Reason in Human Affairs

The Harry Camp Lectures at

Stanford University, 1982

The Harry Camp Memorial Fund

was established in 1959 to make
possible a continuing series of
lecrures at Stanford University on
topics bearing on the dignity and
worth of the human individual.
HERBERT A. SIMON

•
Reason in Human . . . . . . . airs

STANFORD UNIVERSITY PRESS 1983

STANFORD, CALIFORNIA
 Stanford Universirv• Press To the Memory of
Stanford, California
© r983 by the Board of Trustees of the JAS C HA MA R SCHAK
. Stanford Junior Universitv
Pnnted m the Unired States of who had an unshakeable faith in
I SBN o-80+7-11?9 -8 human reason, and an unmatched
LC 82-62448 store of human warmth
 Preface

T H E NATURE of human reason-its mechanisms, its

effects, and its consequences for the human condition-
has been my central preoccupation for nearly fifty years.
When the invitation came to deliver the Harry Camp Lec-
rures at Stanford University, I \vondered if I had anything
left to say on the subject . And if there were some such
topic,.had it not already been tho roughly investigated by
such friends on the Stanford campus as Kenneth Arrow,
'
James March, and Amos Tversky-to mention just a fe\v
who work in one part or another of this domain? Putting
aside this concern, real though it is, I decided to use
the occasion of the lectures to explore some byways that
seemed to me interesting 'a nd important, but that had until
now been off the main paths of my own explorations.
Three topics, especially, were the objects of my inquiry:
the relation of reason to intuition and emotion, the anal-
ogy between rational adaptation and evolution, and the
implications of bounded rationality for the operation of
social and political institutions. In the chapters that follow,
I report on these topics \Vithin the framework provided by
the general viewpoint of bounded rationality.
I am indebted to Stanford University for the occasion
••
Vll
PREFACE

and opportunity to prepare these pages, and for the hospi- Contents
tality and stimulation I always enjoy on my visits to the
Stanford campus. I am grateful, too, to Donald T. Camp-
Richard C. Lewontin, and Edv.rard 0. Wilson, who
provided valuable criticisms of a draft of Chapter 2,
though it is not to be assumed that they would agree with
everything in the final version of that chapter. To them,
and to many friends who have helped guide my education
on evolutionary theory and on other topics addressed in
3
these I offer my warm thanks. I. Alternative Visions of Rationality

H.A.S. 37
2. Rationality. and Teleology

. Rational Processes in Social Affairs 75

3

III
Index

'

.'.
Vlll
 Reason in H11man Affairs

I
 1. Alternative Visions of
Rationality

Q N E KIND of optimism, or supposed optimism, ar-

gues that ifwe think hard enough, are rational enough, we
• can solve all our problems. The eighte_enth century, the
Age of Reason, was supposed to have been imbued with
this kind of optimisn1. Whether it actually was or not I will
leave to historians; certainly the h?pes v-ve hold out for
reason in our world today are much more modest.
It is my purpose in these pages to explore from a con-
temporary standpoint the uses and limits of reason in hu-
man affairs. In order to avoid the kind of unwarranted
optimism I have just mentioned, my first two chapters will
be addressed n1ore to the limitations of reason than to its
uses. I will try to redress the balance in my third chapter,
but as I develop my topic I think you will see why I have
taken up the limitations first. Only as we understand those
limitations can we devise procedures to use effectively the
powers that human reasoning capabilities do give us.
In the first chapter, I will focus initially _on the very
powerful formal models of rationality that have been con-
structed in this century and that must be counted among
;
the jewels of intellectual accomplishment in our time.
Since these models are ·well known, I will describe them

3
 VIS IONS OF RATIONALITY
VISIONS OF RATIONALITY
things that I have said at length in my previous books, and
most of my discussion to showing especially in Administrative Behavior and The ofthe
Wh), In apphcatlon to real human affairs thev deliver Artificial, both of which are deeply concerned \vtth the
less .than they appear to promise. But intent concept of human rationality. In the former, I examined
here IS not mainly The .last half of the chapter will the implications of the limits of human rationality for or-
de:elop. a more realistiC descnption of human bounded ganizational behavior. In the latter, I described
and will consider to what extent the limited that are common to all adaptive ("artificial") systems, g1V-
cap.ablhty for analysis that is provided by bounded ratio- ing us a basis for constructing a general theory of such
nahty can meet the needs for reason in human affairs. systems. In the present volume I have drawn on this pre-
In the second chapter' I will discuss the thesis a- vious work to the extent necessary to provide a framework
days often with the discipline of sociobiology, for my discussion. But within this framework I have con-
that the deficiencies of reason will be corrected, for better centrated on topics that remain problematic or controver-
or for by the sterner rationality of natural selection. sial and that are of critical importance for understanding
quesnons will be of particular concern in that discus- the role of rationality in human affairs. I have already in-
Sl.on :. first, and to what degree altruism can sur- dicated briefly what some of those problematic topics are.
VIVe m a system subjected to the forces ofnatural selection
and.' to what extent selection processes THE LIMITS OF REASON
optmuzatlon processes. Modern descendants of Archimedes are still looking for
In the light of the conclusions reached in these two the fulcrum on which thev can rest the lever that is to move
chapters, I will turn in the third to the question of how -
the whole world. In the domain of reasoning, the difficulty
reason be employed effectively in hwnan social affairs. in finding a fulcrum resides in the truism "no conclusions
In SCienc.e one is. supposed to deliver new truths. The without premises." Reasoning processes take symbolic in-
crushing verdiCt that can be pronounced on a scien- puts and deliver symbolic outputs. The. initial
tific is the fabled referee's report, scribbled in the axioms, themselves not derived by logtc but slffiply ln-
margtn, ,;"hat's ne\v here is not true, and what's true is duced from empirical observations, even more simply
the.se pages are not intended as reports of posited. Moreover, the processes that produce the trans-
discovenes and will not seek novelty. I will be formations of inputs to outputs (rules of inference) are
sa?sfied If what I have to say is mainly true, even though it also introduced by fiat and are not the products of reason.
wtll not at new. As I shall argue in my discussion of Axioms and inference rules together constitute the ful-
rattonabty, attention needs to be called periodically crum on which the lever of reasoning rests; but the partic-
to Important old truths. ular structure of that fulcrum cannot be justified by the
At the same time, I do not wish simply to repeat here
5
\
4
VISIONS OF RATIONALITY VISIONS OF RATIONALITY
'

