Book Read Free

Complexity and the Economy

Page 30

by W Brian Arthur


  problem: it’s a problem in spherical trigonometry. I then associate the prob-

  lem with this framework. From there I associate structures and symbols with

  the sense data of the problem. And I proceed by such associations, stitching

  them together into a pattern. I’m not saying that association is all the human brain does, but cognitively, association is the main thing we do. And we do

  it fast. Our neural system searches fast over many associations before set-

  tling on one as a “meaning.” Occasionally this process slows and we can see

  it in action, as with the three-dimensional optical-illusion pictures that were popular a few years ago that appear flat and two-dimensional until after star-ing for half a minute a 3-D picture “leaps out.” So our brains process a large collection of associations into patterns—and a large set of metaphors which

  are merely more complicated associations with entailments. With metaphors

  we compare this to this and that to that, and if the comparison is good, we

  expect such and such to follow. Metaphor is a form of pattern association,

  and we process much information through metaphors. In sum, we have many

  different forms of associations: pictures; memories; metaphors; and theories,

  which are really elaborated metaphors. And this collection when it’s fully operating, along with the rules for combining these (which are also associations), we call the mind.

  Our minds then are extremely good at associating things, using metaphors,

  memories, structures, patterns, theories. In other words, the mind is not

  given. It’s not an empty bucket for pouring data in. The mind itself is emer-

  gent. This idea is new in Western thinking but there’s plenty of precedent for it in the East. The Neo-Confucian philosopher brothers Ch’eng Yi and Ch’eng

  Hao, writing during the Sung Dynasty about 900 years ago, both saw mind

  as emergent. They did not see the mind as a container, but rather as sets of

  ideas built one upon the other. The mind doesn’t contain our ideas. It’s these ideas—these associations—that instead contain the mind or constitute the

  mind. The mind is not fixed in any way; it consists in its associations and the cogni t ion [ 163 ]

  apparatus to manipulate these. In this sense it’s emergent. So strictly speaking I shouldn’t say as I did earlier that meaning resides in the mind, because deep enough within cognitive philosophy the concept of mind itself dissolves.

  Meaning resides in associations which our neural apparatus connects with the

  data presented. We are now far from seeing reasoning as deduction that takes

  place in a container of variables whose values are updated by “information.” If reasoning is largely association, it depends on the past experiences of the reasoner. The framing of a situation, the “sense” made of it, are therefore dependent on the reasoner’s history. And so is the outcome.

  One final point about cognition. Sometimes we can say roughly that there

  is a “correct” meaning—a single, correct association. More often, in any situ-

  ation of complication, there are multiple interpretations. We may hold one

  or we may hold many. Often, if we are trying to solve a puzzle, or to come to

  a decision such as the next move in a chess game, we make many hypotheti-

  cal associations and search over these, perhaps retaining more than one until

  further evidence presents itself. In the black tail example, if I had indeed seen a small boy a few minutes earlier, I might hold in mind both “cat” and “prank”

  until further evidence arrived.

  MODELING THE COGNITIVE PROCESS

  All this is fine. But as economists how do we make use of it? How might we

  model the thinking process in problems that are complicated or ill-defined?

  I would suggest the following, by way of distillation of the observations

  above: in problems of complication, as decision makers, economic agents look

  for ways to frame the situation that faces them. They try to associate tempo-

  rary internal models or patterns or hypotheses to frame the situation. And

  they work with these. They may single out one such pattern or model and

  carry out simplified deductions (at the level of tic-tac-toe) on it, if they seek guidance for action. As further evidence from the environment comes in, they

  may strengthen or weaken their beliefs in their current models or hypotheses.

  They may also discard some when they cease to perform, and replace them as

  needed with new ones. In other words, where agents face problems of compli-

  cation or ill-definition, they use clues from the situation to form hypothetical patterns, frameworks and associations. These hypothetical patterns fill the

  gaps in the agent’s understanding.

  Such a procedure enables us as humans to deal with complication: we con-

  struct plausible, simpler models that we can cope with. It enables us to deal

  with ill-definedness: where we have insufficient definition, we use working

  models to fill the gap. Such behavior is inductive. It may appear ad hoc and

  messy, but it is not antithetical to “reason,” or to science for that matter. In fact, it is the way science itself operates and progresses.

