Content uploaded by Hartmut Kliemt
Author content
All content in this area was uploaded by Hartmut Kliemt on Feb 04, 2019
Content may be subject to copyright.
A&K
Analyse & Kritik 2018; 40(1): 131–159
Hartmut Kliemt*
On the Nature and Significance of (Ideal)
Rational Choice Theory
https://doi.org/10.1515/auk-2018-0006
Abstract: The increasingly wide spread use of RCM, rational choice modeling, and
RCT, rational choice theory, in disciplines like economics, law, ethics, psychology,
sociology, political science, management facilitates interdisciplinary exchange.
This is a great achievement. Yet it nurtures the hope that a unified account of ratio-
nal (inter-)active choice making might arise from ‘reason’ in (a priori) terms of in-
tuitively appealing axioms. Such ‘rationalist’ characterizations of rational choice
neglect real human practices and empirical accounts of those practices. This is
theoretically misleading and practically dangerous. Searching for a wide reflec-
tive equilibrium, WRE, on RCT in evidence-oriented ways can explicate ‘rational’
without rationalism.
Keywords: rationalism and unity of science, rational choice, bounded rational-
ity, critical rationalism, Goodman’s induction, Selten’s methodological dualism,
Weber’s ideal types
1Introduction and Overview
Emergent from common Hobbesian roots of theorizing ‘more geometrico’ about
how people would and/or should behave modern (axiomatic) rational choice
theory—RCT, henceforth—and its formal language of rational choice modeling—
RCM, henceforth—exist since mid 20th century.¹The increasingly wide spread
use of RCM and RCT—in disciplines like economics, law, ethics, psychology, so-
ciology, political science, management—facilitates interdisciplinary exchange.²
1I am aware, of course, that RCT also stands for randomized controlled trials which form one
of the pillars of proper evidence oriented research on ‘moral’, ‘medical’, ‘political’, ‘managerial’
etc. ‘subjects’. The use of RCT in the rational choice sense is orthogonal to randomized controlled
trials.
2RCM has the invaluable merit of forcing researchers to explicitly formulate their assumptions
about what is and what is not subject to the causal influence of choices within the rules of a given
*Corresponding author: Hartmut Kliemt, c/o Prof. Dr. Max Albert, Volkswirtschaftslehre,
Justus-Liebig-Universität Gießen, e-mail: hartmut.kliemt@t-online.de
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
132 |Hartmut Kliemt
A&K
This is a great achievement. It nurtures the hope that a unified account of ra-
tional (inter-)active choice making might arise from ‘reason’ in (a priori) terms
of intuitively appealing axioms. ‘Rationalist’ characterizations of rational choice
that neglect real human practices and empirical accounts of those practices are,
however, theoretically misleading and practically dangerous.³
To illustrate the achievements, limits and risks of RCT the next two sections
will locate both RCT and my discussion of it on a coarse intellectual map of sci-
ence and philosophy (section 2 and 3). This sets the stage for a critical assessment
of the scope and limits of Reinhard Selten’s methodological dualism and his strict
separation of ‘ideal’ (a priori) and ‘real’ (a posteriori) RCT (section 4). The next sec-
tion opts for the unity of science by interpreting RCT as a nomological discipline
focusing on the explanation of ‘rational’ behavior in empirical terms other than
its rationality (section 5). Then I try to cope in critical rationalist terms with the
notorious question of how the descriptive interact with the prescriptive dimen-
sions in our search for a wide reflective equilibrium, WRE (section 6). Summary
conclusions end the paper (section 7).
2Putting RCT into Perspective
The graphical overview of this section (graph 1) and the comments concerning it
are formed according to the kiss—keep it simple stupid—principle. Without claim-
ing to present a fully-fledged philosophical argument they illustrate possible rela-
tions between ‘mother philosophy’ and ‘moral science’ as its (il-)legitimate social
science ospring was once called. Their sole purpose is that of providing a coarse
account of where—within a history of ideas context—the subsequent discussion
is located on the ‘intellectual map’.
I distinguish between philosophical approaches to human behavior that fol-
low the philosophical tradition of using the terms ‘philosophy’ and ‘science’
mostly synonymous and approaches that strictly separate philosophy and sci-
ence. Endorsing a broadly Humean view that emphasizes the continuity between
interaction (game) and the law-like hypotheses by which consequences of action are predicted;
see in detail on RCM vs. RCT, Güth/Kliemt 2007.
3The entertaining but often absurd ‘story telling’ in so-called economic imperialism forms a
relatively harmless case in point. At best it can deliver interesting explananda but no empirical
explanations; see McKenzie and Tullock 1978. Equilibrium assumptions like that of ‘arbitrage
free financial markets’ will in combination with the apparatus of ideal RCT camouflage ignorance
and uncertainty as if it were risk and thereby nurture some of the dangerous control illusions of
financial engineering whose consequences we experienced in the financial crises.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |133
philosophy and science as rational endeavors (upper branch in graph 1) rejects
the traditional rationalist prima philosophia claim of ruling over science in a
manner impervious to scientific knowledge and criticism.⁴
(Moral) philosophy →(Moral) science 1
Philosophical accounts of
human behavior
Hume: Philosophy = Science
Schlick: Philosophy as
Proto-science
Mindless BehavioralScience
Neo-Classical Economics
Psychological Behaviorism
Mindful accounts of human action
under behavioral laws
Philosophy ≠ Science
Kant-Husserl
Apriori-Non-science
Selten’s ideal RCT
Mises’ Kantian Praxeologywith
apodictic apriori knowledge
Austrian ontologyof the relatively
absolute absolutes
Fig. 1
To the extent that ideal RCT implies such a claim it is subject to basically the same
criticisms as classical rationalism. I will not engage the discussion of classical ra-
tionalism here but rather focus on Reinhard Selten’s ideal theory of interactive
rational choice making which avoids classical criticisms by categorically separat-
ing RCT from disciplinary science (as a system of broadly ‘experiental’ insights).
The ideal rational choice theory part in Selten’s methodological dualism does not
make any of the traditional rationalist claims of characterizing a priori what the
scope and limits of a posteriori considerations may be. Since it also denies the
Humean continuity between philosophy and science it, however, raises questions
concerning the role of such an ideal RCT.⁵
4Beyond the claim of a foundational unity of philosophy and science I follow Moritz Schlick
1986 in regarding it as a—if not the—primary task of philosophy to prepare fields of intellectual
interest such that they can be handed over to a disciplinary evidence-oriented treatment (i.e.
a science in the present disciplinary sense). Schlick took his inspiration from the psychologist
Oswald Külpe (Külpe 1897); see in particular chap. IV, §31, problem 3.
5Many of the so-called Austrian, in particular von Mises type, economists seem to endorse a
methodological dualism with such stronger rationalist claims; see Kliemt 2017.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
134 |Hartmut Kliemt
A&K
It is not accidental that Selten refers to ideal RCT as ‘rationology’—in analogy
to theology. The perfectly rational cognitively unlimited individual envisioned in
ideal RCT is akin to the infinitely powerful, benevolent and knowledgeable be-
ing(s) studied by (‘natural’) theology. In the rationological study of ideal RCT in-
finitary assumptions lead to the same discontinuity with worldly phenomena (see
section 4) as infinitary conceptions in case of the envisioned deity. In both cases
issues of internal coherence as well as the credibility and origins of ‘assumptions’
emerge as central.
Undoubtedly rationology is a potentially fascinating intellectual field. Yet, its
role may be as problematic as that of theology. It invites interpretations of approx-
imability and explanatory power which in fact are non-existent.⁶Like theology,
rationology may support practices that must be regarded as harmful in terms of
our worldly common sense. And, the empirical evidence concerning many (ab-)
uses of RCT seems to suggest that this ‘model risk’ is quite real.
