Report from the Consilience conference, part II

The second day of the consilience conference started with Peter Turchin on cliodynamics (after Clio, the muse of history), or what he thinks may become a science of history. Peter argues that historians are incorrect in thinking of history as part of the humanities, and certainly wrong in having abandoned the search for overarching explanations of historical events. The author’s suggestion is that historians actually do deploy general frameworks for explanations, but do so implicitly, which screens such general frameworks from critical and quantitative analysis.

Peter proceeded to examine one example: the fall of the Roman empire. A German historian at some point counted 210 different explanations for that historical event. (Hmm, that seems hard to believe, I would like to see in what sense these are actually “different,” there just doesn’t seem to be that much room in logical space.) The suggestion is that this proliferation is due to the fact that historians are unable to eliminate hypotheses that don’t work, thereby leading to an ever increasing proliferation of conjectures not followed — to use Popper’s phrase — by any refutation.

Turchin went on to show some interesting data quantifying gridlock in the American Congress over the past several decades (e.g., number of judges confirmed, number of filibusters, etc.). The data clearly show a huge increase in polarization after 1970. He points out an apparent cyclical process of party polarization between 1800 and today, suggesting that if it is cycling, there may be an underlying unifying explanation for it. I think the data are fascinating, but the time series is far too short to pinpoint cycles, and historians (and, more pertinently, political scientists) would probably be able to argue that the factors explaining the ups and downs in political polarization were different for different time periods (e.g., the Civil War was significantly different from the current corporate-ideological wars). It also strikes me as strange to argue that historians don’t make use of quantitative data when they are available and pertinent, and it seems to me that a likely problem with the cliodynamics approach is that the quantitative data (unique historical sequences) often, though not always, will underdetermine possible explanations. The same thing happens in other quantitative social sciences, for instance economics. Again, this doesn’t make a quantitative approach useless, but it also imposes strong constraints on its utility.

Turchin proposed another example, this time based on a much longer time series, quantifying social instability in France between 800 and 1700. The graph shows a number of peaks and valleys. Peter sees four “waves” of instability: the Carolingian break-up, the Early Medieval crisis, the Late Medieval crisis, and the religious wars of the 17th century (he presented a similar graph for ancient Rome, with three periods of instability). While he talks about recurring patterns, I am reminded of similar graphs by paleontologists tracking periods of mass extinction. For a while, it was fashionable to look for periodicities and a single underlying explanation (like a recurring small star companion of the Sun), but they didn’t work out, and we now think that each mass extinction has its own ecological and/or astronomical explanation.

Next we had Christopher Boehm, on social selection and the notorious free rider problem. George Williams set the stage for the problem when he defined free riders as cheaters who defeat altruists within a group. Boehm is interested in a particular kind of free rider, the bully. They are found in any social dominance hierarchy, from chimpanzees to modern humans. He compared 150 modern hunter gatherer societies that are similar in structure to Pleistocene humans. Of these, 49 have data coded for sociopolitical variables.

These societies have a home base with centralized meat provisioning, there is male-female division of labor, and they cooperate altruistically to share meat. Importantly, subordinate band members form coalitions to hold down bullies, resulting into a egalitarian society. The group acts by consensus in sanctioning deviants, with sanctions ranging from social pressure and criticism to shaming, ostracism, ejection and even capital punishment.

The key to moralistic social control is gossip, which permits private evaluation of deviants and allows consensus about a deviant to develop, which leads to group action. The targets of social control are primarily bullies, thieves and cheats, and people engaging in sexually unacceptable behavior. Interestingly, bullies are often executed in these societies, many more than cheaters or sexual transgressives. This, according to Boehm, indicates that bullies are a primary type of free rider in the mind of Pleistocene-like hunter gatherers. Only about 9% of sanctions are irreversible (death, permanent expulsion), with reversible sanctions allowing deviants to reform with only a partial loss of fitness.

