Work in Progress presented at the International Workshop on MODELLING AND SIMULATING CULTURAL TRANSMISSION AND CHANGE (CNR, Università di Siena), Certosa di Pontignano, September 3-4, 2000.

Deep and Shallow Cultural Modeling

Patrick Boylan

Department of Linguistics

University of Rome III

0. Culture, as we experience it, is the organizing principle, worked out through meaningful interaction with others, that enables us to make sense of the events in our lives; in a word, it is our socially-elaborated Weltanschauung ("world-view" or "mindset"). Expressed in these terms, however, the concept of "culture" remains just that: a concept. It can be contrasted with other definitions of "culture" as to conceptual accuracy and usefulness -- do people in fact intend principle as the genus and sense-making as the differentia when using the word "culture"? does this definition stimulate further enquiry? -- but it is not easy to test empirically or use sociometrically. Thus the search for a model capable of making culture amenable to taxonomic classification, quantification, experimentation, computer simulation and, ultimately, social management. But what kind of model ought we to search for?

Models may be shallow or deep: the first kind furnish descriptive accounts of observed phenomena while the second furnish meta-level explanations. In a young science such as cultural studies, both kinds are used independently one of the other. Deep models of culture, usually introspective and difficult to test or use, speak one language while shallow models, generally empirical and thus highly productive, speak another. The present paper addresses this regrettable gap. It presents a deep model of culture, based on the definition given above, and seeks to ground it empirically. The aim is to stimulate further research along lines which are sufficiently introspective to account for culture as a subjective experience, while at the same time offering hooks that are empirically testable and productive.

No claim is made that deep modeling is preferable to shallow modeling in furnishing us with an operational definition of "culture" . Both offer real advantages for the advancement of our knowledge of how human beings -- as well as some animals and perhaps even certain forms of plant life -- concert their actions by means of information learnt from one another and translated into common behavior. Indeed, what this paper suggests is precisely the need for both kinds of model, as well as for a way to make them dovetail.

1. Shallow modeling of culture starts from an empirical description of behavior and then works "from the outside in" without changing the level of analysis (usually physical) or the explanatory metaphor (usually mechanical or probabilistic). One end-product of this approach has been the theoretical construct called "memes" (Dawkins 1976). According to Dawkins, the commonly-shared behavior of the members of a given community, characterized by particular ways of doing things which we may call the community's cultural traits, are associated with commonly-shared data patterns (or, expressed in strictly physical terms, data packets) which the members transmit to one another and store in their individual long-term memories. Each pattern or packet is called a "meme".(1.)

Note that this model is, at its core, an economic model of data transmission in large populations; it does not really explain how behavior determines culture and even less how culture determines behavior. Let us make this point clearer by means of an example. In a wolf pack or in an aborigine clan, the characteristic behavior of the leaders is said to end up somehow in the minds of the young under the form of patterns of data stored in their long-term memories. If "patterns of data" is taken to mean packets of visual cues, the assertion says no more than what we already know from ethology and the psychology of perception. But if the claim is that the "patterns of data" stored in the brain constitute "culture" -- in this case, the culture of an aborigine clan or of a wolf pack-- the assertion begs the central question of what culture is. In gratuitously equating "culture" and "remembered perceptions", it leaves us in the dark as to how the sensory data stored in the brain actually become (collectively willed) imperatives in a given subject's psyche, i.e., imperatives capable of determining (culturally authentic) behaviour. For as we will argue later on in this paper, culture is more than the sum or even the synthesis of sensory data; it is an original elaboration requiring intense cognitive, affective and especially volitional activity (in animals, instinct substitutes for much of the volitional component). Without such activity, the data packets will not constitute a stable, integrated whole that is cognitively consonant (Festinger 1957) and volitionally coherent (Bateson et al. 1968). In other words, without such activity nothing assures us that, among the aborigines if not among the wolves, the data transmitted and stored in memory will necessarily determine acculturation and, as a consequence, acceptable clan or pack behavior. Most communities have, in fact, more than a random number of eccentrics and innovators, not to speak of rebels and outsiders, whose divergent behavior must be due to something more than imperfectly remembered perceptions.