I shall henceforth focus on the uneasy relationship between science and in
some—still to be clarified—sense of those terms ‘ideal’ and/or ‘realistic’ variants
of RCT. With respect to RCT it is crucial to distinguish between approaches that
are at root Humean psycho-logical and those that are Hobbesian decision-logical.⁷
The former are experiental in the modern sense of empirical (cognitive) psychol-
ogy of (boundedly) rational behavior and stand at least in continuity with mod-
ern conceptions of evidence-oriented science. The latter are non-empirical in that
they proceed in the classical spirit of an a priori analysis ‘more geometrico’ or
some form of analysis akin to it (e.g. Kantian transcendental arguments). What-
ever their attractions, they are not in line with modern conceptions of evidence-
oriented science.⁸—Before turning to modern ideal theory conceptions that try to
clarify a ‘normative’ ideal of rational choice under the premise that rational choice
makers behave in ways compliant with the prescriptions of RCT it is useful to go
back to ‘square one’ and to recapitulate essentials of a decision-logical Hobbesian
approach to what traditionally were dubbed ‘moral subjects’ (prescriptive and de-
scriptive).
6As a consequence of its ‘otherworldliness’, ideal RCT cannot be treated as a Weberian ideal
type approximable by real types.
7Perhaps my reading of Hobbes is too extreme, see again Külpe 1897, 10. This stylized account
is meant to emphasize those traits of the Hobbesian approach that render it a clear pre-cursor of
the disciplinary radicalization that later should become economic imperialism but was already
recognized by the British Moralists who responded to Hobbes; see Kliemt, 2009 and below.
8I hasten to add that Reinhard Selten is one of my intellectual heroes because he is a champion
of both fields.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |135
3Pillars of Hobbesian RCT
Spinoza’s streamlined endorsement of originally Hobbesian views is characteris-
tic for the history of ideas trajectory along which RCT developed to its present form
and inherited most of its foundational problems along the way⁹:
“Now it is a universal law of human nature that no one ever neglects anything which he
judges to be good, except with the hope of gaining a greater good, or from the fear of a
greater evil; nor does anyone endure an evil except for the sake of avoiding a greater evil,
or gaining a greater good. That is, everyone will, of two goods, choose that which he thinks
the greatest; and of two evils, that which he thinks the least. I say advisedly that which he
thinks the greatest or the least, for it does not necessarily follow that he judges right. This
law is so deeply implanted in the human mind that it ought to be counted among the eternal
truths and axioms.
As a necessary consequence of the principle just enunciated, no one can honestly forego
the right which he has over all things, and in general no one will abide by his promises,
unless under the fear of a greater evil, or the hope of a greater good [.. .]. Hence though men
make promises with all the appearances of good faith, and agree that they will keep to their
engagement, no one can absolutelyrely on another man´s promise unless there is something
behind it. Everyone has by nature a right to act deceitfully, and tobreak his compacts, unless
he be restrained by the hope of some greater good, or the fear of some greater evil.” (Spinoza
1951[1670], 203–204)
In view of the preceding citation it seems somewhat strange that economists tend
to refer to Adam Smith as the founding father of their discipline. He certainly was
the first institutional economist and he shared with his older friend Hume the fo-
cus on what nowadays became experimental and (psychological) behavioral eco-
nomics. However, Hobbes and, even more so, his direct followers like Spinoza,
were much closer to the RCT approach endorsed by modern economists than was
Adam Smith (and, for that matter, Hume).¹⁰
9Spinoza shook o the last residuals of the older tradition that lingered on in the workof Hobbes
to endorse a fully ‘economic’ rational choice account; see on this also Steinberg 2013.
10 Hume endorsed the homo oeconomicus model only as a contrary to fact ‘stress test’for mor al-
political institutions. He suggests “that, in contriving any system of government, and fixing the
several checks and controls of the constitution, every man ought to be supposed a knave and to
have no other end, in all his actions, than private interest” (Hume 1985, VI/I, 42). But Hume says,
too, that “it appears somewhat strange, that a maxim should be true in politics, which is false in
fact” (Hume 1985, VI/I, 42–43).
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
136 |Hartmut Kliemt
A&K
The Hobbesian empire of moral science was built on three pillars (that should
re-emerge as defining characteristics of the so-called ‘economic imperialism’ of
the second half of the 20th century):
–First, theories of human (inter-)action should be derived ‘more geometrico’
from first ‘self-evident’ a priori principles of individual behavior (‘a priorism’).
(1.)
–Second, overt human behavior is to be explained as serving individual actors’
interests in reaching aims, ends or values (‘self-regarding individualism’). (2.)
–Third, behavior can be explained exclusively by relating it to (future) causal
consequences of each act taken separately within consistent individual
choice-making (‘consistent case by case opportunism’). (3.)
The three elements show up in Hobbes’ original account. The briefest confirma-
tory passage (1.) is supported in Chap. 4 Leviathan and passim while for (2.),
(3.), the briefest confirmatory passage is the first sentence of the central Chap. X
Leviathan (Hobbes 1968[1651]) which commences with: “The Power of a Man, (to
take it universally,) is his present means, to obtain some future apparent good.”
The present means part is important since it draws attention to power as potential
of an individual to reach her aims, ends, or values. This potential is not power
over somebody as in the Weberian concept of power. It is a potential “to obtain
some future apparent good”. Here the ‘some’, I take it, indicates that there is
a pluralism of goods, ‘future’ implicitly acknowledges the teleological element
in human action, ‘apparent’ makes it clear that human action aims at what is
subjectively perceived by the actor as an end.¹¹
On the whole, present neoclassical economics still seems to be built on the
three pillars of Hobbesian RCT. Individuals’ motives need not necessarily be self-
ish. As rational choice theorists have argued time and again consistent pursuit of
11 There is no externally given objective end as in Aristotle. I readily admit that such an end
would emerge in an interpretation of Hobbes moving the ‘natural obligation’ to secure survival
center stage. If this medieval element in Hobbesian terminology and thought is taken seriously
a completely dierent type of moral science emerges. It is either leading back to natural law tra-
ditions that do not cohere with the body of Hobbes’ work—in particular not with his concept of a
natural right as absence of any obligation—or it may lead to an anticipation of evolutionary com-
petition of the Schumpeterian kind. In the latter case one has to take into account the Hobbesian
insight that humans need to engage in pre-emptive strikes out of ‘defensio’; they are in foro in-
terno under an obligation to hope that this may not be necessary (they understand in modern
parlance the Pareto superiority of non-pre-emption) yet in foro externo they are justified to act
otherwise; see on this also Kant 1977[1798] who in his metaphysics of morals 39–43] expresses
similar views.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |137
ends according to well-defined complete preference orders suces to represent
human opportunity seeking action as if maximizing a well-defined function.¹² To
put it slightly otherwise, to unleash the analytical power of RCT—as a ‘tidy’ theory,
in which preference orders can be represented by the natural order of the values
of a real valued function—the aims ends or values must be pursued consistently
in a self-regarding manner but otherwise may be of whatever content.¹³
Much of the neo-classical tradition in economics has focused on the consis-
tency aspect of rational choice making. It became often oblivious of the fact that
at the Hobbesian origin pursuing aims consistently was merely one and perhaps
the less important aspect of rational choice as compared to the faculty to act in an
opportunity-seeking case-by-case manner that takes into account only the future
causal consequences of each act taken separately. As John Hicks succinctly put
it in his discussion of ‘causality in economics’: “people would act economically;
when an opportunity of an advantage was presented to them they would take it.”
(Hicks 1979, 43)
Spelled out in terms of opportunity-seeking the assumption that ‘people
would act economically’ can be interpreted in at least two dierent ways.¹⁴ These
lead to dierent conceptions of idealization in RCT: On the one hand, we are in-
formed that human actors will seize opportunities to an extent that allows to build
economic models on the assumption that the underlying generalization is as a
matter of fact approximately fulfilled. On the other hand, we are informed about a
presupposition of an ideal theory discussion that intends to develop prescriptions
for ideally rational behavior under conditions of universal compliance with the
characterization of ideal behavior and the prescriptions or recipes derived from
it.