  [ 164 ] Complexity and the Economy

  More practically then, in a typical economics problem that plays out over time, we might set up a collection of agents, probably heterogeneous, and

  assume they make associations in the form of mental models, or hypotheses,

  or subjective beliefs. These beliefs might themselves take the form of simple

  mathematical expressions that can be used to describe or predict some vari-

  able or action; or of statistical hypotheses; or of condition/prediction rules (if situation Q is observed/predict outcome or action D). These will normally be subjective—they will differ among the agents. An agent may hold one

  in mind at a time, or several simultaneously, keeping track of the perfor-

  mance of each. When the time comes to make choices, the agent acts upon

  his currently most credible (or possibly most profitable) one. The others he

  keeps at the back of his mind, so to speak. As economists we will be tempted

  to say the agent rationally combines his several hypotheses. But cognitive

  psychology tells us we don’t do this, we hold in mind many hypotheses at a

  time and act on the one currently most plausible. Once actions are taken the

  aggregative picture is updated, and agents update their confidence in each

  of their hypotheses.

  This scheme I’m suggesting is of course also a simplification and abstrac-

  tion. But it captures the idea that the agent is imposing meaning on the problem situation, or making sense of it by associating multiple frameworks, or

  belief structures, or hypotheses with it and allowing these to “compete.” This is also a system in which learning takes place. Agents “learn” which of their

  hypotheses work, and they “learn” also in the acts of discarding poorly per-

  forming hypotheses and generating new “ideas” to put in their place. Notice

  there is a built-in hysteresis: agents linger with their currently most believable hypothesis or belief model, but drop it when it no longer performs, in favor

  of a better one. A hypothesis or association or belief model is clung to not

  because it is “correct”—there is no way to know this—but rather because it

  has worked in the past and must cumulate a record of failure before it is worth discarding.

  A key question remains. Where do the hypotheses or mental models come

  from? How are they generated? Behaviorally, this
is a deep question in psy-

  chology, having to do with object representation, and pattern recognition.

  I will not go into it here. But there are some simple and practical options for modeling. Sometimes we might endow our agents with focal models—patterns or hypotheses that are obvious, simple and easily dealt with mentally.

  We might generate a “bank” of these and distribute them among the agents.

  Other times, given a suitable model-space, we might allow some similar intel-

  ligent search device such as the genetic algorithm to generate suitable mod-

  els. The reader should note that whatever option is taken, the framework I’ve

  described is independent of the specific hypotheses or beliefs used, just as

  the consumer theory framework is independent of particular products chosen

  among.

  cogni t ion [ 165 ]

  Can such a scheme be put in practice in economics? The answer is yes.

  There is now a growing body of examples: the work of Sargent (1993); the El

  Farol problem (Arthur 1994); the Santa Fe stock market study (Arthur et al.

  1997). This type of study typically finds that “solutions”—patterns of beliefs and actions predicated upon these—need to be generated by computation

  because of the increased complication of heterogeneous beliefs. It also typi-

  cally finds a richer world, a psychological world, where an ecology of beliefs about the problem in question emerges. Sometimes this ecology of hypotheses converges to some standard equilibrium of beliefs. More often it remains

  open-ended, always discovering new hypotheses, new ideas.

  COGNITION AND GRADUATE ECONOMIC EDUCATION

  Let me turn from modeling in economics to quite a different area that can

  benefit from the insights of cognitive science: the education of economists.

  I want to start here by drawing attention to two ways in which we make

  sense: two types of association, not completely different and at opposite ends of a spectrum. Let me call one “theory” and the other “experience.”

  Theories are metaphors with entailments. If in 1705, Edmond Halley sub-

  scribed to Newton’s gravitational theory and applied it to a comet that had

  previously appeared in 1531, 1607 and 1682, one entailment was that the

  comet would return in the year 1759. In using Newton’s theory, Halley was

  making an association between a comet and the heavenly bodies Newton dealt

  with; and the entailments of the association allowed Halley to predict. I want to suggest that theories are thin associations: the theory fits if a narrow and precise set of conditions is fulfilled; and the entailments are also narrow and precise. Providing the theory fits correctly—is a good association—and is

  consistent within itself, then the entailments can be relied upon. Narrow fit, narrow entailments. Theories are in this way thin but powerful associations.

  Experience is different. Suppose I’m an executive sent to Korea, and I’ve

  never done business there. I arrive in Korea and I’m wondering how I shall act.

  I have no idea of how many times I should bow to my host, or if anybody bows

  to the host, or whether I should take my shoes off, or if I want to close the

  deal do I wait till the end of dinner or do I try to close the deal up front? But I do have a lot of experience in Japan and in China, and so I use these. In this case hundreds of pictures are going through my mind. This sort of association

  is more dream-like. It’s richer. It covers a wider set of cases. It is suggestive of what will follow given what is. But it’s much less accurate and less precise and less reliable than theory. So experience in the form of a wide collection of memories and pictures of situations— thick association—is also powerful. Its power lies in its width of coverage and its suggestiveness. Such experience is what we seek from human conversation and from taking in stories and novels

  [ 166 ] Complexity and the Economy

  and plays. We seek to draw into ourselves other people’s experiences, to make their situations into our memory pictures that we can use later. In this way

  we construct and conjure a whole dream-like world where logic doesn’t mat-

  ter and precision doesn’t matter, but where suggestiveness and coverage give

  power.

