Journal of Memetics -Evolutionary Models of Information Transmission
http://cfpm.org/jom-emit/1999/vol3/cambridge_conference.html
Journal of Memetics -
Evolutionary Models of Information Transmission
http://cfpm.org/jom-emit/1999/vol3/cambridge_conference.html
The following report summarizes the issues which arose in the workshops associated with the King's College conference on memes held in June 1999. It is organized along the general lines of academic disciplines, since that is the way in which the workshops were arranged. Details of a forthcoming book deriving from the conference (including a chapter by Boyd and Richerson) can be found at http://www.cus.cam.ac.uk/~rva20/Darwin.html.
Given its origin in the work of the zoologist and evolutionary theorist Richard Dawkins, the memetics literature has continued to exhibit a strong flavor of evolutionary biology. Many of the problems with pursuing this line of research therefore tended to arise from consideration of memes as analogous to genes -- that is, as cultural replicators.
Where are memes? It was generally agreed that the concept of memes should not be restricted to being only thoughts or behaviors or artifacts, but should be generally conceived (i.e., substrate neutral) for now. The possibility that it is behaviors rather than ideas which replicate was not specifically considered, although that is a position adopted by others within the memetics community (e.g., Derek Gatherer, William Benson). David Hull forcefully argued, however, that definitional uncertainties should not stand in the way of progress; it certainly did not in the case of genetics during the first part of this century.
The relationship of memes and genes: It was argued that memes can certainly influence genetic evolution. One need only think of lactose intolerance (an example discussed by Kevin Laland), where cultural practices (such as drinking milk) influence gene frequencies in a group. Likewise, genes may determine meme frequencies, if indirectly, through psychological biases against the adoption of some kinds of beliefs or values. The general consensus was that both replicators should be seen as coevolving in potentially complex ways. It has been a failure of some work in memetics to believe that memes operate without constraint.
Mention was also made of the immune system as another replicator system, acting at a different temporal and spatial scale (within the lifetime of a single organism, and confined to that body). The implication is that we need not rely only on the gene analogy to understand cultural replication; a variety of replicator systems exist, the investigation of each of which can provide insights and heuristic principles for memetics, as well as a general awareness that there is likely to be more than one way to accomplish any evolutionary goal.
A major problem -- which many adherents admit to troubling over for a long time -- is establishing how the genotype/phenotype distinction might work for memes. This functional difference in the genetic system has been generalized by Dawkins and Hull as the replicator/interactor distinction. Although it is possible for a replicator to serve both as replicator and interactor (as in the case of a ribosome, for instance), it it is generally considered unlikely to persist, if only because replicators and interactors have fundamentally different roles to play in the evolutionary drama (as store of information and as survivor/transmitter, respectively). It is usually inefficient for the same entity to play both roles, so a competitor system with greater specialization would almost certainly win out in an evolutionary race, if only because a more specialized replicator would likely be more robust in its ability to duplicate itself. So it seems memeticists must develop a notion of a memetic phenotype, or "phemotype." While there are a number of contenders for this role, none has achieved a widespread tip of the hat.
Part of the problem with developing a rigorous notion of a memetic interactor is coming up with a criterion which surely identifies it as distinct from its progenitor, the memetic replicator. A related example is the prion: is this protein a replicator or interactor (or both)? David Hull put forward one criterion for making this distinction, which is generalizable regardless of substrate and replicator system (and thus a candidate for Universal Darwinism): the relative difficulty of reconstituting the replicator from an interactor. This is a generalization of the Weismannian notion that you can't go "backwards" from protein to gene. This inability arises because there tends to be some slop in the production of phenotypes: genes don't code for one phenotype, they code for a reaction norm, thanks to the impact of environmental conditions on development. So the relationship between replicators and their material instantiations is not simple (i.e., not one-to-one). This implies that information will be lost in the translation from meme to phemotype. It is this loss of information which makes the project of "reverse engineering" (or inferring the instructions from the product, as Susan Blackmore puts it) so difficult.
If one allows that memes can be present in artifacts, then memes such as ink on paper can be replicated with very high fidelity: using photocopiers, we have direct replicator-replicator reproduction, and consequently no loss of information. However, many memetic replication cycles seem to involve stages of translation from one form to another, and hence some information leakage. In particular, more "traditional" memetic life cycles involve brains. If a meme must pass through someone's head, the general inability of bits of brain to duplicate themselves directly means that they must travel between hosts to replicate. But to do this, they must be translated into another form for social transmission -- for example, as bits of speech -- since bits of brain don't themselves make the journey from one head to another.