In conclusion, meme theory can be said to predict "culture" from behavior and vice versa only if all we are interested in are statistical correlations and do not ask what culture is, nor why the relationship between so-called culture (remembered perceptions) and behavior holds in a significant number of cases, nor why that relationship does not hold in a non-random number of apparently similar cases (i.e., cases in which subjects receive the same sensory data input). These explanatory weaknesses are inherent in shallow modeling and outside-in approaches; they are precisely the weaknesses that this paper will try to address.

2. In spite of its inherent weaknesses, the rigor, elegance and theoretical simplicity of shallow modeling makes it nonetheless attractive. In addition, it satisfies the anti-Cartesian slant that modern science has chosen to give its research. Instead of positing two distinct spheres, the mental/transcendent and the physical/immanent, shallow modeling places all phenomena on a single cline, stretching from the purely physical to the mental-as-a-product-of-physical. The often-cited analogy is with ice, water and steam: three physical states which look different but which, in reality, are simply different energy level states of the same H₂O molecules. Similarly, in meme theory data packets start out as photons, then become chemical-electrical charges and finally neural nodes. Moreover, precisely because it is a shallow model, meme theory is able to place on a single cline both human acculturation (for example, that of young aborigines) and non-human adaptation-through-learning (for example, that of young wolves). What varies is only the complexity of the mental representations (clusters of neural paths) in the brains of the various subjects. The sheer simplicity of shallow modeling makes it a very powerful explanatory device, indeed.

Shallow modeling of culture is also attractive because it lends itself to explaining large-scale cultural changes in evolutionary terms, i.e. as a probabilistic product of random variability and natural selection (Heylighen 1992). This satisfies the Darwinian slant that modern science (not to speak of mainstream market-oriented economics and social policy) often prefers for its "efficiency".

Finally and most important, shallow modeling of culture makes it easy to verify hypotheses empirically and translate them into simulations and practical applications. It gives researchers a hold on the otherwise slippery concept of "culture". The promise held out is that, if the "mental data packet" model proves to be a satisfactory representation of "culture", then the jump from such packets to real OSI-protocol data packets in a computer network will present no difficulties. At that point, we will be able to simulate, analyze and manage cultural phenomena with ease. (2.)

3. Such buoyant optimism, typical of all young sciences, should not be discouraged lest the fervor (and the funding) come to an end. But it should be put more into perspective. Being able to map the "cultural configuration" of a human brain in terms of data packets may indeed enable us one day to do things like define and localize the cultural families of the world, as Cavalli-Sforza & Feldman (1981) have done with human genetic families. But while the "meaning" of genes is exhausted in the immediately evident physiological traits they contribute to produce, the "meaning" of cultural traits goes more than skin deep. To grasp it, we need to move beyond shallow modeling.

Cultures are not simply collections of artifacts that the members of given societies have produced over time: jewelry, laws, greetings, hunting techniques. Nor are they simply the instructions for making and using such things, largely unwritten and stored in the memory of individuals as patterns or packets of data called memes. Cultures are laboriously worked out and tenaciously willed existential "stances" or ways of positioning oneself with respect to certain objects or activities. To use an analogy, cultures are not nouns and verbs but rather noun affixes and verbal moods. People in a community develop a culture in order to give sense to the world as they have chosen to see it; words follow (and do not precede) that vision. Indeed, a very simple definition of culture is "that which links what we see to what we say". This is why, in order to grasp the culture of another society, we must enter into the mindset underlying its concrete manifestations; for it is by means of that mindset that natives "see" those manifestations and give them their names.(3.)

