Prelude to Chapter 10
GENERATING THOUGHTLESS CONFORMITY
TO NORMS
LET ME TRY to get at something big with something small. Here are two small experiments: (1) Stand four inches too close to someone when conversing. (2) The next time someone at a conference table makes a clever academic quip, try laughing heartily for ten full seconds after the obligatory polite chuckles have subsided. In both cases, you'll feel uncomfortable. Why? Because we have a very finely tuned sense of what's inside a norm and what's outside it. But we don't think about that boundary in any conscious way; not, that is, until it is crossed.
We don't think about how to walk normally, but we can detect remarkably slight departures from a normal human gait, or a normal human face. We don't think about how to talk normally, but we'd notice immediately if someone put a one-second silence after every third word. That which is normal is maintained as a kind of homeostatic target, and we only notice when the observed value wanders outside the acceptable range, just like a thermostat. Perhaps we are endowed with something like a normostat.
Societies may differ in the normostat's target value (what constitutes normalcy), and a given social group can change its value (as when fashions change) endogenously. But once a norm is entrenched—once the normostat is set to a value—we don't think about it, until a violation is observed. That is at once understandable and, arguably, adaptive. Clearly, if we spent the entire day consciously deciding whether or not to conform to every norm of locution, dress, gesture, hygiene, social comportment, and so on, we'd have no time left to accomplish anything. Thoughtless conformity to entrenched norms has an adaptive computation-saving aspect. But it is also very disturbing that we can just as thoughtlessly adopt discriminatory norms: racial biases, xenophobias, and so on.
A great deal of social science is concerned with conscious choices and decisions. Homo economicus collects information, arrays alternatives, and makes conscious strategic choices. He's thinking. To me, however, one of the more remarkable aspects of homo sapiens is how much of our social behavior is entirely thoughtless. Indeed, as argued in the Generative chapter, the major patternings in the fabric of humankind are of this sort: We didn't decide to identify with our particular ethnic group; we didn't consider the alternatives and select our particular native tongue; we didn't lay out a range of religions and pick orthodox Judaism; we didn't choose, in any meaningful sense, to prefer ragas to sonatas, or pancakes to monkey brain for breakfast.
Not all of the above examples are equally momentous. And none of them is specifically modeled in the next chapter. Rather, I try to get at the general phenomenon of thoughtless conformity in a simple two-norm setting.
As emphasized in the Generative chapter, space and heterogeneity are typical of agent models, and they are important here. Agents are arranged on a one-dimensional ring. Although they differ by norm, they are most heterogeneous by their search radius—that is, by how much they are thinking about what norm to adopt. In regions of the space where there is norm variation (i.e., on norm boundaries), agents are worrying about how to behave. In zones where one norm or another is monolithic (i.e., norm interiors), there is no thinking. Norm innovators (represented as noise) can jog the thoughtless into worrying, a process that can eventuate in local “tipping” from one norm to another. The dynamics can then exhibit what Young has identified as hallmarks of complex systems: local conformity (norm patches), global diversity, and punctuated equilibria.1
Best Reply to Adaptive Sample Evidence
The essential point of the model is to generate thoughtlessness itself. Hence, the number of sites an agent samples in figuring out how to behave is adaptive. It increases in the face of variation (observed norm violations), and, most importantly, it shrinks—in principle, to zero—once a norm is entrenched. Agents play a coordination game with others in this adaptive search radius. Hence, I have dubbed the learning dynamic Best Reply to Adaptive Sample Evidence. Now the attentive reader will notice that this departs from precisely the scheme used in the earlier Classes model (chapter 8). There, agents played Best Reply to Sample Evidence, but the sample size was fixed, not adaptive. In a sense, then, this chapter criticizes that one. So be it. As Emerson said, “A foolish consistency is the hobgoblin of little minds.”2 Best Reply misses something important in cases where thoughtless conformity (not merely conformity) is the “emergent” phenomenon of interest.
Unobservables
The chapter introduces a graphical technique also exploited in the subsequent chapter's Civil Violence models. Two screens are used. On the left is the screen of observable action: one's norm (e.g., left-hand driving) in the present model; one's rebellious state (active or not) in the next. On the right screen is an unobservable: one's search radius (how much one is thinking) in the present model; one's level of political grievance in the next. Unobservable theoretical entities have a long and venerable history in science, and I do not shrink from using them in this and other chapters. Contrary to popular inductivist myth, moreover, it often occurs in science that theory precedes observation. Electromagnetic theory predicted the existence of radio waves, which were only later observed. Further examples abound.
Without theory, it is not obvious what data to collect. In any case, when deep in a norm, the adaptive agents you are about to meet (and, in my view, many humans) collect no data at all!
1 H. P. Young, Individual Strategy and Social Structure: An Evolutionary Theory of Institutions (Princeton: Princeton University Press, 1998).
2 Ralph Waldo Emerson, “Self-Reliance,” in Essays by Ralph Waldo Emerson: First Series (New York: Houghton, Mifflin, 1904), 57.