The first aspect of the problem is that, per the above argument, such phemotypes are compromised as message carriers. This information loss means there must be a reconstitution of the message by its receiver: this is the famous Chomskian "poverty of the stimulus" argument concerning linguistic message-passing. But if there is significant reconstruction of the informational content of a meme by each host brain, then the likelihood of message replication is low, thanks to the vagaries of how each brain processes in-coming information (due to the different background information individuals have acquired, the inferencing algorithms they use, etc.).
One way out of this problem, suggested by Dan Sperber (who also brought this problem to our attention), would be for the brain to have a general decoder -- a utility enabling it to reliably infer the intention of the sender, and hence the substance of the message. In fact, this is what the much-lauded "theory of mind" module is likely to be for (although this, in itself, is contentious; most cognitive scientists believe agents reason upon others' minds as a special application of "instrumental reasoning" with special application to social world, rather than granting such a capacity a special modular status). In this view, brains have evolved filters to assess the utility of information coming in from the social environment to keep us from rapidly being swamped with bad information (i.e., duped into stupid behaviors by people with ulterior motives). This normalizing inferential machine might also ensure the replication of memetic material during social transmission. However, its operation is unlikely to be perfect, so a high mutation rate remains a potential problem.
The need to communicate memes between brains also introduces another problem. If such psychological normalization of memetic inputs is important for communication to be successful, then memetic information is not, strictly speaking, inherited because it is not passed from person A to B. Instead, the similarity of socially-acquired information between individuals has another cause: inherently structured inferential processing by the brain. These reconstructive processes depend on a long history of genetic selection on the human cortex, not the passing of information from person to person in cultural lines of descent. In effect, the cause of the similarity between the information in A's and B's brains is thanks to evolutionary psychology, not memetics. Since the causes are different, one can expect the population-level dynamics (e.g., rates of mutation, types of selection) to also be different. This creates a fundamental problem for memetics as an inheritance process (the most general view on memetics).
Dan Dennett pointed out, however, that the memetic process -- even if dependent on error-correction routines in the brain to produce cultural similarity of beliefs and values -- still confers an evolutionary advantage. This is because the same information is acquired through transmission-plus-correction more efficiently and cost effectively than individual learning through trial-and-error could have done. Further, error correction is an important aspect of genetic inheritance as well, so replicator systems can operate with such assistance without having to be called something else.
It was also noted by Susan Blackmore that Dan Sperber's reasoning leads to the expectation that, if there is a cultural replicator, there should also be selection for improved mechanisms for its transmission over time. In this way, the reliance on reconstituting information from local resources each iteration would be reduced and the proportion of information actually being transmitted increased. Her presumption is that this is indeed what has happened during the major transitions in cultural evolution, such as language, writing, and computer-based communication. But whether these have increased the transmissability of memes, or merely their copying fidelity, remains to be determined.
A somewhat different view of memetics was presented by Liane Gabora. She takes her cue from neuroscience and complexity theory. More particularly, her critique of the replicator perspective is that memes do not function in isolation, but as parts of a complex conceptual network or worldview; this should be the basic level at which cultural evolution is analyzed. In Gabora's vision, the tendency to see memes as discrete, identifiable units is problematic. As parts of the brain, memes are intimately involved in highly distributed, highly structured and interconnected networks of neurons. Therefore, the means by which they are stored and evoked are highly contextualized. So, memes spend most of their time as parts of streams of thought which are continuously renewing, revising, and reassessing the constituent memes. It is the analysis of these more holistic streams of information which should be our primary focus. This can be accomplished, Gabora argues, along lines she has pursued in her own work.
Initally, Richard Dawkins argued it was important to consider the evolution of memeplexes such as religious doctrines. What happens if some component of the complex is missing? Does its ability to function degrade significantly? Memeplexes are another instance of a transition in evolutionary complexity, which should be approachable using "major transition theory" as it is being developed by Eoers Szathmary, John Maynard Smith, Richard Michod and others.