The role of idealizations of the first kind is akin to the role of Weberian ideal
types in social and assumptions like frictionless motion in natural science. The
role of idealizations of the second kind is akin to Morgenstern’s condition of ‘the-
ory absorption’—or rather absorbability—according to which it must be possible
that the prescriptions of a theory of ideally rational action in situations of inter-
active choice making hold good if the theory is commonly known and universally
observed.
If idealizations in RCT were exclusively of the first, Weberian kind they could
play an obviously legitimate role as approximately valid stylized generalizations.
The corresponding RCT would be in continuity with science in that it could be
12 As if maximization and fulfilling the appropriate consistency and continuity conditions char-
acterizing rationality amount to the same.
13 See on this clearly and succinctly Hausman 1992, ch. 2.
14 Not as an exegetic question concerning Hicks’ work but as a systematic, general one.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
138 |Hartmut Kliemt
A&K
critically assessed by means of evidence-oriented science. Contrary to this, ide-
alizations of the second kind need not make any claim to approximate real be-
havioral facts. The failure of such ideal RCT to live up to some reality test is not
decisive. In fixing the ‘prescriptions’ of what should ideally be done the prescrip-
tions are assumed to be ‘descriptions’ of what would be done under contrary to
fact assumptions.
The ideal theory leg of the methodological dualism of Reinhard Selten is not
Weberian. As a philosophically advanced approach to ideal RCT it holds that the
theoretical idealizations of ideal RCT are projections of pre-theoretical thoughts
about ideal rationality. It intends to assess the speculative projections in terms
of their inner logic and under the ‘side-constraint’ of absorbability of ideal RCT
among perfectly rational reasoners with unlimited reasoning capacities.In short,
ideal RCT shows how humans would project their intuitive boundedly rational
views concerning full or ideal rationality if they were equipped with the mathe-
matical skills necessary to spell out such a projection precisely and completely.¹⁵—
As illustrated next, rejecting the continuity between philosophy and science, Sel-
ten’s methodological dualism leads along the lower branch to the lowest of the
end-nodes of graph 1.
4Methodological Dualism and RCT
In terms of widely shared stereotypes the two arguably most prominent self-
declared methodological dualists in economics, Ludwig von Mises and Reinhard
Selten, make strange bed-fellows. Mises and his followers cultivate resentment
against RCM (unfortunately identifying it with an old-fashioned conception of
‘quantitative methods’). Selten and fellow game theorists endorse RCM and try to
push it to its mathematical limits. Yet, Mises and Selten in his incarnation as a
full-rationality theorist (his cognitive psychologist incarnation will be addressed
below as a dierent matter altogether) are close to each other in that both are to
be placed on the non-science branch of graph 1.¹⁶
15 The complete separation between ideal RCT and real theory implied by Selten’s methodolog-
ical dualism also implies that large parts of, for instance, economic general equilibrium theory
are to be classified not as behavioral science but rather as non-science.
16 The bounded rationality theorist Selten who tries to develop a ‘mindful account of human
action under behavioral laws’ is firmly located on the upper branch of graph 1 and its middle end
node.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |139
As a theorist of full rationality Selten avoids the more extreme claims to gener-
ate (apodictic) a priori insights about rationality and intentionality that Mises and
his adherents tend to raise. Yet, due to Selten’s focus on what individuals think
‘about rational behavior’ his approach to full rationality is very close to Mises’
praxeology and its Kantian undertones.¹⁷
4.1 Selten’s Ideal Theory
The problem of absorbability of theories—that is, whether theories about hu-
man interaction can remain true if the fact is accounted for that the entities
whose actions are described and predicted by theories, understand the theories
and respond to this knowledge by opportunity taking changes of their intended
actions—is akin to familiar ideas about self-arming and self-refuting theories
(prophecies). Through Oskar Morgenstern it found its way into game theory.¹⁸
Here is what leading game theorist Reinhard Selten says about ideal RCT to which
he refers as ‘normative’ decision and game theory:¹⁹
“[... ] let me explain my epistemological position on methodological dualism. In my view,
there is a fundamental dierence between normative and descriptive decision and game
theory. Normative decision and game theory has the aim of exploring full rationality and its
consequences. Full rationality is an ideal about the adequacy and coherence of decision-
making. It is not meant to be descriptive of how human beings actually behave, but rather of
what they think about the structure of the behavior of an idealized decision-maker without
any cognitive limitations. This idealized decision-maker is a mythical hero, whom we may
call ‘fully rational man’. Real people have limited powers of logical deduction and com-
putation, but fully rational man has instant access to everything that needs to be logically
deduced or computed for adequate and coherent decision-making. [.. .]
17 Praxeology tries to develop RCT in the Hobbesian spirit assubstantive knowledge a priori (ei-
ther of the Kantian kind or that of classical Euclidean geometry). Selten and fellow gametheorists
along with mainstream mathematical economists endorse modern more formalist conceptions of
mathematics and axiomatization. Selten is accepted as an important mathematical economist by
the present economic mainstream whereas von Mises and his followers are—in view of their in-
terpretation of and resentment against RCM with some justification—sidelined as ‘old fashioned’,
non-orthodox social theorists. Yet there is much more in common between the two main method-
ologically dualist camps than superficially meets the eye.
18 For more on Austrian influences in the conception of game theory by von Neumann and Mor-
genstern, see Leonard 2010.
19 From now on I will focus on the Selten position and only occasionally comment on its rela-
tions.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
140 |Hartmut Kliemt
A&K
Empirical arguments are irrelevant for normative decision and game theory. What counts is
the appeal to underlying tendencies in the thinking about what fully rational man is like.
The situation is similar to theology, which is concerned about what we should think about
God. [.. . ] In view of the analogy to theology, the study of fully rational man may be called
‘rationology’.” (Selten 1999, 303–304).
The ideal rationality of ‘fully rational man’ is, as Selten puts it in the citation, “not
meant to be descriptive of how human beings actually behave, but rather of what
they think about the structure of the behavior of an idealized decision-maker”
(emphasis added, H. K.).²⁰ In delivering his stylized account of ‘thoughts’ concern-
ing ideal rationality Selten relies on two crucial contrary to fact assumptions one
explicit and one implicit. Explicitly he states that “fully rational man has instant
access to everything that needs to be logically deduced or computed for adequate
and coherent decision-making”. Implicitly he accepts the constitutive premise of
game theory that the concept of ‘full rationality’ in interactive decision-making
is to be developed under the presumption that real behavior complies with ideal
RCT.²¹
In terms of familiar philosophical discussions Selten aims to develop a so-
called ideal theory.²² Within Selten’s rather precise game theoretically informed
framework the ideal theory should be absorbable under ideal conditions.²³ There
are no cognitive limitations and the theory of rational play is common knowledge
among fully rational actors. The rational actors ascribe rationality as explicated
by the theory itself to all actors and behave accordingly themselves presuming
that all others do and know that they do etc. Under full compliance with the the-
ory no actor should have a rational reason to behave other than the (‘normative’)
theory of ideally rational behavior recommends under the presumption that the
theory is fully complied with by all other actors.
20 As in the case of Herbert Hart’s The Concept of Law, to explicate ‘law’ a whole theory must be
developed in case of ‘ideal rationality’ in interactive choice making. Otherthan in the case of Har t
1961 who wanted to provide a realistic theory built on stylized facts, Selten intends to present an
ideal theory separate from behavioral facts; see on ‘explication’, Carnap 1956, appendix.
21 Examples of the by now extended discussion of ‘ideal theory’ are Brennan/Pettit et al. 2005;
Anton Leist drew my attention to Ypi 2010 as background for the present context.