methods of reasoning. For an attempt at such a justifica- are capable of generating normative outputs purely from
tion would involve us in an infinite regress of logics, each descriptive inputs. 1 The corollary to "no conclusions vvith-
as arbitrary in its foundations as the preceding one. out premises" is "no ought's from is's alone." Thus,
This ineradicable element of arbitrariness-this Origi- as reason may provide povverful help in finding means to
nal Sin that corrupts the reasoning process, and therefore reach our ends, it has little to say about the ends themselves.
also its products-has two important consequences for There is a final difficulty, first pointed out by Godel, that
our topic here. First, it puts forever beyond reach an unas- rich systems of logic are never complete-there always
sailable principle of induction that would allow us to infer exist true theorems that cannot be reached as outputs by
infallible general laws, without risk of error, from specific applying the legal transformations to the inputs. Since the
facts, even from myriads of them. No number of viewings problem of logical incompleteness is much less important
of white swans can guarantee that a black one will not be in the application of reason to human affairs than the
seen next. Whether even a definite probability statement difficulties that concern us here, I shall not discuss it
can be made about the color of the next swan is a matter ther. Nor will I be concerned with whether the standard
of debate, with the negatives, I think, outnumbering the axioms oflogic and the rules ofinference themselves are to
affirmatives. some extent arbitrary. For the purpose of this discussion, I
Further, the foundations of these inductions-the shall regard them as unexceptionable.
facts-rest on a complex and sometimes unsteady base of Reason, then, goes to work only after it has been sup-
observation, perception, and inference. Facts, especially in plied with a suitable set of inputs, or premises. If reason
science, are usually gathered in with instruments that are is to be applied to discovering and choosing courses of
themselves permeated with theoretical assumptions. No action, then those inputs include, at the least, a set of
microscope \Vithout at least a primitive theory of light and should's, or values to be achieved, and a set of is's, or facts
optics; no human verbal protocols without a theory of about the world in which the action is to be taken. Any
short-term memory. Hence the fallibility of reasoning is attempt to justify these should's and is's by logic will simply
guaranteed both by the impossibility of generating un- lead to a regress to new should's and is's that are similarly
assailable general propositions from particular facts, and postulated.
by the tentative and theory-infected character of the facts
VALUES
themselves.
Second, the principle of "no conclusions without prem- We see that reason is wholly instrumental. It cannot tell
ises" puts forever beyond reach normative statements us where to go; at best it can tell us how to get there. It is a
(statements containing an essential should) \vhose deriva- 1I will not undertake to make the argument here. It was stated well many
tion is independent of inputs that also contain should's. years ago by Ayer, in Language, Truth, and Logic, re\'. ed. (New York, 1946),
None of the rules of inference that have gained acceptance chap. 6.

6 7
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

gun for hire that can be employed in the service of what- was to combat its program by reason resting on better
ever goals \Ve have, good or bad. It makes a great dif- facmal premises.
ference in our view of the human condition \vhether .we But somehow that calm response does not seem to
attribute our difficulties to evil or to ignorance and irra- match the outrage thatMein J(ampfproduces in us. There
tionality- to the baseness of goals or to our not knowing must be something more to our rejection of its argument,
how to reach them. and obviously there is. Its stated goals are, to put it mildly,
incomplete. Statements of human goals usually
Method in Madness guish between a "we" for w hom the goals are shaped and
A useful, if outrageous, exercise for sharpening one's a "they" whose welfare is not "our" primary concern.
thinking about the limited usefulness of reasoning, taken Hitler's ''we" \vas the German people-the definition of
in isolation, is to attempt to read Hitler's M ein ](ampf "we" being again based on some dubious "facts" about a
analytically-as though preparing for a debate. The exer- genetic difference between Aryan peoples.
cise is likely to be painful, but is revealing about how facts, Leaving aside this fantasy of Nord1c most
values, and emotions interact in our thinking about hu- would still define "we" differently from H 1tler. Our 'we
man affairs. I pick this particular example because the might be Americans instead of Germans, or, if \Ve h ad
reader's critical faculties are unlikely, in this case, to be reached a twenty-first-century state of enlightenment, our
dulled by agreement with the views expressed. "we" might even be the human species. In .case, w_e
Most of us would take exception to many of Hitler's would be involved in a genuine value conflict w1th Metn
"facts," especially his analysis of the causes of Europe's J(ampf, a conflict not resolvable in any obvious way by
economic difficulties, and most of all his allegations that improvements in either facts or reasoning. Our posmla-
Jews and Marxists (whom he also mistakenly found indis- tion of a ''we" -of the boundary of our concern for oth-
tinguishable) were at the root of them . H owever, if we ers-is a basic assumption about what is good and what
were to suspend disbelief for a moment and accept his
is evil.
"facts'' as true, much of the Nazi program would be·quite Probably the greatest sense of outrage that M ein
consistent with goals of security for the German nation or generates stems from the sharpness of the boundary Hnler
even of welfare for the German people. Up to this point, draws between "we" and "they." Not only does he give
the unacceptability of that program to us is not a matter of priority to "we," but he argues that .an! treatment of
evil goals-no one would object to concern for the welfare "they," however violent, is justifiable If 1t advances the
of the German people-or of faulty reasoning from those goals of "we." Even if Hitler's g?als and "facts"
goals, but rests on the unacceptability of the factual postu- were accepted, most of us would st11l obJect to the m ea-
lates that connect the goals to the program. From this sures he proposes to inflict on "they" in o rder to nurture
viewpoint, we might decide that the remedy for Nazism the welfare of ''we." If, in our system of values, we do not
8 9
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