Another major set of issues concerns the psychology of memes. The question which dominated discussion in this workshop is whether memetics can proceed without a clear idea of what kinds of transformations memes undergo during storage and retrieval by brains. Can memetics leave the brain as a black box, and deal only with social transmission aspects? The virtue of ignoring psychology is that we can simply talk about inheritance processes at the population level and not worry about something we don't know too much about anyway: how the brain processes information. On the other hand, if memetics disregards psychology, and there are major transformative processes at work in the brain, then memetics is only explaining part of the cultural evolutionary process, and may therefore "get it wrong." It was felt by some (particularly Rosaria Conte) that no social theory, including memetics, can succeed without a proper psychological underpinning.
Related to this question is the relationship of memetics to imitation. Two interconnected questions pop up here: First, Does imitation require a complicated brain to do? This issue is important because it determines who gets to have memes: only complex intentional agents like people, or more lowly creatures without cortices, such as birds? Many (including Henry Plotkin) argue that there is no consensus concerning the psychological mechanisms of imitation. This is significant because, as Rosaria Conte says, you cannot define imitation without reference to the mental abilities involved, because using behavior as the sole criterion leads to confounds. For example, automatic contagion (such as yawning) is direct phenotypic copying without the inferencing of mental contents. Counting contagion as a kind of imitation suggests that agents don't need to correctly infer another's intention (plus her beliefs and needs, etc.) in order to adopt or imitate her behaviors. What psychological resources imitation demands remains an open question.
Second, Should memetic transmission be restricted to imitation? Susan Blackmore restricts memetics to cases of imitative behavior because, she asserts, only imitation serves as a direct copying process, and if memetics is to be founded on replication events, then only imitation can be counted as a memetic mechanism. But as we have just seen, the jury is still out on whether imitation is behavior copying or mental state inferencing (as assumed in the "theory of mind" literature). This leaves Blackmore's contention somewhat up in the air.
Because of the general discontent with imitation (due to its problematic psychological status), a satisfactory resolution of these interlocking issues was not achieved. However, it was generally felt that all social learning, rather than imitation alone, is a better psychological foundation for the cultural evolutionary process. The famous example of milk bottle-top opening by birds was presented as evidence. The pecking of bottle-tops has now gone on for many bird-generations, and spread through several European countries. Since it is generally felt that birds learn this bit of cleverness not by observing others, but by seeing opened bottle-tops, which inspires their own creativity (a process psychologists call "stimulus enhancement"), it seemed a pity to exclude such an example from the purview of memetics by limiting it to imitation-based diffusion.
However, if this liberal position on social learning is adopted, many repercussions ensue. For example, the phylogenetic history of memes suddenly becomes considerably longer, with birds and perhaps even more "primitive" creatures being allowed to have meme-based "protocultures." If so, then what distinguishes human culture from non-human culture? In addition, it means that direct contact between hosts is no longer required for memetic transmission, since the source of a meme (such as the tit which pecked a bottle-top) can be absent when a new, naive tit arrives on the doorstep. It is the artifact left behind -- that is, the pecked bottle-top itself -- which serves as the proximate stimulus for transmission of the pecking meme to the new arrival. This implies, in turn, that memetics must account for artifact production, since memes can be associated with these constructions, and not just brains. Such implications were not discussed at the meeting.
A second major point of contention was whether the memetic dynamic can be extended into the brain. Can we call individual learning a selection process just like the social transmission process? This proposal met with some disdain, and Henry Plotkin noted that at least among academic psychologists, this is definitely a minority position. Susan Blackmore argued that whatever is happening inside the head should not be considered part of the memetic process; if it is in fact selectionist, it still should be recognized as an independent replicator system. Others argued that including selection among alternative mental representations was crucial to a successful memetics. Two benefits were seen to result from this move by proponents. First, only through an analysis of mental properties and processes can good models of transmission mechanisms, such as imitation, be understood. Second, by extending the Darwinian process into the brain, the confusion of calling thinking --the manipulation of memes -- "directed," "intentionalist" or "Lamarkian" could be avoided. It was felt that the unpopularity of a selectionist psychology may be largely due to the appearance it gives of there being no room for human agency or decision-making, that all human psychology is merely a random selection process among alternative behavioral choices. Of course, the abandonment of intentionality and free will was hailed as a victory for memetics by the mental selectionists. But no consensus was reached on this issue either.
From the perspective of this group, memetics is largely a promise at present, with no real results to show for itself. As such, the question is whether it will contribute anything new. It was felt that a quasi-epidemiological approach similar to memetics is already in widespread use in the social sciences. The idea that some cultures are more stable, or produce a higher quality of life because certain ideas spread better than others, has long been around.