Boehm considered six theories of altruism: (1) Mutualism (you scratch my back, I scratch yours);  it is so immediate that no cheating is possible, therefore it raises no free rider problem. (2) Reciprocal altruism (long-term mutual back scratching); it is subject to the free rider problem. (3) Group selection (between group effects trump within group effects); subject to free rider. (4) Piggybacking n. 1 (misplaced nepotism, bonding and generosity transferred to non-kin as a maladaptive effect); susceptible to some free riding. (5) Piggybacking n. 2 (cultural learning is good for you, but one of the things your culture is teaching you is to be altruistic, and if you listen to it, you’ll lose out); subject to deceptive free riding, for instance by bullying. (6) People’s choice as an agency of selection; this can take place by reputation or collective sanctioning. I’m not sure in which sense these count as “theories,” they rather seem to me to be different scenarios, actualized to different degrees in different societies.

The third speaker of the day was Joseph Carroll (the organizer of the conference), on Graphing Jane Austen and the evolutionary basis of literary meaning. He set up a web questionnaire on 2000 characters from 202 British novels of the 19th century, and he sent requests for comments to a number of literary scholars encouraging them and their students to provide feedback. This approach strikes me as analogous to, say, “experimental philosophy’s” surveys of people’s attitudes toward issues like free will or consciousness. I find the latter to be somewhat interesting psychological surveys of how people think about philosophical issues; I do not think they have anything to do with philosophy, though. (Here, of course, a difference is that Carroll and his collaborators asked professionals and their students, not lay people.)

The point was to gain insight into the ethos of individual novels (based on how people perceived the different characters), and then to be able to generalize to the entire Victorian period, once enough novels had been so analyzed. The underlying reasons for the study were to demonstrate that major components of literary meaning can be reduced to the elements of human nature (reduced? Built upon, perhaps?); to test empirically the hypothesis that agonistic structure reflects evolved social psychology (evolved here means biologically, since the author deferred to evolutionary psychology); and to generate new empirical knowledge and begin a process of knowledge acquisition in literary study.

(Full disclosure, the author took pot shots at my own, somewhat skeptical, presentation from the first day, incidentally deeply mischaracterizing it. Oh well.)

The model of human nature used in the study included motives (impulses, instincts, goals of action), mate selection, personality differences, and emotions (of the readers, not the characters). A factor analysis highlighted “dominance” (quest for power, wealth and prestige, and tendency to not helping non-kin) and “nurture” (negatively correlated to short-term mating and positively to helping offspring and kin). This is beginning to look like the famous scene in Dead Poet Society where Robin Williams graphs the characteristics of a novel only to show his students how silly that exercise is, missing the point of literature. But Carroll was very serious about this. (Another analogy might help: most word processors can collect data on the structure of your writing, such as average sentence length, average word length, and so on. I seriously doubt, however, that a good English teacher needs that sort of quantification to tell you whether your essay is worth a crap.)

Mate selection was described by other factors, including extrinsic attributes (power, prestige, wealth) and intrinsic qualities (reliability, kindness, intelligence). Not at all surprisingly, male protagonists are interested in physical attractiveness of prospective males, not in their intrinsic or extrinsic qualities, while the reverse was true for female protagonists. In terms of the big five personality traits, the good guys and gals score highly on positive attributes and lowly on negative ones, while the reverse was true for the bad guys and gals (again, surprise, surprise!).

The data apparently disprove the following claims: (i) meaning in literary texts is undecidable; (ii) the novels center thematically on a struggle of power between the sexes; (iii) the novels merely push pleasure buttons. Of course, the first “theory” is postmodern, not difficult to argue that it is crap; the third one is by Steven Pinker, not a literary critic. The second would deserve more discussion, but Carroll went too quickly through the data apparently showing that the novels do not display a power struggle between the sexes. He prefers an interpretation according to which the novels show instead a struggle against cheaters and bullies. Surely both elements are present?

Finally Carroll gave a few remarks trying to tie his results to recent views in evolutionary psychology, according to which gossiping is important, and members of egalitarian societies are preoccupied with keeping bullies and cheats under control. Ironically, I think, this type of approach typifies a sort of mirror image of postmodernism-deconstructionism: according to the latter, the meaning of text is forever fluid and undecidable; according to the former, Pleistocene biology pretty much determines what is going on in literature. The multifarious and ever changing dynamics of human culture that makes Jane Austin and other authors such a pleasure to read get lost in both cases.