regard "they'' as being without rights, reason will disclose cipal shield against Nazism. Our principal shield was con-
to us a conflict of values a conflict between our value of trary factual beliefs and values.
helping "we" and our general goal of not inflicting harm
on "they." And so it is not its reasoning for which \Ve must De Gustibus Est Disputandum
fault Mein l(ampf, but its alleged facts and its outrageous Recognizing all these complications in the use of rea-
values. . . son, hot or cold, and recognizing also that ought's cannot
There is another lesson to be learned from Mein !(ampf. be derived from is's alone, we must still admit that it is
We cannot read many lines of it before detecting that possible to reason about conduct. For most of the ought's
Hitler's reasoning is not cold reasoning but hot reasoning. we profess are not ultimate standards of conduct but only
have long since learned that when a position is de- subgoals, adopted as means to other goals. For example,
claimed with passion and invective, there is special need to taken in isolation a goal like "live within your income"
examine carefully both its premises and its inferences. We may sound unassailable. Yet a student might be well ad-
have but we do not always practice it. Regret- vised to go into debt in order to complete his or her
tably, 1t ts prectsely when the passion and invective reso- education. A debt incurred as an investment in future
with our own inner feelings that we forget the warn- productivity is different from a gambling debt.
tog and become uncritical readers or listeners . Values can indeed be disputed (r) if satisfying them has
. Hitler was an effective rhetorician for Germans pre- consequences, present or future, for other values, (2) if
ctsely because his passion and invectives resonated with they are acquired values, or (3) if they are instrumental to
beliefs and values already present in many German hearts. more final values. But although there has been widespread
The of his rhetoric rendered his readers incapable of consensus about the rules of reasoning that apply to fac-
applymg the rules of reason and evidence to his argu- tual matters, it has proved far more difficult over the cen-
ments. Nor was it only Germans who resonated to the turies to reach agreement about the rules that should
facts and values he proclaimed. The latent anti-Semitism govern reasoning about interrelated values. Several vari-
and overt anti-Communism of many Western statesmen eties of modal logic proposed for reasoning about impera-
made a number of his arguments plausible to them. tive and deontic statements have gained little acceptance
2
And so we learned, by bitter experience and against our and even less application outside of philosophy.
first quick judgments, that we could not dismiss Hitler as a In the past half century, however, an impressive body of
madman, for there was method in his madness. His prose formal theory has been erected by mathematical statisti-
met standards of reason neither higher nor lower than we 1 state the case against modal logics in Section 3 of my Models ofDiscomy
2

are accustomed to encountering in writing designed to (Dordrecht, 1977) and in "On Reasoning about Actions," chap. 8 of H. A. Si-
mon and L. Sik16ssy, eds., Representation and Meaning (Englewood Cliffs, N.J.,
persuade. Reason was not, could not have been, our prin- 1972).

ro II
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

cians and economists to help us reason about these mat- Third, it assumes that the decision maker can assign a
ters- without introducing a new kind of logic. The basic consistent Joint probability distribution to all furore sets of
idea of this theory is to load all values into a single func- events. Finally, it assumes that the decision maker will (or
tion, the utility function, in this way finessing the question should) choose the alternative, or the strategy, that will
of how different values are to be compared. The compari- maximize the expected value, in terms of his utility function,
son has in effect already been made when it is assumed that of the set of events consequent on the strategy. With each
a utility has been assigned to each particular state of affairs. strategy, then, is associated a probability distribution of
This formal theory is called subjective expected utility future scenarios that can be used to weight the utilities of
(SEU) theory. Its construction is one of the impressive those scenarios.
intellectual achievements of the first half of the twentieth These are the four principal components of the SEU
century. It is an elegant machine for applying reason to model: a cardinal utility function, an exhaustive set of
problems of choice. Our next task is to examine it, and to alternative strategies, a probability distribution of sce-
make some judgments about its validity and limitations. narios for the future associated with each strategy, and a
policy of maximizing expected utility.
SUBJECTIVE EXPECTED UTILITY

Since a number of comprehensive and rigorous ac- Problems with the Theory
counts of SEU theory are available in the literature, 3 I \Vill the SEU model is a beautiful object de-
give here only a brief heuristic survey of its main com- serving a prominent place in Plato's heaven of ideas. But
ponents. vast difficulties make it impossible to employ it in any
literal way in making actual human decisions. I have said
The Theory so much about these difficulties at other times and places
First, the theory assumes that a decision maker has a (particularly in the pages ofAdministrative Behavior) that I
well-defined utility function, and hence that he can assign a will make only the briefest mention of them here.
cardinal number as a measure ofhis liking of any particular The SEU model assumes that the decision maker con-
scenario of events over the future. Second, it assumes that templates, in one comprehensive view, everything
the decision maker is confronted with a well-defined set of lies before him. He understands the range of alternative
alternatives to choose from. These alternatives need not be choices open to him, not only at the moment but over the
one-time choices, but may involve sequences of choices or whole panorama of the future. He understands the conse-
strategies in which each subchoice will be made only at a quences of each of the available choice strategies, at least
specified time using the information available at that time. up to the point of being able to assign a joint probability
distribution to future states of the world. He has recon-
3
For example, L. J. SaYage's classic, The Fou11datio11S ofStatistics (Ne,,· York,
1954). ciled or balanced all his conflicting partial values and syn-
'
12 13
 VISIONS OF RATIONALITY VISIONS OF RATIONALITY