22 See for a useful (economist’s) account of the philosophical discussion Hamlin/Stemplowska
2012, where rationology would presumably be subsumed under “the theory of ideals”.
23 The theory addresses all actors by its prescriptions that recommend what they should do and
by its descriptions that predict what they would do under certain circumstances.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |141
4.2 Why Selten’s RCT Does Not Characterize a Weberian Ideal
Type of Rational Behavior
Selten’s rationology drives the aforementioned (Hobbesian) philosophical con-
ception of opportunity taking behavior to its extremes. Dynamic choice-making is
about exerting proximate causal influences without emotional and cognitive con-
straints exclusively in a non-myopic view of all future expected consequences of
each choice taken separately. Case-by-case opportunism of the underlying teleo-
logical action model is exercised without any emotional or other decision-‘inertia’.
Rational individuals can and—by assumption—will immediately shift their behav-
ioral gears should they perceive an opportunity for exerting a causal influence on
the future (serving their ‘given’ aims, ends or values).²⁴
To see why the preceding ‘idealizations’ create models of RCT that are behav-
iorally non-approximable consider the central RCT example of the repeated pd,
prisoner’s dilemma, that is typically taken to show exactly the opposite: by ex-
tending the time horizon, ideal model behavior increasingly approximates real
behavior. So, imagine an n-times, n>0, identically repeated 2X2 pd basic game
pd(n) of two actors i=A,B. Let Ci, Di, i=A,B refer to the co-operation, respectively,
defection move of A,B in the base game.²⁵ The unique dominant strategy solution
of the base game is (DA, DB). In the game of length n=1, on round t=1 the move
combination (DA, DB)(1) will be in equilibrium. However, with ‘Pi(t)’ interpreted
as ‘preferred by i at time t’ we have ‘(CA, CB)(1) Pi(DA, DB)(1), i=A,B, that is the equi-
librium result is Pareto dominated by the result of mutually co-operative behavior.
Note also, if at stage ‘t=0’—before the game is played—both A and B plan on mak-
ing a co-operative move the execution of these strategic plans could not be made
credible in pre-play communication among fully rational actors who know of each
other’s full rationality. Consistent joint pursuit of gains clashes with opportunity-
seeking behavior: A strategy can be chosen as a plan at t=0 yet the plan to move
24 In particular the familiar generalizing counterfactuals of ‘what if everybody would do the
same?’ and ‘what if I would always do the same?’ are treated as rationally irrelevant and motiva-
tionally impotent since the rational actor knows that fulfillment of the contrary to fact statement
is not among the causal eects of any single action. Genuine rule following behavior as well as
retributive inclinations (in particular retributive emotions) are ruled out by future directedness of
choice in view of the causal consequences of each single act taken separately. Together with the
assumption of unbounded cognitive capabilities such counter-intuitive consequences as payo
dominated solutions based on backward induction in centipede and finitely repeated prisoner’s
dilemma-games can be derived.
25 A few remarks on the formal structures involved must suce to indicate the basic line of ar-
gument. How it would have to be pursued in a fuller account seems rather obvious but would
create a lot of notational clutter.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
142 |Hartmut Kliemt
A&K
is not the move at t=1. The plan at t=0 can be executed only in the future when
the occasion to move (to exert a causal influence) at t=1 arises. Then, however,
since with n=1 there is no t>1 the expectation that there are no future causal con-
sequences will dictate—see Spinoza and Hobbes—that the co-operative strategy
will not be executed in view of the dominance of defection.
As the work of some collaborators of Selten vividly illustrates,²⁶ if the human
faculty to act opportunistically (the pillar 3 of Hobbesian RCT) is taken seriously
all continuity between ideal RCT and real behavior is lost. Accounts of orderly
behavior in terms of evolutionary selection will not work if full opportunism is as-
sumed²⁷—evolution can only operate on what is invariant rather than on fully flex-
ible behavior—and even so-called Folk Theorems of eductive game theory (Bin-
more 1987/88) will not ‘survive’. The conventional claim that ideal RCT may be
regarded as approximately representing central characteristic features of real ra-
tional behavior brakes down.
To see why exactly, return to the simple example of the pd(n) but now consider
pd(∞) a game identically repeated indefinitely (at least potentially).²⁸ Then—
except for renumbering—the future looks strategically identical, independently
of any preceding history.²⁹ An exclusively future directed analysis of structurally
identical expected futures should come to the same conclusion regardless of the
preceding history. That is, if the history has been (CA, CB)(k) for k=1, 2, . . . , n the
continuation strategy with respect to the future must be the same as after (DA,
DB)(k) for k=1, 2, ... , n and also for any other starting point of a subgame starting
at t=n.
An additional argument is required to show what the identical supergame
strategy should be. In view of backward induction arguments³⁰ that single out
26 See in particular Güth/Leininger/Stephan 1991.
27 Only indirect evolution that operates on the rules of the game can be modeled to work while
upholding the teleological case by case model of opportunistic action which precludes any-
thing but choosing strategies dominant with respect to the future; see on indirect evolution
Güth/Kliemt 2000; Güth/Kliemt/Peleg 1999 or Berninghaus/Güth/Kliemt 2003.
28 The assumption that the end point is stochastic will not help; for, if backward induction is to
be avoided, a continuation probability strictly greater than the positive threshold below which
backward induction would kick in must apply indefinitely.
29 Think of the natural numbers, if you cut o the first n, you can renumber and start with n+1
as the new first number in the progression which is structurally identical with the original one.
30 If there is a last round n in which defection is dominant, then on round n-1 it must be com-
monly known among fully rational actors that behavior on round n-1 cannot induce rational ac-
tors on round n to behave other than dictated by dominance. Therefore, the dominance of defec-
tion on round n-1 will be eective. This is known among the actors on round n-2 . .. etc. until the
all-D behavior emerges.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |143
strategies implying move combinations (DA, DB)(k) for k=1, 2, . . ., n and all finite
n, only (D∞, D∞) will be approximable in pd(∞).³¹ Under conditions of ideal ra-
tionality, contrary to folk wisdom, extending the time horizon to infinity, though
avoiding backward induction, will not yield any co-operation as observed in real
human behavior.
The upshot of all this is that in a world of fully rational strategic actors who
decide exclusively opportunistically in view of substantive pay os so-called con-
ditional co-operation cannot emerge. Consequently, the so-called ‘order problem’
has no plausible solution among fully rational actors who are extrinsically mo-
tivated by substantive payos.³² Taking the ideal model seriously co-operative
orderly behavior cannot be approximated at all. Since we do as a matter of fact
observe human co-operative behavior, strictly Hobbesian ideal RCT becomes an
empirical absurdity.
If we assume that ideal RCT correctly identifies and drives to its extreme what
we think is an essential human faculty—to seize opportunities—we must conclude
that ideal opportunistic rationality is otherworldly. Other than in case of Weberian
ideal types there is no continuity between the theory of ideally and the theory
of realistically rational choice. In Selten’s methodological dualism there are no
‘bridges’ between the ideal and the real theory strands.³³
If the preceding diagnosis holds good, then whatever the merits of rationology
it does not contribute to our experience-based scientific knowledge of the world.³⁴
To put it slightly otherwise, what we think about the world in terms of ideal RCT
does not tell us anything about the world as object of that thinking.³⁵ Selten is
right, ideal RCT is categorically separated from real RCT. In line with this in graph
31 Only so-called subgame consistency according to which structurally identical games should
have structurally identical strategic equilibria as solutions is plausible if only the future matters.
Strict future orientation rules out all path-dependence of play in structurally identical sub-games
and thereby co-operation conditional on past events.
32 Parsons 1968 is a locus classicus concerning the so-called Hobbesian problem of social order.
33 As far as science is concerned Reinhard Selten’s dualism is reduced to the monistic psycholog-
ical founding of rational choice modeling in terms of so-called bounded rationality which Selten
himself has always propagated. As an activity outside of science the a priori reasoning of pure
rational choice theory may play a role.