Only if we understand cultures at this level of abstraction will we be able to furnish international negotiators and intercultural policy makers with the kind of training they require to operate effectively. Only then will we know what mental framework we, too, must acquire if we wish to interact successfully with members of another culture, whether they be our guests or our hosts. For simply knowing the "facts" of their culture, i.e., having in our minds data packets similar to theirs, does not in itself make us see things their way, just as it does not necessarily enable us to say things their way, i.e., in a way that is as easy for them to understand and accept as possible. What we need instead, in order to achieve a real cultural entente with them (just as with people from our own culture), is to undergo a particular kind of volitive transformation so that, instinctively, we:

want the kind of relationship they would try to create if they were in our position (not just the "tollerantly neutral" kind found in most international bureaucracies);

laugh at their jokes (not just understand them intellectually) and know, in heated discussions, when enough is enough by empathy (not just by trial and error);

end up speaking their lingo (not just their language).

This kind of volitional, affective and epistemic knowledge is the product of a reworking of our value system, our appetites and our experiences in the light of a newly introjected Weltanschauung (that of the other culture). It cannot be properly described in terms of patterns of information or packets of data that are somehow transmitted from external sources and stored in the brain as such.

4. It should come as no surprise that shallow outside-in modeling cannot adequately represent the complexity of cultural knowledge. After all, such modeling has earned its name precisely because it is fundamentally tautological. What it hypothesizes inside the mind (e.g., memes) cannot, by definition, be any richer in information than what it verifies empirically in the outside world (e.g., social behavior).

True, as in algebraic equations, we can use a shallow cultural model to reformulate our data in such a way as to shed new light on what we already know. The concept of "the viral transmission of culture" is a good example. By taking inspiration from epidemiology, memetics (the science of memes) is able to trace how cultural hackers -- from advertising copywriters to pop singers -- get people mindlessly to pass on "virus-laden" ad slogans, clothing styles, and so on (Dawkins1993). In other words, insofar as we human beings do in fact act mindlessly a great deal of the time, our behavior can be modeled as though we were body cells beset with biological viruses (or PC memories beset with computer viruses). Nothing surprising here: works on mob psychology came to a similar conclusion a century ago: Gustave Le Bon's The crowd: A study of the popular mind was published in 1895. Where memetics innovates, however, is in taking the body metaphor literally and in applying the descriptive tools of epidemiology to meme diffusion, a potentially fruitful tack indeed.

But if memetics, using shallow modeling, is able to explain how cultural hackers can fool most of the people most of the time, it is not able to explain why cultural hackers are unable to fool all of the people all the time, or why, every so often, not just random but numerically consistent segments of society -- even entire populations -- rise up and say no to the sociocultural, political or religious models currently prevailing. In these cases the images transmitted "from the outside in" somehow fail to acculturate any more. A complete theory of cultural development and propagation must explain these cases as well.

Some theorists claim that evolutionary memetic theory is enough to account for massive, non-random changes in a species' behavior (Heylighen 1992). The cause of such changes, it is said, are to be sought in some corresponding radical change in the environment. To take an example from the animal world, if a disease kills off all the deer in a mountain region, leaving only rabbits for prey, wolves will learn to change their cultural habits from group to individual hunting. (We may even exclude learning by simply hypothesizing that some wolves will randomly select individual hunting tactics and these are the wolves that will survive; the others, hunting in a pack as before, will expend, for every rabbit caught, too much energy for too little nourishment per head). In other words, by adding an evolutionary component, our shallow model of culture is able to explain the wolves' apparent "change of behavior" in terms of photons, neurons and neural path clusters, just as before: in certain wolves' brains, "prey" becomes mechanically associated with "rabbits" (because of changes in the environment) and randomly associated with "one-on-one hunting".

This kind of causal explanation, however, does not always work with human beings, since their environment is largely composed of the cultural artifacts that they have made for themselves and that they use to make their habitat distinctly artificial. How does change occur in an environment where Mother Nature takes back seat and where most of the significant changes are strictly man-made (memes changing memes)? Once harmony with the environment is achieved, how do occasional "dissident" memes manage to gain a foothold and eliminate rival memes?