thesized them into a single utility function that orders, by production levels, inventories, and work force in a factory
his preference for them, all these future states of the ·world. under conditions of uncertainty. 4 The procedure fits the
The SEU model finesses completely the origins of the SEU model. The utility function is (the negative of) a cost
values that enter into the utility function; they are simply function, comprising costs of production, costs of chang-
there, already organized to express consistent preferences ing the level of production, putative costs of lost orders,
among all alternative futures that may be presented for and inventory holding costs. The utility function is as-
choice. The SEU model finesses just as completely the sumed to be quadratic in the independent variables, an
processes for ascertaining the facts of the present and fu- assumption made because it is absolutely essential if the
ture states of the world. At best, the model tells us how mathematics and computation are to be manageable. Ex-
to reason about fact and value premises; it says nothing pected values for sales in each future period are assumed to
about where they come from. be known. (The same assumption of the quadratic utility
When these assumptions are stated explicitly, it be- function fortunately makes knowledge of the complete
comes obvious that SEU theory has never been applied, probability distributions irrelevant.) The factory is as-
and never can be applied-with or without the largest sumed to have a single homogeneous product, or a set of
computers-in the real world. Yet one encounters many products that can legitimately be represented by a single-
purported applications in mathematical economics, statis- dimensional aggregate.
tics, and management science. Examined more closely, It is clear that if this decision procedure is used to make
these applications retain the formal structure of SEU the- decisions for a factory, that is very different from employ-
ory, but substitute for the incredible decision problem ing SEU theory to make decisions in the real world. All
postulated in that theory either a highly abstracted prob- but one of the hard questions have been ans\vered in ad-
lem in a world simplified to a few equations and variables, vance by the assumption of a known, quadratic criterion
with the utility function and the joint probability distri- function and known expected values of future sales. More-
butions of events assumed to be already provided, or a over, this single set of production decisions has been
microproblem referring to some tiny, carefully defined carved out of the entire array of decisions that manage-
and bounded situation carved out of a larger real-world ment has to make, and it has been assumed to be describ-
reality. able in a fashion that is completely independent of infor-
mation about those other decisions or about any other
SE U as an Approximation
aspect of the real world.
Since I have had occasion to use SEU theory in some of
I have no urge to apologize for our decision procedure
my own research in management science, let me throw the
stone through my own window. Holt, M odigliani, Muth, 4
C. C. Holt, F. Modigliani, J. R. M uth, and H. A. Simon, Planning Produc-
and I constructed a procedure for making decisions about tion, Inventories and Work Force (Englewood Cliffs, N.J., I96o). ,

14- IS
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

as a useful management science tool. It can be, and has situations, decision makers, no matter how badly they
been, applied to this practical decision task in a number of want to do so, simply cannot apply the SEU model. If
factory situations and seems to have operated satisfac- doubt still remains on this point, it can be dissipated by
torily. What I wish to emphasize is that it is applied to a examining the results of laboratory experiments in which
highly simplified representation of a tiny fragment of the human subjects have been asked to make decisions involv-
real-world situation, and that the goodness of the deci- ing risk and uncertainty in game-like situations orders of
sions it will produce depends much more on the adequacy magnitude simpler than the game of real life. The evi-
of the approximating assumptions and the data support- dence, much of which has been assembled in several arti-
ing them than it does on the computation of a maximizing cles by Amos Tversky and his colleagues, leaves no doubt
value according to the prescribed SEU decision rule. whatever that the human behavior in these choice
Hence, it would be perfectly conceivable for someone to tions-for whatever reasons-departs widely from the
contrive a quite different decision procedure, outside the prescriptions of SEU theory. 5 Of course, I have already
framework of SE U theory, that would produce better de- suggested what the principal reason is for this departure.
cisions in these situations (measured by real-world conse- It is that human beings have neither the facts nor the
quences) than would be produced by our decision rule. consistent structure of values nor the reasoning power at
Exactly the same comments can be made about eco- their disposal that would be required, even in these rela-
nomic models formed within the SEU mold. Their verid- tively siinple situations, to apply SEU principles.
icality and usefulness cannot be judged from the fact that As our next task, we consider what they do instead.
they satisfy, formally, the SED assumptions. In evaluating
them, it is critical to know how close the postulated util- THE BEHAVIORAL ALTERNATIVE

ities and future events match those of the real world. I will ask you to introspect a bit about how you actually
Once we accept the fact that, in any actual application, make decisions, and I will make some assertions that you·
the SED rule supplies only a crude approximation to an can check against your introspections. First, your deci-
abstraction, an outcome that may or may not provide sat- sions are not comprehensive choices over large areas of
isfactory solutions to the real-world problems, then we are your life, but are generally concerned with rather specific
free to ask what other decision procedures, W1related to matters, assumed, whether correctly or not, to be rela-
SEU, might also provide satisfactory outcomes. In partic- tively independent of other, perhaps equally important,
ular, we are free to ask what procedures human beings dimensions of life. At the moment you are buying a car,
actually use in their decision making and what relation you are probably not also simultaneously choosing next
those actual procedures bear to the SEU theory. 5See A. Tversky and D. Kahnemann, "Judgment under Uncertainty: H euris-
I hope I have persuaded you that, in typical real-world tics and Biases,'' Science 185: II24- 3I (1974-), and references there.