34 The other forms of a priori reasoning represented in the same end node box will not be dis-
cussed here since the focus is on RCT in its present mathematically sophisticated form not bur-
dened with the epistemic and ontological liabilities of the traditional variants.
35 Making their distinction between how the world is perceived and how it is ‘as such’ even Kan-
tians would concede this, yet insist that something can be known a priori about the world as an
object of experience.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
144 |Hartmut Kliemt
A&K
1 the lower branch on which Selten’s ideal RCT is located is separate from science.
The continuity between science and philosophy is interrupted.
Selten—like many Austrian economists and broadly Kantian scholars of so-
cial theory—allows for forms of rationality that transcend the rationality of sci-
ence. Where in particular Austrian economists mix their claim that a meaningful
fact-insensitive conception of intentionality and rationality exists with a (mis-)
conception of the limits of mathematization, Selten’s discontinuity claim arises
from a consequent application of mathematical reasoning to teleological forward
looking choice making. Both camps believe that there is a realm of meaningful
ideal RCT beyond human practices that are conventionally regarded as rational
(as in particular those of empirical science). In Selten’s case there is even an argu-
ment akin to a proof that ideal rationality as emergent from a projection of human
intuitive conceptions of rationality and real rationality fall apart.
The thesis that the human ability to make choices teleologically puts human
behavior beyond (causal) explanations based on law-like regularities—at least at
the present state of empirical knowledge—is independent of any limits of logic and
mathematics per se. As in particular the example of Selten’s methodological du-
alism shows it deserves to be taken seriously. It cannot be brushed aside by other
adherents of mathematical social science as a manifestation of the resentment of
those who feel left behind due to their lack of mathematical skills. In particular
the critics of ‘internalization of values’ and other concepts akin to intrinsic moti-
vation and purely subjective factors in choice making will get into trouble if they
take their own rhetoric seriously. Their a priori preference for behavioristic expla-
nations in terms of extrinsic motivation is within an evidence oriented conception
of scientific rationality unconvincing as long as they cannot beef up their claims
by presenting superior explanations in terms of ‘mindless RCT’. The alternative to
this kind of approach is a cognitive psychology version of ‘mindful RCT’. Both are
incoherent with ideal RCT.
To put it slightly otherwise, what is needed is an account of human ratio-
nal choice making that either avoids teleological concepts altogether (as in be-
haviorism) or embeds them in an empirical conception of teleological reasoning
(as in cognitive psychology). Some concept of ‘rational choice’ may be explicated
in terms of overt behavior and its ‘successes’ in (inter-active) choice making. A
‘mindless RCT’ that explains behaviorally ‘rational choice’ in terms of causal fac-
tors other than ‘reasoning’ (including in particular selection and adaptation in
competitive processes) should be explored. There is room for this, on the upper
branch of graph 1. However, there is also clearly room on the upper branch of
graph 1 for ‘mindful RCT’ which links rationality psycho-logically rather than log-
ically to reasons and reasoning.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |145
5The Nature and Significance of RCT
Mindful RCT acknowledges as a brute fact of human self-experience that real hu-
man actors command a basic faculty to distinguish between acting/intervening
(teleologically) into the course of the world and representing/predicting the
course of aairs under (causal) laws. An intentionally acting human being will
perceive herself as (‘self-controlled’) author of her actions. She will imagine her-
self as being able to reverse her intentions at any point in time. She is to herself a
maker rather than a predictor of her choices.³⁶ Whether or not she regards herself
as rationally or morally justified she will think of herself as intentionally bringing
about future causal consequences.
From an internal point of view, in a ‘first-person’ perspective, making a choice
is incompatible with predicting it. Yet this does not rule out predicting choices
from a point of view external to the choice maker. Humans can do better than
chance in predicting choices of fellow humans.³⁷ It seems that there can be predic-
tive and explanatory RCT while accepting the first-person agent-relative thinking
about choice making as separate from choice prediction.
The explanatory subjectivism of mindful RCT accepts the presence of other
mindful choice makers as a fact. It conjectures that to emulate or to reconstruct
the internal point of view to the action situation may be helpful for external ex-
planatory and predictive purposes. Whether or not this in fact is true depends on
whether or not reliable law-like hypotheses exist. Here as in other fields prophe-
cies turn into evidence based predictions to the extent that they can be based on
tested and corroborated law-like (stochastic) hypotheses.³⁸
Within a realist conception of RCT, law-like hypotheses can themselves not
be merely a matter of thinking about the world. They are so to say not only ‘in
the expecting mind’ of an observer but are conceived by the observer to operate
on observed actors as other ‘laws of nature’. In a ‘mindful’ psychological science
36 This will hold good even if the actor knows that according to results of neuroscience her
decisions have been fixed in her brain before they become conscious to her. On the surface of
conscience she will experience herself as a ‘self-propelled’ actor ‘pulled’ by the expected future
rather than being pushed by the past. Even if she is driven by retributive emotions of a positive
or negative kind and is conscious of it there will be the ‘illusion’ of commanding the faculty to
act otherwise.
37 To what extent predictions from an external point of view to actors are possible is an alto-
gether contingent empirical question; see on empirical evidence concerning the scope and limits
of prediction in human aairs, Tetlock 2009; 2015.
38 Like those identified in the heuristics and biases approach; see Kahneman 2012 for a popular
discussion.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
146 |Hartmut Kliemt
A&K
approach to human behavior the observer will typically pre-suppose that law-like
regularities relate subjective perceptions of action situations (stochastically) to
overt behavior of actors.³⁹
If an observer uses his reconstruction of a ‘subjective’ action situation of an-
other actor to predict the behavior of the other individual according to objective
law-like hypotheses concerning cognitive processes then—despite the reconstruc-
tion of the subjective situation—he adopts what has been called an ‘objective atti-
tude’ (Strawson 1962). Such an observer might use objective law-like regularities
to rationally ‘manipulate’ the actions of another rational individual in a purely
causal way. If the observer feeds back his information to the actor then the actor
can respond in a possibly ‘falsifying’ way in rational pursuit of her own ends (as,
e.g., in case of self-refuting prophecies) yet this alone does not change the situ-
ation into one in which the observer approaches the actor with what Strawson
called a ‘participant’s attitude’.
It is obviously hard to draw the relevant distinction between an argument of-
fered from a participant’s and one used for manipulation from an objective point
of view precisely.⁴⁰ An argument may be used for the sake of its causal eects or
for the sake of its internal validity and often for both reasons simultaneously.⁴¹ In
any event, ‘subjectivism’ as methodological precept of taking into account sub-
jective mental states as well as mental models of action situations in explanatory
RCT eorts is compatible with science in general and a ‘technological’ conception
of the application of scientific knowledge.
What individuals think about what is rational according to RCT can matter
as a causal factor. For this it needs to enter mental models of real subjects as a
real factor. In principle the corresponding RCT remains a branch of (cognitive)
psychology.⁴² That RCT may be known to the individuals who are described by it
39 The logic of a situation can be very compelling but empirically a law-like regularity that in-
dividuals will act according to what they perceive as implied by that logic is needed for a fully-
fledged explanation.
40 In a related context this distinction is more carefully discussed in Baurmann 1987. Whether
an act of manipulation or of, say, genuine agreement seeking occurred is contingent on the aims,
ends or values of the external actor. He can address ‘arguments’ towards ‘another mind’ with the
intention that the other should ‘own’ and thereby ratify them or as causal forces meant to bring
about predicted responses.
41 Without a ‘normative’ fact that somebody as a matter of fact intends to pursue certain aims,
ends or values there is no instrumental or technological ‘ought’. Which aims, ends or values as
a matter of fact prevail (are ‘given’ in the sense of Robbins 1935) decides to which technological
uses knowledge will be put.