To use a historical example, the change from aristocratic feudalism to democratic capitalism in Europe took place when the economic system, prevalently based on agriculture, became industrialized. What explains this change in the mode of economic production? The absence of natural causal phenomena leaves us in a quandary. If we say that democratic capitalism arose in response to industrialization but that industrialization arose in response to the rise of capitalism, we have a vicious circle that does not explain either phenomenon in terms of strict causality. Moreover, if we seek to ascribe the rise of industrialism to other cultural changes (the Guilds, the Crusades, etc.) we have to explain what cultural modifications produced those changes. The chain could go on forever. Thus, by excluding a quirk natural catastrophe as a cause (as in the case of the wolves and rabbits), we are faced with the task of justifying, in strictly cultural terms, how a consistent part of the population of Europe rejected, in a given historical moment, the acculturation process that had held the feudal system together for centuries. If we invoke the evolutionary paradigm by hypothesizing a series of random changes producing greater fitness among the social innovators, we have to explain why those changes didn't occur in the previous centuries. And if, to do so, we invoke the concept of "socio-economic conjuncture" (the catalytic effect of a fortuitous coming together of quirk favorable conditions), we have to explain how those memes, i.e. the quirk favorable conditions, survived when they were not to become truly useful until the moment of conjuncture.

To be sure, the theorists of evolutionary memetics are an ingenious lot: they will probably have no trouble in devising explanations to answer such objections. But why flog a dead horse? The very kind of question we are raising suggests that the kind of answer we are looking for lies elsewhere.

Dialectical materialism furnishes a plausible explanation of radical historical changes: the cultural world -- and thus economic, social and political history -- is intrinsically unstable because it is a product of the human mind, which reasons dialectically. This explanation does not, however, account for the subjectivity of culture. For that, we must turn to another anti-Hegelian current of philosophy, stretching from Dilthey through Brentano to Husserl, which we may loosely term "phenomenology". (Sartre's existentialism is a distant cousin.) In this perspective, cultural phenomena, indeed all knowledge and beliefs, are said to be a product of concerted human wills interacting with the constraints of reason (just as, in biology, phenotypes are said to be the product of genotypes interacting with the constraints of the environment). Now, if knowledge is intentional, then it follows that culture is willed into being. What consequences this affirmation has on the modeling of culture we shall see further on.

In conclusion, evolutionary (or epidemiological) explanations of cultural changes are insufficient: genes and viruses do not require motivated consent to spread. Culture does. It is not necessary at this point to decide whether "motivated consent" implies "free will" or, indeed, whether "free will" actually exists. We know introspectively that the subjective reality of will exists and this must be accounted for, since it is on the basis of this subjectivity that people are said to act, believe they act, and are held accountable for the way they act. Thus, a satisfactory theory of cultural development and propagation must define how humans choose (or think they choose) to be what they are not yet, and how culture is experienced -- at least in part -- as a project and not simply as a conditioning coming from the environment.

This way of proceeding -- founding theory on the subjectivity of human experience -- may seem anti-scientific. After all, a physicist does not seek to account for what people subjectively feel "gravity" to be, but rather what gravity does independently of how it is experienced. A geneticist tracks the historical displacements of populations by interrogating, above all, the subjects' genes (what the subjects themselves offer as oral testimony is precious but complementary: see Cavalli-Sforza & Feldman 1981). But cultural studies, along with all "moral sciences"(4.), require a different approach. For it is the subjective experience of willed existential meaningfulness, synthesized in a Weltanschauung, that determines the behavior both of individuals and of communities -- or at least insofar as individuals and communities may be said to act intentionally.

----------------------

The following sections await completion:

The lesson from L2 learning: culturally authentic discourse

"Being" and "willed into being"; stance-taking

Volitional knowledge (cf. affective knowledge and cognitive knowledge)

Modeling stance-taking: Parisi and Castelfranchi's "goals" not enough

Need to switch levels in modeling culture with respect to other learned behavior

Dawkins etc. portrayal of culture as viral (blind transmission) is therefore inappropriate.