I6 17
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

week's dinner menu, or even deciding how to invest in- Fourth, a large part of whatever effort vou devote to
-
making your car-buying decision ·will be absorbed in gath-
come you plan to save.
. Second, when you make any particular decision, even an ering facts and evoking possibly relevant values. You may
Important one, you probably do not work out detailed read Consumer Reports and consult friends; you may visit
of the future, complete with probability distri- car dealers in order to learn more about the various alter-
butions, conditional on the alternative vou choose. You natives, and to learn more about your own tastes as ·well.
have a general picture of your life-style prospects, and Once facts of this sort have been assembled, and prefer-
perhaps of one or two major contemplated changes in the ences evoked, the actual choice may take very little time .
near future, and even of a couple of contingencies. When
Bounded Rationality
are considering buying a car, you have a general no-
Choices made in the general way I have just been describ-
tion of your use of automobiles, your income and the
ing are sometimes characterized as instances of bounded
demands on it, and whether you are thinking of
rationality. Good reasons can be given for supposing that
a new job in another city. You are unlikely to
evolutionary processes might produce creatures capable of
env1s1on large numbers of other possibilities that might
bounded rationality. Moreover, a great deal of psychologi-
affect what kind of car it makes sense to buy.
cal research supports the hunch to which our introspec-
Third, the very fact that you are thinking about buying a
tions have led us, namely that this is the way in which
car, and not a house, will probably focus your attention on
human decisions-even the most deliberate-are made.
some aspects of your life and some of your values to the
Let us call this model of human choice the behavioral
neglect ?f others. The mere contemplation ofbuy-
model, to contrast it with the Olympian model of SEU
mg a car may stimulate fond memories or dreams of travel
theory.
and divert the pleasures oflistening
Within the behavioral model of bounded rationality,
stereo or gtvtng dinner parties for friends at home. Hence
one doesn't have to make choices that are infinitely deep in
it _is unlikely that a single comprehensive utility
time, that encompass the whole range of human values,
Will watch over the whole range ofdecisions you make. On
and in which each problem is interconnected with all the
contrary, particular decision domains will evoke par-
other problems in the world. In actual fact, the
ticular values, and great inconsistencies in choice may re-
ment in which we live, in which all creatures live, is an
sult from fluctuating attention. We all know that if we
environment that is nearly factorable into separate prob-
to diet, we should resist exposing ourselves to tempt-
lems. Sometimes you're hungry, sometimes you're sleepy,
Ing food. That would be neither necessary nor useful if our
sometimes you're cold. Fortunately, you're not often all
choices were actually guided by a single comprehensive
three at the same time. Or if you are, all but one of these
and consistent utility function. '

19
r8
VIS IONS OF RATIONALITY
VISIONS OF RATIONALITY

needs can be postponed until the most pressing is taken some way of focusing attention of avoiding distraction
care of. You have lots of other needs, too, but these also do (or at least too much distraction) and focusing on the
not all impinge on you at once. things that need attention at a given time. A very strong
We live in what might be called a nearly empty \vorld- case can be made, and has been made by physiological
one in which there are millions of variables that in princi- psychologists, that focusing attention is one of the princi-
ple could affect each other but that most of the time don't. pal functions of the processes we call emotions. One thing
In gravitational theory everything is pulling at everything an emotion can do for and to you is to distract you from
else, but some things pull harder than others, either be- your current focus of thought, and to call your attention to
cause they're bigger or because they're closer. Perhaps something else that presumably needs attention right no\v.
there is actually a very dense network of interconnections Most of the time in our society we don't have to be out
in the world, but in most of the situations we face we can looking for food, but every so often we need to be re-
detect only a modest number of variables or considera- minded that food is necessary. So we possess some mecha-
tions that dominate. nisms that arouse periodically the feeling of hunger, to
Ifthis factorability is not wholly descriptive ofthe world direct our attention to the need for food. A similar account
we live in today and I will express some reservations can be given of other emotions.
about that it certainly describes the world in which hu- Some of an organism's requirements call for continuous
man rationality evolved: the world of the cavemen's ances- activity. People need to have air access to it can be inter-
tors, and of the cavemen themselves. In that world, very rupted only for a short time and their blood must circu-
little was happening most of the time, but periodically late continually to all parts of their bodies. Of course,
action had to be taken to deal with hunger, or to flee human physiology takes care of these and other short-
danger, or to secure protection against the coming winter. term insistent needs in parallel with the long-term needs.
Rationality could focus on dealing with one or a fe\v prob- We do not have to have our attention directed to a lack of
lems at a time, with the expectation that when other prob- oxygen in our bloodstream in order to take a breath, or for
lems arose there would be time to deal with those too. 6 our heart to beat. But by and large, with respect to those
needs that are intermittent, that aren't constantly with us,
Mechanisms for Bounded Rationality
we operate very much as serial, one-at-a-time, animals.
What characteristics does an organism need to enable it
One such need is about as many as our minds can handle at
to exercise a sensible kind of bounded reality? It needs
one time. Our ability to get a\vay with that limitation, and
6
A simple formal model of such rationality is provided by my "Rational to survive in spite of our seriality, depends on the mecha-
Choice and the Structure of the Environment,'' Psychological Review 63: 129-38 nisms, particularly emotional mechanisms, that assure new
(I9S6).
problems of high urgency a high priority on the agenda.
20
21
 VISIONS OF RATIONALITY
VISIONS OF RATIONALITY

