or the gesture with which the catholic faithful cross themselves.
In these casee also, the interest of the researcher is secondary.
So, to define a memeplex a good approach is to focus to its function rather than to some researcher’s interest.
The next objection comes with a little détour to the creationist’s universe to come to Popper’s objection presented as valid for memetics: « the only way to tell which memes are going to be successful, which tunes are going to stick in your mind, or which religions are going to become popular is by waiting and seeing what happens« . Well, he is right, quite difficult to say which tune will be a success or which sect will become popular and raised at the religion status.
As much difficult as to say which allele will prevail without knowing what function they serve. Let’s use something more simple and see if one can make predictions taking out Popper’s objection.
There are at least two populations of chimps termite-fishing.
One can hypothesize that if the first group observe the practice of the second group it is possible that they will get the memeplex consisting of using two different tools, make artificial holes and get more termites from each stick insertion. Knowing that chimps can adopt new memes serving foraging, one may make the prediction that the memeplex will win over the less efficient single meme and that there will be new converts. I hope I’m reasonable if I make such a prediction, as the memeplex seem to improve the fitness of the carrier.
If so, Popper’s objection would invalidated for memetics. Someone should do the test and see if my prediction is verified, to free memeticists from this burden. That will not make memetics scientifically full all of a sudden, but this would be a start. (
well, I’m such an ignorant in the memetics field and primates ethology (and so many much more domains…) that it could be that the results are already available and I don’t know about them; if so, drop a comment Found exactly what I wanted: Transmission of Multiple Traditions within and between Chimpanzee Groups, just the right keywords to use, say Dawkins in the references 😉).
Pigliucci, « despite the questionability of memetics, [consider] Dawkins’s claim about universal Darwinism [as] probably correct » and use a non biological example to study, coming from the computer scientists, genetic algorithms, mimicking biological processes. He should have gone one step further and mention that the very same computer scientists developed memetic algorithms, mimicking memetic processes and find, usually, that an hybrid of genetic and memetic algorithms outperform genetic algorithms.
Genetics italicized, memetics not, intentionally.
It is true that there is no way to transfer and express genetic information otherwise than by using a particular physical substrate, roughly one or another of the nucleic acids, which make it central for biology. On the other hand, there is quite easy to use a wide range of substrates, including all this computer stuff mentioned, to transfer and express memes. Memory supports can be just common between several systems (with appropriate read/write accesses), be physically exchanged, duplicated, support intermediated encoding as algorithms and the meme can be expressed if the system dispose of the appropriate effectors: a tune can be played by my iPod, but my iPod doesn’t play drums (yet). The robot is a non-biological system able to reproduce a phenotype using memories of it and the appropriate effectors. The robot replicates the meme in a way it can be observed and imitated by another system, biological or not, able to learn by imitation, that is, instantiate the meme as its proper memories. For robot Learning from Demonstration (LfD) check Google and your local library.
Memes and memetics concern as much living organisms as artifacts, thus I’ll keep memetic algorithms in contrast to genetic algorithms, despite the fact that memetic algorithms are the result of biomimicry as genetic ones.
There is an interesting point concerning artifacts and memes transmission: artifacts can transmit memes vertically by sharing memories, not just through imitation. One could make the prediction that they would be a more efficient medium for transmission of acquired skills.
What vocabulary specific to learning from imitation, starting at the simplest living organisms able to do so, and encompassing the new technologies in development, known that the replicators corresponding to learning from imitation were termed for simplicity memes, 33 years ago by Dawkins, one could use? Meme and memetics seem appropriate and already in use. That’s a meme gaining thrust, you like it or not.
Massimo Pigliucci attempt to critic memetics and the idea that they may build up as a specific scientific domain studying learning from imitation, and that this would be a new way to address how cultural elements evolve, is not really convincing.
Focused to disanalogies between genes and memes, which are not, he fails to address the main point: are darwinian principles applying to memes (and you may spell it « culture » if meme discourages you)? If so, the analogies, homologies and, most importantly, specificities, will be found progressively.
The main objection to see memetics as a scientific enterprise is the one phenocopied from Popper’s objection about biological evolution and it probably could be refuted by not-waiting-to-see-what-it-happens but trying to make a prediction and see if it is falsified or not. That is being constructive and I’m sure Pigliucci is more ingenious and constructive, when he is in the right mood, than I do. Shall we expect his proposal of a smart prediction to test or event better, how to build appropriate predictions to test memetics?
And I still agree with him, memes have a long way to go before becoming a sufficiently fecund concept for scientists to work with (with the exception of computer guys who are already successfully with it and certainly some ethologists).
1. Pigliucci, M. 2007. The trouble with memetics. Skeptical Inquirer 5:23-24
2 Pigliucci, M. 2007. Is Dawkins deluded? Skeptical Inquirer 4:20-21
3 Pigliucci, M. 2007. Beyond selfish genes. Skeptical Inquirer 6:20-21
4 cited from 3 « […]where natural selection can act directly on genes (there are trully selfish genetic elements that reproduce parasitically inside the genome itself) […] »
5 memetology, the term is terrible, I’m keeping it as a proxy for something more appropriate