  As I said earlier, these two types of association are not completely distinct; associations arrange themselves on a spectrum from narrowness and precision to width and suggestiveness.

  What has this to do with graduate economic education? A lot. A great deal

  of education is the formation of associations; and the spectrum ranges from

  collections of narrow but precise theories on one side to wide but suggestive

  and imprecise pictures on the other. We need both types of association to

  function successfully as human beings.

  In economics, graduate education at least in the first year or two consists

  in mastering 20 or 30 theoretical economic models—thin associations. These

  include the principal–agent model, the overlapping generations model, the

  prisoner’s dilemma model, and so on. The idea is that these theoretical meta-

  phors will later become useful associations. We hope that if the student is

  later employed say at the World Bank, she will be able to look at a situation

  and say, “This problem in African agriculture is partly a principal–agent problem. It does have some overtones of overlapping generations, and it’s also got this game theory component. So I can put together a hybrid version of the

  three models to get insight.” All this is fine. It is fine that economics has recognized recurring structures that it has rendered into theories. We can hope and expect that a well-educated student will use these as association components

  later.

  But models cannot be all that we teach. There’s been a tendency in many

  graduate schools to increase teaching in theory at the expense of teaching

  in economic history and in case example. Students of course can still choose

  to study the experience-details of the economy; but they are aware that this

  may not enhance their graduate careers. In 1990 Colander and Klamer asked

  students how important having a “thorough knowledge of the economy” was

  to succeeding as an economist. Three percent thought it very important, and

  68% thought it unimportant. Important was: “Being smart in the sense of

  being good at problem solving,” and “excellence in mathematics.” With this

  bias toward theory and away from experience, we eliminate the wider meta-

  phors that come from history-experience—the thick associations. These allow

  students to put their models into perspective; they provide the vocabulary,

  so to speak, where theory provides the grammar; they provide a richness of

  thought and a breadth of association that theory cannot possibly match.

  When a decision maker faces a situation of high complexity, say Bosnia

  in the mid-1990s, applying theory prematurely—a set of precise but nar-

  rowly applicable metaphors—can be dangerous. Let’s say he is in the State

  cogni t ion [ 167 ]

  Department looking at Bosnia and has been in graduate school in political science, doesn’t have much experience and is full of theories. His reaction may be to shoehorn Bosnia into a pre-constructed framework. But in this situation

  it is better to wait and observe. And in observation to invoke a variable set

  of pictures on which he may conjure up a richer set of associations. Such free association comes from a study of history, not theory. “Well, it could be a bit like the Bosnian crisis of 1908, but it’s not unlike the situation under Turkish rul
e in 1831 when Husein seized power. On the other hand there are elements of the ethnic rivalries of 1875 that resulted in the Austro-Hungarians

  taking over.” What’s of use is to have thousands of such pictures from his-

  tory, available for pondering and perusal. Eventually from such pondering

  and perusal—from dreamlike association—a composite set of hypotheses

  or composite picture may emerge. It’s at this stage that theory might apply.

  Premature association without going through the richness of a wide set of

  pictures may be disastrous. Where I come from, Belfast—another complicated

  situation—we say: “If you’re not confused, you don’t know anything.”

  I am saying here that students need experience—details as well as theory.

  That is, they need economic history, not as an adjunct to theory, but as a supplier of cognitive understanding in its own right. What about teaching the

  history of economic thought—another threatened discipline in economic

  education? From the cognitive point of view, the history of economic thought

  bestows on us an awareness of the associations we make. Without such aware-

  ness, associations can be unconscious and poorly suited to the case in hand.

  Consider the English painters who came to Australia in the late 1700s or early 1800s. These artists depicted trees in Australia as they would have depicted

  English trees; they were well trained in English art schools and knew how

  to paint trees. But in Australia the leaves of most trees—often eucalyptus

  trees—are thinner, and the sun shines through them. Trees there look dif-

  ferent—lighter, more airy. It took a generation of Australian-born painters

  before the trees in paintings started to look like Australian trees. Before that European painters were unconsciously making European associations and

  imposing these upon Australia. Similarly Europeans depicted aboriginals in

  this early period as Europeans with dark skins. This is not to criticize artists.

  It is to be aware that the actions we take are built upon our unconscious associations. We need to be conscious of our associations and where they come

  from. We need to be suspicious of them. We need a Zen-like standing back

  and seeing from beginner’s mind. We need an awareness that theories aren’t

 

‹ Prev