At the risk to be a bit simplistic here, the first question that comes to mind is whether one needed to do all this work to arrive at the conclusion that novels are written about the dramatic aspects of human foibles. Moreover, how does this sort of approach explain the differences between, say, Victorian and late 20th century novels, or between novels in the Western canon vs Japanese ones? Presumably “human nature” is the same across times and cultures (at least according to evolutionary psychologists), so what gives? Now, of course one can legitimately engage in the kind of data gathering that Campbell and his colleagues carried out. And one can obtain stats and graphs as a result. But I honestly didn’t learn anything new about Victorian novels throughout the presentation, not to mention that I doubt the memory of those stats and graphs will make it any more pleasurable to read Jane Austin from now on (unlike, say, a perceptive review by a good literary critic).

After lunch, we resumed with Michael Rose, on Darwinian evolution of free will and spiritual experience. His approach to these issues is based on evolutionary game theory as it pertains to behavioral strategies for social life. This consists in exploring quantitatively the dependence of fitness on strategies that specify an animal’s behavior in case of conflict.

Michael stressed that human intelligence must have evolved (and is currently maintained) by intense directional selection for increased brain size, otherwise natural selection would shrink our brains quickly. That is, there is a high cost to being smart. He directed a pretty good jab at other speakers by pointing out that detailed reconstructions of what early humans were doing in the Pleistocene’s savannas amount to little more than “science fiction” (ouch).

Hypotheses to explain human intelligence include the use of technology, but it turns out that the tools employed by early humans (already with large brains) were not much more sophisticated than those used by chimpanzees (which don’t need big brains). The currently fashionable hypothesis is the machiavellian one, that we evolved intelligence to deal with complex social situations. Again, though, early human societies were not that different from those of the Australopithecines, which did not need large brains to deal with their societies.

Another possibility is that selection favored a combination of tool use (particularly weapons) and social intelligence, which can trigger an arms race leading to a rapid and sustained evolution for larger and larger brains. Michael suggested that some (few) human behaviors can be analyzed by the methods used to study animal behavior, like incest. But that doesn’t get you very far from pop evolutionary psychology (hence the usefulness of game theory as an alternative approach). Incidentally, what he calls “free will” isn’t the metaphysical concept, but rather the social science idea that the human mind is not (much) constrained by past genetic evolution, that it’s a general purpose computer capable of making decisions that sometimes violate fitness requirements.

Rose’s suggestion is that this free willing feeling hides the fact that our computer brain does have anatomical compartments with specific functions, such as the prefrontal cortex, which calculate (subconsciously) the possible fitness outcomes of our actions. Evidence for this comes from neurobiological studies on what happens when the Darwinian fitness calculator is damaged: we get a host of psychopathic behaviors (1-2% of the population, committing 40-50% of total crimes). These individuals have a high mortality rate, and given their high incarceration rates, this behavior comes with a high fitness cost.

Michael then moved on to spiritual experiences. The neurobiological evidence suggests that we have this “second executive function” that subconsciously monitors your fitness functions (and is occasionally subservient to our conscious self). This second executive function is the source of the occasional awareness of some Other interfering in our lives. From there the step is (allegedly) brief to having spiritual experience and hence formulating the idea of gods. But, one wonders, how frequently do people actually have spiritual experiences? Enough to justify widespread belief in god?

Next to last was David Linden, on what notions from neurophysiology are useful if one wishes to connect evolution and culture. (Hmpf, another talk without slides.) Neurons are slow and unreliable, and yet they make possible “clever us.” This was achieved by expanding enormously the number of neurons during hominid evolution, something that had to happen without any radical structural redesign. Indeed, only the rough structure of the brain is specified genetically, the rest is the result of wiring induced by experience, beginning early on in utero. The rest of this talk was an informative, if didactic, overview of neurophysiology with comments on how the way the brain works affects our behavior in the world.

Finally, we get to Brian Boyd with a talk on “experiments and experience.” Art is like science in that it experiments with experience, yet literature doesn’t aim at building models, but rather at engaging people with the human experience. Evolution and cognition may “bear on” literature by way of application of general principles like cost-benefit analysis and pattern detection. Boyd talked about different levels of explanation, from the details of a particular work to general principles applying to human nature and derived from social evolution. The idea is that evolution bears on the broader levels of explanation, while attention to specific cultural-historical and even psychological conditions become more relevant at narrower levels of explanation.