Second, \Ve need a mechanism capable of generating thing is closely connected with everything else, in
alternatives. A large part of our problem solving consists problems can be decomposed into their In
such a world, the kind of rationality I've been descnb1ng
in the search for good alternatives, or for improvements in
alternatives that we already know. In the past 25 years, gets us by.
research in cognitive psychology and artificial intelligence Consequences ofBounded Rationality .
has taught us a lot about how alternatives are generated. I Rationality of the sort described by the behavioral
have given a description of some of the mechanisms in model doesn't optimize, of course. Nor does it even guar-
Chapters 3 and 4 of The Sciences of the Artificial. 7 antee that our decisions will be consistent. As a matter of
Third, we need a capability for acquiring facts about the fact, it is very easy to show that choices made by an organ-
environment in which we find ourselves, and a modest ism having these characteristics will often depend on the
capability for drawing inferences from these facts. Of order in which alternatives are presented. If A is presented
course, this capability is used to help generate alternatives before B, A may seem desirable or at least but
as well as to assess their probable consequences, enabling ifB is presented before A, B will seem desirable and w1ll be
the organism to maintain a very simple model of the part chosen before A is even considered.
of the world that is relevant to its current decisions, and to The behavioral model gives up many of the beautiful
do commonsense reasoning about the model. formal properties of the Olympian model, in
What can vve say for and about this behavioral version, for giving them up it provides a way of looking at
this bounded rationality version, of human thinking and nality that explains how creatures with our mental capaci-
problem solving? The first thing we can say is that there is ties or even, with our mental capacities supplemented
now a tremendous weight of evidence that this theory with all the computers in Silicon Valley-get along in a
describes the way people, in fact1 make decisions and solve world that is much too complicated to be understood
problems. The theory has an increasingly firm empirical from the Olympian viewpoint of SEU theory.
base as a description of human behavior. Second, it is a
theory• that accounts for the fact that creatures stay alive INTUI T IVE RATIONALITY
and even thrive, who-however smart they are or think A third model of human rationality has been much less
they are have modest computational abilities in com- discussed bv social scientists than the two that I've consid-
parison with the complexity of the entire world that sur- ered so far: but is perhaps even more prominent in the
rounds them. It explains how such creatures have survived popular imagination. I've referred to it as the intuitive
for at least the millions of years that our species has sur- model. The intuitive model postulates that a great deal of
vived. In a world that is nearly empty, in which not every- human thinking, and a great deal of the success of human
7
Second ed. (Cambridge, Mass., 1981). beings in arriving at correct decisions, is due to the fact

22 23
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

that they have good intuition or good judgment. The that any kind of complex thinking that involves taking
notions of intuition and judgment are particularly promi- in information, processing that information, and doing
nent in public discussion today because of the research of something \vith it employs both of our hemispheres in
Roger Sperry and others, much supplemented by specula- varying proportions and in various \Vavs. •

tion, on the specialization of the left and right hemi- Of course, brain localization is not the important is-
spheres of the human brain. sue at stake. Regardless of whether the same things or
different things go on in the two hemispheres, the impor·
The Two Sides of the Brain tant question is whether there are two radically different
In the minds and hands of some writers, the notion of forms of human thought-analytic thought and intui-
hemisphere specialization has been turned into a kind of tive thought-and whether \Vhat we call creativitv relies
'
romance. According to this romanticized account, there's largely on the latter.
the dull, pedestrian left side of the brain, which is very
analytic. It either, depending on your beliefs, does the Intuition and Recognition
Olympian kind of reasoning that I described first, or- if What is intuition all about? It is an observable fact that
it's just a poor man's left hemisphere-does the behavioral people sometimes reach solutions to problems suddenly.
kind of thinking I described as the second model. In either They then have an "aha!" experience of varying degrees of
case, it's a down-to-earth, pedestrian sort of hemisphere, intensity. There is no doubt of the genuineness of the
capable perhaps of deep analysis but not of flights of fancy. phenomenon. Moreover, the problem solutions people
Then there's the right hemisphere, in which is stored hu- reach \vhen they have these experiences, \vhen they make
man imagination, creativity all those good things that intuitive judgments, frequently are correct.
account for the abilities of human beings, if they would Good data are available on this point for chess masters.
entrust themselves to this hemisphere, to solve problems Show a chess position, from a mid-game situation in a
• •
1n a creanve way. reasonable game, to a master or grand master. After look-
Before I try to characterize intuition and creativity (they ing at it for only five or ten seconds, he will usually be able
are not always the same thing) in a positive way, I must to propose a strong move-very often the move that is
comment on the romantic view I have just caricatured. objectively best in the position. If he's playing the game
When we look for the empirical evidence for it, we find against a strong opponent, he won't make that move im-
that there is none. There is lots of evidence, of course, for mediately; he may sit for three minutes or half an hour in
specialization of the hemispheres, but none of that evi- order to decide whether or not his first intuition is really
dence really argues that any complex human mental func- correct. But perhaps 8o or 90 percent of the time, his first
tion is performed by either ofthe hemispheres alone under impulse will in fact show him the correct move.
normal circumstances. By and large, the evidence sho\vs The explanation for the chess master's sound intuitions