What memeticists don't recognize, these critics argue, is that one can have a theory of cultural change which is not memetic because it does not involve social learning or the spread of particulate bits of information. For example, evolutionary psychology explains cultural change simply by invoking variation in what the environment stimulates people to recall. All the information for cultural processes is considered by evolutionary psychologists to be already in place in people's heads; the inheritance of these bits of information is genetic. What remains to be explained is not social transmission dynamics, but recall dynamics: what kinds of responses do different environments cause to arise? Thus, existing explanations of diffusion and spread exist which do not invoke memes. Memeticists miss this "Big Picture" because they are largely unaware of the comprehensive literature which has accumulated in anthropology on cultural change, or the actual history of earlier views (e.g., the cultural diffusionists of the early twentieth century). This leads them to reinvent old wheels and postulate new terms for rejected ideas. Memeticists do not even recognize that the concept of culture -- the thing which memetics intends to explain -- is itself sufficiently problematic that some social scientists advocate its abandonment. The notion simply covers too complex and varied a set of processes to be useful in their view. (What exactly would replace the concept of culture, or what sub-concepts it should be divided into, is not obvious, however.)
What remains unclear to this group is the central claim of memetics: whether there is a novel replicator-based process underlying the population-level, epidemiological dynamic that is culture change. The primary problem of memetics, therefore, is whether there is a new entity on the horizon in whose interests things can be said to happen (the "meme's eye view"). This would be a new kind of function which a social institution might serve: that of the memes. As such, it would represent a real and novel alternative to group-level functionalism, or the various flavors of structuralist thought current in the social sciences. Unfortunately, this central claim has not yet been proven.
In sum, memetics is seen as simply another case of those from outside the discipline, in this case largely biologists, "having a go at explaining culture," but without taking into account many of the complexities this project is widely recognized by contemporary social scientists to entail. The meme critics are happy with the general notion that cultural change involves the diffusion of some vaguely characterized entity, but not with an explanation couched solely in terms of the selection, variation and inheritance of a particulate replicator.
It was also remarked that there is a problem of circularity in the way memetics is generally conducted. Memeticists only study things which seem likely to follow a memetic process, like fashions and fads (e.g., the infamous backward baseball cap). The perceived success of such empirical adventures leads memeticists to self-congratulation. But many aspects of culture aren't small, isolatable bits of information or practices that readily diffuse in observable time. Take the example of language, which permeates every aspect of culture. How does memetics expect to explain these more fundamental components of culture?
The word "meme" itself has problems. Its close parallel to "gene" may lead memetics astray, if in fact memes are not the same kind of thing. It also produces a "revulsion factor" among those who would otherwise be friendly to the Darwinian cause. Memetics is perceived from outside as an arrogant usurper, making extreme, unwarranted claims. This only serves to put memetics in the same basket with a related attempt at explaining human social life, sociobiology, which was widely seen as what Dan Dennett calls "greedily reductionist." Sociobiology left no ground for social scientists to stand on, and all the interesting questions were subsumed under a single algorithm: the maximization of biological fitness. This is unpalatable to social scientists not just because of territoriality disputes, but because such a greedy reduction is bound to failure. Can all social processes really be reduced to selection and transmission? The box of concepts available from Darwinism doesn't impress this group. It seems a very small toolkit when so many theoretical alternatives are already available and there is so much complexity to explain. In fact, theory abounds in the social sciences. What is lacking is insight into real social processes. Explaining these seems a goal quite far removed from the concerns of most memeticists, who are laboring much further down the organizational hierarchy, worrying about replicators. An uphill battle against a wide variety of other approaches therefore lies ahead for memetics in the social realm.
Many participants observed that despite the shared belief that an evolutionary approach to culture was necessary, significant barriers to communication remained between those from different disciplines. This perhaps derived from the varying histories these disciplines have with evolutionary approaches. In particular, social anthropology has a long history of such thought, which has generally not proven successful. Indeed, a common refrain among those social anthropologists participating in the meeting was "been there, done that." It was difficult for "believers" in memes to convince these historically mindful and hence reticent social scientists that this time around things might be different. Similarly, it was difficult for the anthropologists to explain exactly what went wrong previously, or specifically how the memetic perpsective was likely to go wrong itself, even if given a clear run at explaining culture.