42 The law-like regularities underlying cognitive psychology accounts are beyond human inter-
vention as much as those invoked by (‘old fashioned’) psychological behaviorism.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |147
does not in itself change its explanatory nature. However, it will make predictions
more and sometimes exceedingly complicated.⁴³
Psychological explanatory behaviorism that strives to explain overt behav-
ior exclusively in terms of regularities in overt behavior (uppermost end-node of
graph 1) would be a nice way out of the problems of unpredictability arising from
symmetric knowledge of RCT.⁴⁴ Law-like regularities that could explain regulari-
ties in overt behavior without taking recourse to the directly non-observable the-
oretical concepts of cognitive psychology would be methodologically preferable
for other reasons as well. However, the condition that the relevant explananda
of (inter-active) choice making can—at least approximately—be explained in be-
havioral terms will typically not be fulfilled. It seems that the regularities in overt
behavior that we as a matter of fact observe and desire to explain require taking re-
course to cognitive psychology concepts like in particular genuine rule following
behavior by a choice making actor from the actor’s internal point of view.
To the extent it is required for and conducive to explanations explanatory sub-
jectivism seems unavoidable. It is based on the presumption that the laws of na-
ture that govern rational choice relate to subjective perceptions and cognitive pro-
cesses. Yet the law-like regularities are assumed to exist ‘for real’ and not only as
a logic of situations etc. (locating such RCT at the lower end node of the upper
branch of graph 1).
The explanandum of the factual existence of social order that was used in the
preceding section to illustrate that ideal RCT will not work as adequate explanans
may serve as an illustration here. The game analyzed in RCT terms arises from a
game form that exhibits pd like material or substantive payos as extrinsic motives
or incentives. It is a pd in terms of directly observable substantive ‘pay os’ but
not necessarily in subjective payos. The subjective payos may be quite dierent
from the objective or substantive ones. Often, in a 2X2 pd game form individuals
i=A,B will form ‘subjective’ preferences according to which not only (CA, CB) Pi
(DA, DB), i=A,B but also, (CA, CB) PA(DA, CB) and (CA, CB) PB(CA, DB). Then the
unique dominant strategy solution of the base game is not anymore (DA, DB). If
43 See for an exploration of so-called theory absorption among boundedly rational actors,
Güth/Kliemt 2004 and on theory absorption more generally and precisely Dacey 1976; 1981, build-
ing on Morgenstern.
44 As for instance the example of finance shows rational choice makers who know RCT and
other theories about the world and know that they know will engage in theorizing in symmetric
ways that will render the predictive eorts self-refuting and the so-called random walk down Wall
Street may be the result. Reasoning about knowledge seems as a matter of fact influencing the
world.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
148 |Hartmut Kliemt
A&K
(DA, DB) PA(CA, DB) and (DA, DB) PB(DA, CB) a so-called AG, assurance game,
arises from a pd game form in substantive payos due to subjective factors.
This is merely one of the better known examples for the necessity of including
subjective factors in adequate explanations of (inter-active) rational choice mak-
ing.⁴⁵ It obviously generalizes to the claim that RCT can have adequate explana-
tory force only to the extent that it is a variant of cognitive psychology (focusing
on intentional human behavior in inter-active choice making). For, the subjective
perceptions of situations leading to games and not only the game forms matter. In
particular economists are prone to point this dierence between game and game
form out. However, they then take resort to their conventional strategy of ‘let us
assume that the game form and the game motivationally coincide’ which avoids
all the interesting and crucial empirical issues of the relationship between games
and game forms in realist RCT.
Under what conditions individuals who are exposed to a pd game form are in
subjective preference terms interacting in games of the PD or AG type is an em-
pirical issue. Our ability to answer the related empirical questions on the basis
of nomological knowledge will be decisive for our ability to predict behavior. Yet,
however this may be, descriptive explanatory RCT in both, its cognitive psychol-
ogy and its behaviorist incarnation, remains firmly located on the upper branch
of graph 1 (regardless of the mindless or mindful nature of explanations).
Though I intend to stand by this account of the descriptive-explanatory nature
and significance of RCT I readily admit that RCT has always been closely associ-
ated with aspirations to make ‘better’ choices. These aspirations reach from im-
proving the prospects of getting one’s way in pursuit of self-regarding aims, ends
or values to realizing broadly speaking moral standards of showing interpersonal
respect in interactive choice making (as an ‘ethical’ extension of the participant’s
attitude). An account of the nature and significance of RCT would be thoroughly
incomplete without at least a few remarks relating such ‘normative’ RCT concerns
to descriptive-explanatory RCT. In doing so one has to bear in mind in particular
that Selten referred to rationology also as a ‘normative’ theory (though not one
that could be used to derive prescriptions for real behavior).
45 As opposed to games that represent the preferencesover plays of the game (all things consid-
ered), game forms represent only the material or substantive paths and outcomes. In particular, a
game form of a pd may in fact typically give rise to an assurance game with no dominant strategy
in preferences regardless of the fact that it exhibits a pd structure in material or substantive pay-
os; a fact to which Amadae 2016 rightly draws attention though due to her lack of the category
of a game form her presentation and understanding seems somewhat deficient.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |149
6RCT as a ‘Normative’ Account of Human Behavior
Awareness that ideal RCT is not an explanatory scientific theory makes it harder to
claim the prestige of science for normative purposes of prescribing what rational
actors ideally should do. In a certain sense the ‘rational’ in real RCT has, however,
as a matter of fact normative undertones in the wide sense in which, say, maxims
of prudence, recipes expressing ‘technological know-how’ and standards of ‘cor-
rect’ measurement also are normative.⁴⁶ In our search for a WRE on the theoretical
and practical merits of RCT we need to account for the normative uses of RCT as
well. An answer to this challenge would require a whole book. This being acknowl-
edged the subsequent remarks on WRE present merely some essential demands
that more complete and possibly more convincing answers to the challenge of ex-
plicating the normative aspect of real RCT would have to meet.
6.1 WRE on Descriptive and Prescriptive Claims
The term reflective equilibrium stands for systematically seeking a coherent ac-
count of general and specific claims which may in principle be of a prescriptive or
descriptive, an a priori or an a posteriori nature. The search is ‘wide’ if background
theories (which themselves may have been subject to separate equilibrium con-
siderations) are included in the process.
Without going into details and without making any claim to exegetically rep-
resent the views of Rawls or Daniels with whom the WRE concept is normally as-
sociated in philosophy the basic intuition may be presented the following way:⁴⁷
When humans try to systematize and to critically assess bodies of knowledge or
advice, coherence is the guiding aim. Typically, specific claims representing spe-
cific cases in which observations of what is true or false and also prescriptive or
evaluative intuitive judgments of what is right, wrong, better or worse are ‘(rela-
tively) prioritized’ in that the burden of proof is assigned to those who intend to
overrule them as invalid when they do not cohere with general or abstract claims.
To give a most simple example, assume somebody looks out of the window
observing the killing of another human being. He considers this as repugnant.
Assume also that he never thought about such killings before. Even such a single
46 See on what may be called minimum normativity of standard-settingthe special issue of Anal-
yse & Kritik 2016 (38), and the account of and examples of norm justification in WRE given there.
47 See Rawls 1951; 1971; Daniels 1979. Goodman 1983 whose views are closer to RCT consider-
ations in substance and form will be discussed towards the end this section; see for a succinct
account Hahn 2004.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
150 |Hartmut Kliemt
A&K
strong situational encounter may induce him to conclude that the general claim
that the killing of other human beings is allowed must be rejected. Let us assume
also that after further considerations of a general as well as additional observa-
tions of a particular kind he feels assured that killing of fellow humans is in gen-
eral wrong (and stops there in a temporary equilibrium of his reflections since for
the time being he sees no counter argument).
After a while the person is confronted with an attack of a robber armed with a
baseball club. To prevent becoming a victim, the attacked considers using his gun.