25
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

is well known to psychologists, and is not really surpris- acquaintances to them. The estimates come out, as an
ing. 8 It is no deeper than the explanation of your ability, in order of magnitude, around fifty thousand, roughly com-
a matter of seconds, to recognize one of your friends parable to vocabulary estimates for native speakers. Intui-
whom you meet on the path tomorrow as you are going to tion is the ability to recognize a friend and to retrieve from
class. Unless you are very deep in thought. as you walk, the memory all the things you've learned about the friend in
recognition will be immediate and reliable. Now in any the years that you've known him. And of course if you
field in which we have gained considerable experience, \Ve know a lot about the friend, you'll be able to make good
have acquired a large number of "friends"-a large num- judgments about him. Should you lend him money or
ber of stimuli that we can recognize immediately. We can not? Will you get it back if you do ? Ifyou know the friend
sort the stimulus in whatever sorting net performs this well, you can say "yes" or "no" intuitively.
function in the brain (the physiology of it is not under-
stood), and discriminate it from all the other stimuli we Acquiring Intuitions and Judgment
might encounter. We can do this not only with faces, but Why should we believe that the recognition mechanism
with words in our native language. explains most of the "aha!" experiences that have been
Almost every college-educated person can discriminate reported in the literature of creativity? An important rea-
among, and recall the meanings of, fifty to a hundred son is that valid experiences happen only to people
thousand different words. Someho\v, over the years, we who possess the appropriate knowledge. Poincare rightly
have all spent many hundreds of hours looking at words, said that inspiration comes only to the prepared mind.
and we have made friends \vith fiftv or a hundred thousand Today we even have some data that indicate how long
'
of them. Every professional entomologist has a compara- it takes to prepare a mind for world-class creative per-
ble ability to discriminate among the insects he sees, and formance.
every botanist among the plants. In any field of expertise, At first blush, it is not clear why it should take just as
possession of an elaborate discrimination net that permits long in one field as in another to reach a world-class level
recognition of any one of tens of thousands of different of performance. However, human quality of performance
objects or situations is one of the basic tools of the expert is evaluated by comparing it with the performance of
and the principal source of his intuitions. other human beings. Hence the length of human life is a
Counts have been made of the numbers of "friends" controlling parameter in the competition; we can spend a
that chess masters have: the numbers of different configu- substantial fraction of our lives, but no more, in increasing
rations of pieces on a chessboard that are old familiar our proficiency. For this reason, the time required to pre-
pare for world-class performance (by the people whose
For a survey of the evidence, see my Models of'Thought (New H aven, Conn.,
8
talents allow them to aspire to that level) should be
1979), chaps. 6.2 -6 .5.
roughly the same for different fields of activity.
26
27
VISIONS OF RATIONALITY
VISIONS OF RATIONALITY

Empirical data gathered by my colleague John R. Hayes have gained through our past searches. Hence we would
for chess masters and composers, and somewhat less sys- expect what in fact occurs, that the expert will often be able
tematically for painters and mathematicians, indicate that to proceed intuitively in attacking a problem that requires
ten years is the magic number. Almost no person in these painful search for the novice. And we would expect also
disciplines has produced world-class performances ·with- that in most problem situations combining aspects of nov-
out having first put in at least ten years of intensive learn- elty with familiar components, inruition and search will
ing and practice. cooperate in reaching solutions.
What about child prodigies? Mozart was composing
world-class music perhaps by the time he \Vas seventeen- INTUITION AND EM OT ION
certainly no earlier. (The standard Hayes used for music is Thus far in our discussion of intuitive processes we have
five or more appearances of recordings of a piece of music left aside one of the important characteristics these pro-
in the Schwann catalog. Except for some Mozart juvenilia, cesses are said to possess: their frequent association with
which no one would bother to listen to if thev hadn't been emotion. The searching, plodding stages of problem solv-
"
written by• Mozart, there is no world-class Mozart before ing tend to be relatively free from intense emotion; they
the age of seventeen.) Of course Mozart was already com- may be described as cold cognition. But sudden discovery,
posing at the age of four, so that by age seventeen he had the "aha!" experience, tends to evoke emotion; it is hot
already been educating himself for thirteen years. Mozart cognition. Sometimes ideas come to people when they are
is typical of the child prodigies whose biographies Hayes excited about something.
has examined. A sine qua non for outstanding \Vork is
diligent attention to the field over a decade or more. Emotion and Attention
Hence, in order to have anything like a complete theory
Summary: The Intuitive and Behavioral Models of human rationalitv, we have to understand what role
J

There is no contradiction bet\veen the intuitive model emotion plays in it. Most likely it serves several quite dis-
of thinking and the behavioral model, nor do the two tinct functions. First of all, some kinds of emotion (e.g.,
models represent alternative modes of thought residing in pleasure) are consumption goods. They enter into the
different cerebral hemispheres and competing for control utility function of the Olympian theory, and must be
over the mind. All serious thinking calls on both modes, counted among the goals we strive for in the behavioral
both search-like processes and the sudden recognition of model of rationality.
familiar patterns. Without recognition based on previous But for our purposes, en1otion has particular impor-
experience, search through complex spaces would proceed tance because of its function of selecting particular things
in snail-like fashion. Intuition exploits the knowledge we in our environments as the focus of our attention. Why
28 29
VISIONS OF RATIONALITY VISIONS OF RATIONALITY

was Rachel Carson's Silent Spring so influential? The

Emotion in Education
problems she described were already known ecologists
I would like to take a brief excursion at this point in
and the other biologists at the time she descnbed them.
order to consider the role of emotion in education. If
But she described them in a \Vay that aroused emotion,
literarv and artistic \vorks have a considerable power to
that riveted our attention on the problem she raised. That ' .
evoke emotions as thev certainly do, does this power sug-
emotion, once aroused, wouldn't let us go off and worry ' .
gest any special role for them in the educational process?
about other problems until something had been done
We all know that the humanities feel a bit besieged
about this one. At the very least, emotion kept the prob-
today. A large proportion of the students. in our
lem in the back of our minds as a nagging issue that
sities appear to want to enroll in law, business, or medi-
wouldn't go away.
cine, and the humanities suffer neglect, benign or other-
In the Olympian model, all problems are permanently
wise. One argument that is often advanced by those who
and simultaneously on the agenda (until they are solved).
would counter this trend is that it may be better, more
In the behavioral model, by contrast, the choice of prob-
effective, for students to learn about the human condition
lems for the agenda is a matter of central importance, and
bv exposure to the artist's and humanist's view of the
emotion may play a large role in that choice.
than by exposure to the scientist's. Of course
Emotion does not always direct our attention to goals
own professional identifications put me on the other. s1de
we regard as desirable. If I may go back to my example of
of the argument, but I think we should look at the tssue
Meinl(ampf, we observed that the reasoning in that book
quite carefully. What are the optimum for effi-
is not cold reasoning but hot reasoning. It is reasoning
cient human learning about central and Important mat-
that seeks deliberately to arouse strong emotions, often
ters? Which is better, cold cognition or hot? And \vhich-
the emotion of hate, a powerful human emotion. And of
ever is better, will \Ve find that this is the kind we associate
course, the influence of Mein J(ampf, like that of Silent
with the sciences or the humanities?
Spring or Picasso's Guernica, was due in part to the
I should say here that I have heard physicists argue for a
fact that it did have evocative power, the abthty to arouse
strong infusion of hot cognition in teach_ing their subject.
and fix the attention of its German readers on the particu-
The problems that excite them, and motivate them to un-
lar goals it had in mind.
derstand rather abstruse matters,. are the cosmological and
A behavioral theory of rationality, with its concern for
philosophical problems associated with fundamental
the focus of attention as a major determinant of choice,
particles, and with astrophysics and the the
does not dissociate emotion from human thought, nor
universe. So perhaps I should not have associated sc1ence
does it in any respect underestimate the powerful of
strictly with cold cognition.
emotion in setting the agenda for human problem solvtng.