This incommensurability of ethos led to an undercurrent of dissatisfaction on both sides. One side seemed to feel that having to address the concerns of "non-believers" kept progress back, while the opposite side felt that the believers "just weren't getting it." Nevertheless, most agreed that bringing both sides together decreased the likelihood that proponents would engage in unchecked, hubristic claims about having explained culture (along with other conundrums such as consciousness), or that social anthropologists would continue to ignore the memetic alternative. Nevertheless, while I don't think anyone was persuaded to jump from one camp to the other, both sides did go away with a lot to think about, and increased respect for those who disagree with them.
A general disappointment was the lack of discussion about what might be called "applied memetics." More time certainly needs to be devoted in future to thinking of ways to do memetics. This should include discussion of existing empirical studies that don't go under the banner of memetics but which could be interpreted as falling within the general purview of this incipient discipline, as well as the development of methodologies for conducting specifically memetic studies in the future. This is because the ultimate test -- which would preempt theoretical objections -- is whether memetics can produce novel empirical work or insightful interpretations of previous results. Everyone agreed it has not yet done so, but must do so in the near future, given the extensive theoretical work already accomplished and the high level of current interest in the subject. Otherwise, it is likely that memetics will soon be perceived to be a failure. This might be considered unlikely if only because, as one participant remarked, just being able to assemble such an eminent, multidisciplinary group to discuss the topic underlines how these ideas are coming to have real force in contemporary intellectual discourse.
Dr. Robert Aunger King's College Cambridge CB2 1ST
Dr. Michael Best Media Laboratory E15-325 MIT 20 Ames Street Cambridge, MA 02139 USA
Professor Susan Blackmore Department of Psychology University of the West of England St Matthias College Bristol BS16 2JP
Professor Maurice Bloch Department of Anthropology London School of Economics Houghton Street London WC2A 2AE
Professor Robert Boyd Department of Anthropology University of California 405 Hilgard Avenue Los Angeles, CA 90024 USA
Professor Rosaria Conte Division "AI, Cognitive and Interaction Modelling" PSS (Project on Social Simulation) IP/Cnr, V.LE Marx 15 - 00137 Roma, Italia
Professor Richard Dawkins Department of Animal Behaviour University of Oxford
Professor Daniel C. Dennett Center for Cognitive Studies Tufts University Medford, MA 02155-7059 USA
Professor James E. Doran University of Essex Department of Computer Science Wivenhoe Park Colchester, Essex CO4 3SQ
Liane M. Gabora Center Leo Apostel Brussels Free University Krijgskundestraat 33, B-1160 Brussels, Belgium
Professor David Hull Department of Philosophy Northwestern University Evanston, Illinois 60208 USA
Professor Nicholas Humphrey Centre for Philosophy of Natural and Social Science (CPNSS) Tymes Court Building London School of Economics Houghton Street London WC2A 2AE
Professor Adam Kuper Department of Human Sciences Brunel University Uxbridge, Middlesex UB8 3PH
Dr. Mark Lake Institute of Archaeology University College London 31-34 Gordon Square London, WC1H 0PY
Dr. Kevin Laland Royal Society University Research Fellow Sub-Department of Animal Behaviour University of Cambridge Madingley, Cambridge CB3 8AA United Kingdom
Dr. Neil Manson Fellow in Philosophy King's College Cambridge CB2 1ST
Professor John Odling-Smee Institute of Biological Anthropology University of Oxford 58 Banbury Road Oxford OX2 6QS
Professor Henry Plotkin Department of Psychology University College London London WC1E 6BT, U.K.
Lord W.G. Runciman Trinity College University of Cambridge Cambridge CB2 1TQ
Professor Stephen Shennan Insitute of Archaeology University College London 31-34 Gordon Sq. London WC1H 0PY
Hans-Cees Speel School of Systems Engineering, Policy Analysis and Management Technical University Delft Jaffalaan 5 2600 GA Delft PO Box 5015 The Netherlands
Professor Dan Sperber Centre de Recherche en Epistémologie Appliquée Ecole Polytechnique 1, rue Descartes 75005 Paris, France
Dr. Richard Webb Centre for Philosophy of Natural and Social Science (CPNSS) Tymes Court Building London School of Economics Houghton Street London WC2A 2AE
Thanks to Rosaria Conte, Liane Gabora, David Hull and Henry Plotkin for feedback on this report.