To do so would, of course, violate the general principle that ‘thou should not kill!’
that he came to endorse. Coherence will be reached once the principle is modified
to demand ‘killing is forbidden unless in self-defense’. New reflections may be
necessary if a robber attacks with a toy pistol or, say, with a banana. Depending
on often subtle circumstances, self-defense might not justify the use of the gun in
reflective equilibrium in such cases.—So much for the toy example which should
be sucient to illustrate the guiding idea.
If a reflective equilibrium is sought on ‘great questions’ like what are the prin-
ciples systematizing our views on justice or how to account for our practices of
induction (corroborating or refuting claims) systematically, a simple stylized pro-
cess as described before will not work. Without going into details a few remarks
may be sucient to sketch some of the necessary basic additions.
First, the simple linear process characterized before will not apply. General
principles may become so well entrenched that particular claims that contradict
them may be discounted in favor of general and abstract principles. This is tricky
business since the hurdles must be high before some ‘refuting’ evidence may be
discounted say in case of a general law-like hypothesis of natural science. Not
by chance do we have additional rules like the demand that it must be possi-
ble to replicate the falsifying phenomenon in a reliable and predictable way etc.
These additional rules will normally be sucient to neutralize fake evidence. Still,
clashes between specific claims and general evidence are not rare but very com-
mon. For instance, in cases like evidence-based medicine the intuitions of doc-
tors that stem from their personal experience often clash with general evidence-
based rules or recipes of good practice. Though there are rather well supported
secondary principles that would guide rational doctors to discard their own spe-
cific intuitions in cases where, say, the evidence of randomized controlled trials
speaks against their specific contrary experiences some of them would still rather
follow the testimony of their particular case-based experience. They could often
point to arguments that indicate that their cases are special and therefore the gen-
eral conclusions of the randomized trial do not apply. Nevertheless, the results of
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |151
randomized trials should carry much weight in light of the empirical evidence on
how randomized trials fare as compared to expert judgment in general.⁴⁸
A second and perhaps even more important point about addressing the ‘great
questions’ in a search for WRE is their relation to established practices. The ‘great
questions’ do not arise out of the blue. Specific claims arise in a context of real
(smaller) problems. Against this background the implicit empty slate premises
raised by claims to fact-insensitive intuitions of pure reason—or what have you—
seem rather absurd. We must start from a contingent state of mind and knowledge
when we search for a WRE and what arises will be dependent on the starting point.
The vision of path-independent answers to the ‘great questions’ seems—at least
to me—absurd.
More concretely, take Rawls’ intention to reach a WRE on fundamental prin-
ciples of justice. Whatever Rawls and many of his followers may have also said in
favor of path-independent claims, his original inclination—which he never fully
abandoned—to conceptualize the WRE search as relative to the experiences made
in a Western society under rule of law and operating under conditions of interme-
diate scarcity seems the only plausible one—even if in a world of plural values and
diverse cultures the Rawlsian search for an overlapping consensus may not suc-
ceed at all. In any event, to systematize particular realms of political convictions
will remain a worthwhile endeavor.
Whether or not one agrees in substance with Rawls, his focus on questions of
priority—as in his ‘priority of liberty’—in the WRE search for a Western ‘rule of law
society’ seems intuitively right (at least to Westerners like me). Yet, when acknowl-
edging path dependence, granting priority to certain claims vis a vis others will
be convincing only relative to some contingent base and some value judgments.
The acknowledgment of the relativity of the results of WRE search to a context
and a starting point is as essential for other ‘great questions’ than those addressed
by developing ‘a theory of justice’. The ‘great question’ of developing a theory of
rational choice is at least as fundamental if not even more so, yet as dependent
on contingent facts. As a paradigm of a reflective equilibrium search in such cases
Goodman’s discussion of induction may be used. Goodman very strongly empha-
sizes the role of entrenched practices and in that sense ‘relativity’ to factual con-
text, relevance of a starting point and path dependence. Intuitions formed on an
empty slate are sidelined as of minor relevance.
That Rawls thought of claims that concern merely fictitious situations origi-
nally as irrelevant should be noted, too. In case of RCT, claims derived from such
48 Ever since Meehl 1954 such second order evidence has been considered; see in a dierent
context impressively Tetlock 2015.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
152 |Hartmut Kliemt
A&K
fictions are at least of lesser force than those derived from real cases. The fiction
of a decision maker who has unlimited reasoning capacities may trigger certain
intuitions in us yet this ‘intuition pump’ will create a flow of specific claims that is
of minor relevance as compared to responses to real decision situations. It cannot
completely be ruled out—and certainly not on a priori grounds—that there is some
legitimate role for ideal a priori theory in indirectly shaping our WRE—as there is
a role for thought experiments as opposed to real experiments—yet it seems that
the relatively higher weight or the priority of claims should be with those derived
in real situations (the same way as real experiments trump thought experiments,
say, in physics).⁴⁹
6.2 Weighing the Evidence in Search for WRE
In search for an equilibrium, trade-os between what has to be given up and
what can be maintained as part of a coherent system of mutually supportive
claims must be made. All claims are treated as revisable. Yet, there are—at least
for the ‘time being’—certain fundamental principles guiding the process of weigh-
ing claims. For instance, we would on the most fundamental level not accept as
‘minimally rational’ the elimination of a well-entrenched (coherent with and
supported by many other claims) factual claim simply because it expresses an
inconvenient truth. In a similar vein, we would accept that ‘ought pre-supposes
can’ as bridging the gulf between the descriptive and the prescriptive in ways that
allow the elimination of claims that demand the impossible. Finally, and perhaps
most importantly, in an adequate equilibrium search claims that represent time
honored human practices have special weight. Yet again this does not mean that
they cannot in principle be revised and eliminated in a reflective equilibrium. It
cannot be ruled out a priori that intuitive particular or general claims that are not
supported by practices can successfully discharge the burden of proof against
claims supported by practices and established institutional facts.
In any event, the preceding tentative rules of weighingthe evidence and mak-
ing trade-os when seeking for a wide reflective equilibrium give a certain prece-
dence to the factual as opposed to the counter factual.⁵⁰ In particular, the accep-
tance of factual claims cannot directly be overruled by evaluative ones. If we have
49 So-called trolleology is another case in point similar to rationology in many respects. It may
be noted, though, that trolleology can give rise to interesting considerations concerning causal
relations to brain activity; see on this Greene 2013.
50 Merely imagined possibilities, prescriptive demands that something that does not exist is to
be brought about.. .
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |153
good reasons to believe that x is, but good reasons to desire that non-x should
be the case the latter should not aect our factual believe in x. If it comes to the
relation between representing and intervening the priority is with representing
and explaining. Interventions are planned and executed on the basis of factual
knowledge—or at least this should be so.
The WRE-search sets in where disciplinary science and explanatory RCT end.
Without turning to ‘fact-insensitive’ ideal RCT the search for WRE is emulating
empirical science eorts to reach coherence of general and singular ‘statements’
for purposes of assessing ‘normative’ matters. In doing so the search incorporates
‘witness-evidence’ or ‘intuitions’ in ways similar to how the testimony of scientists
is used in the critical assessment of scientific theories. Yet concerning normative
claims the testimony of witnesses is non-observational in that it is not merely used
to describe what witnesses think about what is ‘normatively adequate’. The testi-
mony of ‘representative individuals’ is rather used directly as a non-descriptive
normative premise to support normative standards or rules of correct conduct.
As Nelson Goodman states in Fact, Fiction, and Forecast (1983, 63 f., cited after
Hahn/Schlaudt 2016, 314):
“How do we justify a deduction? Plainly by showing that it conforms to the general rules
of deductive inference. [.. .] Yet of course, the rules themselves must eventually be justified.