30 31
VISIONS OF RATIONALITY VISIONS OF RATIONALITY
'

But let me go to a domain \Vhere the point can be made ates a great deal of literature today-at the very time when
more unequivocally and convincingly. Perhaps some of Freudian theories are being revised radically by new psy-
you are familiar with Arthur Koestler's Darkness at Noon. chological kno\vledge. There are few orthodox Freudians
It is a novel that describes what happens to a particular left in psychology today. Hence there is a danger, if we
person at the time of the Russian purge trials of the 1930's. take this route of asking the humanities to provide an
Now suppose you wish to understand the history of the emotional context for learning, that a kind of warmed-
Western world between the two world wars, and the over Freud will be served to our students in a powerfully
events that led up to our contemporary world. You will influential form. We have to re-evaluate the great human-
then certainly need to understand the purge trials. Are you ist classics to see to what extent they suffer from obsoles-
more likely to gain such an understanding by reading cence through the progress of our scientific knowledge.
Darkness at Noon, or by reading a history book that deals Homer is still alive because the Iliad and the Odyssey
with the trials, or by searching out the published tran- treat mainly of matters in which modern social science has
scripts of the trial testimony in the library? I would vote not progressed much beyond lay understanding. Aristotle
for Koestler's book as the best route, precisely because of is barely alive-and certainly his scientific works are not,
the intense emotions it evokes in most readers. and his logic hardly. And we could have a great argu-
I could go down a long list of such alternatives: War and ment with philosophers as to ·whether his epistemology
Peace versus a treatise on military sociology, Proust and or his metaphysics has anything to say to students today.
Chekov versus a textbook on personality. If I were in a And Lucretius, of course, talking about atoms, has gone
position ·where I had to defend the role of the humanities entirely.
in education, to provide an argument for something like The moral I draw is that, whereas works capable of
the traditional liberal arts curriculum of the earlv, twen- evoking emotion inay have special value for us just by
tieth century, I would argue for them on the grounds that virtue of that capability, if we wish to use them to educate,
most human beings are able to attend to issues longer, to we must evaluate not only their po\ver to rouse emotion
think harder about them, to receive deeper impressions but also their scientific validity when they speak of matters
that last longer, if information is presented in a context of of fact.
emotion- a sort of hot dressing than if it is presented If the humanities are to base their clain1s to a central
whollv without affect.
J place in the liberal curriculum on their special insights into
But educating with the help of hot cognition also im- the human condition> thev must be able to sho\v that their
'
plies a responsibility. If we are to learn our social science picture of that condition is biologically, sociologically, and
from novelists, then the novelists have to get it right. The psychologically defensible. It is not enough, for this par-
scientific content must be valid. Freudian theory perme- ticular purpose, that humanistic works move students.

32 33
VIS IONS OF RATIONALITY VISIONS OF RATIONALITY

They must move them in \vays that will enable them to live processes of intuition. The intuitive theory, I have argued,
with due regard for reason and fact in the real world. I do is in fact a component of the behavioral theory. It empha-
not mean to imply that the humanities do not now meet sizes the recognition processes that underlie the skills hu-
this standard; a detailed assessment of the liberal curricu- mans can acquire by storing experience and by recogniz-
lum in any existing university would certainly not give a ing situations in which their experience is relevant and
simple yes-or-no answer to that question. But I do suggest appropriate. The intuitive theory recognizes that human
that any examination of the appropriate roles of differ- thought is often affected by emotion, and addresses the
ent fields of knowledge in providing the materials of a question of what function emotion plays in focusing hu-
liberal education needs to give close attention both to the man attention on particular problems at particular times.
emotional temperature of material and to its empirical I have left for the next chapter a fourth theory: the
soundness. vision of rationality as evolutionary adaptation. The evo-
lutionary model is a de facto model of rationality; it im-
CONCLUSION
plies that only those organisms that adapt, that behave as if
In this first chapter, I have sought to present three vi- they \Vere rational, will survive. In the next chapter, I shall
sions of rationality: three ways of talking about rational examine these claims of the efficacy and centrality of natu-
choice. The first of these, the Olympian model, postulates ral selection as applied to the exercise of human rationality.
a heroic man making comprehensive choices in an inte-
grated universe. The Olympian vie\v serves, perhaps, as a
model of the mind of God, but certainly not as a model of
the mind of man. I have been rather critical of that theory
for present purposes.
The second, the behavioral model, postulates that hu-
man rationality is very limited, very much bounded by the
situation and by human computational powers. I have
argued that there is a great deal of empirical evidence
supporting this kind of theory as a valid description of
ho\v human beings make decisions. It is a theory of how
organisms, including man, possessing limited computa-
tional abilities, make adaptive choices and sometimes sur-
vive in a complex, but mostly empty, world.
The third, the intuitive model, places great stress on the

34 35