But how is the validity of rules to be determined? Here again we encounter philosophers
who insist that these rules follow from some self-evident axiom, and others who try to show
that the rules are grounded in the very nature of the human mind. I think the answer lies
much nearer the surface. Principles of deductive inference are justified by their conformity
with accepted deductive practice. [.. . ] A rule is amended if it yields an inference we are
unwilling to accept; an inference is rejected if it violates a rule we are unwilling to amend.
The process of justification is the delicate one of making mutual adjustments between rules
and accepted inferences; and in the agreement lies the only justification needed for either.”
If the role of established practices along with de facto singular convictions is re-
garded as decisive as in Goodman (1983) and in Rawls’ (1951) Outline of a Decision
Procedure for Normative Ethics the dependence (‘relativity’) of normative justifi-
cations on factual convictions, aims, ends and values is acknowledged. Reflective
equilibrium contains a stylized account of practices that as a matter of fact are
prevailing. Yet the aspiration of the reflective equilibrium search goes beyond evi-
dence based empirics of prevailing practices and the opinions of what is and what
is not compliant with the ‘spirit’ of such practices. There is a clear view that the
reconstruction of what people would basically deem appropriate—according to
the de facto standards of practices and the ‘intuitions’ shaped by them—can after
critical assessment and ‘extrapolation’ yield general standards of what in a ‘local’
sense should be done in particular instances.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
154 |Hartmut Kliemt
A&K
With respect to developing prescriptive standards of rational action against
the background of descriptive RCT it should be noted that the process is sensitive
to the facts of general practices and specific judgments that accompany them. The
critically rational account of these ‘normative facts’ yields a concept of what is to
be deemed rational as opposed to what is rejected as irrational. It is dynamic in the
sense that the self-applicable ‘technological’ concept of rationality may develop
through time. It is like a living constitution of a community of mindful actors who
mutually recognize each other as ‘members of the club of boundedly rational ac-
tors’.
The underlying theories need not classify all intentional behavioras r ational—
qua intentionality as in Mises’ methodological dualism—nor is it necessary that
they rely on an ideal theory of rational choice—as in Selten’s methodological
dualism. They specify what is rational in a self-selection process. Such bootstrap-
ping may seem unsatisfactory in many regards yet in view of the list of preceding
rules of thumb it is at least not completely arbitrary.
6.3 WRE on Normative Standards of Morality
The broad discussion of the role and legitimacy of ideal theories in philosophy
that was triggered by a certain interpretation of the WRE search in Rawls’ (1971)
A Theory of Justice will not be discussed in any detail here. Suces it to note that
the role of a fact insensitive ideal theory that Rawls himself seems to allow for in
his exploration of ‘justice’ is not in line with the contextual interpretation of the
WRE search in Goodman’s and in Rawls’ own earlier account.
That philosophers who reject the continuity of philosophy and science and in
particular that of ideal RCT and boundedly rational practice were eager to jump on
the ideal theory train of Rawls’ theory of justice is unsurprising.⁵¹ Without claim-
ing to oer the only correct interpretation it seems to me, however, that an inter-
pretation of Rawls’ WRE-search in ‘a theory of justice’ is possible in ways that are
in line with the continuity view of science and philosophy and the principles of
contingent path dependent equilibrium search enunciated before. Then the Rawl-
sian search for WRE in developing his theory of justice is framed as anchored in
51 The ideal theory interpretation of Rawls is supported by the fact that Rawls uses in his WRE
rational choice models for justifying moral principles formulated in ideal RCT terms. Yet accord-
ing to the argument concerning the nature and significance of RCT oered here this will not work
in favor of ideal theory since ideal RCT itself is rejected in the account of rational practice oered
here.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |155
stylized accounts of socio-political practices of well working liberal Western legal
orders under conditions of ‘intermediate scarcity’.
The principles that emerge after several circles of the search and equilibrating
process can at best make a claim on those who endorse the underlying practices in
general. Such normative standards may be regarded as “relatively absolute abso-
lutes” (Buchanan 1999, vol. 1, 442–454). They are ‘absolute’ only relative to some
contingent facts. In that they are very similar to what Goodman required in giving
an account to inductive practices. For the present purposes it seems sucient to
note that a WRE on moral standards will be sought for along the same lines as that
on any other standards.
7Concluding Remarks
Selten’s ideal RCT systematizes what ‘we’ think about what is rational in search
for a reflective equilibrium. The basic intuitions that serve as springboard for this
search are not found by an empirical study of what ‘we’ are as a matter of fact
thinking, though. In his search, Selten conducts no high-level opinion poll, no
fact-finding mission concerning normative facts like established rules. Neither are
there elaborate descriptions of practices that might serve as paradigm examples
of what is conventionally regarded as rational according to human practices of
judgment. Quite to the contrary the initial ‘intuition material’ is ‘laundered’ and
driven to its extremes in an ideal theory equilibrium search in which empirical
issues of feasibility of behavioral recipes do not figure at all. As in the ideal theory
interpretation of Rawls’ theory of justice these aspects are eliminated by the a
priori assumption of behavioral compliance.
In a Selten type WRE the real practice dimensions would be eliminated to a
degree that alienates it from science. What Selten calls normative RCT appeals to
reason and aesthetic feelings as do certain mathematical theories. It spells out
what the human imagination may fabricate as ideal rationality. According to the
Humean view of this paper, it seems that such ideal rationality theory is too far
removed from actual human practice to have real relevance. It does not respect
principles like ought pre-supposes can and it does not even try to uphold continu-
ity of its substantive prescriptions as rooted in pre-equilibrium intuitions shaped
by human practices. Neither is it a Weberian ideal type representing reality in a
stylized yet approximable way.
As the experience with the ascent of evidence-based medicine shows, dis-
criminating between practical prescriptions that are merely speculative and those
for which tested evidence exists may be expected to lead to an improvement of
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
156 |Hartmut Kliemt
A&K
our practices. Though it is completely open to what extent all practices may be
evidence-based, arguments that assume away the relevance of the evidence-base
seem unconvincing. Ideal RCT that simply assumes away empirical limits to RCT
will not do as a substitute for evidence-based theories of what we deem rational
behavior. The demand for an a priori conception of rationality that pre-cedes all
practices (including what we as a matter of fact deem desirable in actor relative
ways) is no more such a theory than hunger is bread (see Bentham 1843) or that
the demand for an ideal moral theory will provide that theory.
Strict RCT leads to models that cannot explain since the underlying behav-
ioral assumptions are not approximately true. Despite this discontinuity with real
behavior, models based on RCT invite ascribing explanatory power to them even
in cases where they are not even a far cry of the real causal mechanisms of interac-
tive human choice behavior. By their inherent misrepresentation of the behavioral
facts, models based on ideal RCT tend to suggest mistaken technological recipes
leading to misguided interventions into real interactive processes. This model risk
can be avoided in principle by Selten type strict methodological dualism which in-
sists on the separation of pure a priori RCT and behavioral science a posteriori. Yet
the price for this is that the relationship to science becomes discontinuous and the
rationality embodied in science and its practices cannot be brought to bear on the
transcendental (v. Mises) or mathematical (Selten) a priori speculations of RCT.
As a way out, this essay indicates that going back to Nelson Goodman’s and
John Rawls’ original ideas on reflective equilibrium search in a non-a priori man-
ner may be useful. Goodman dealt with the problem of inductive reasoning by rely-
ing on real human practices and trying to systematize them in a critically rational
way open for all sorts of argument but insisting on some primacy of real human
practices. Rawls’ original ‘outline of a decision procedure for normative ethics’
along with his—not always consistent—insistence that the reflective equilibrium
we can meaningfully search for will always be rooted in particular practical expe-
riences under particular circumstances located in time and space is another case
in point. How exactly in such contingent wide reflective equilibrium searches sci-
entific results and scientific and non-scientific value judgments should be brought
to bear on each other is itself subject to a reflective equilibrium search on the re-
flective equilibrium search ‘method’. Not much about this can be said except that
it would seem highly desirable and that ‘absorbability’ of some kind would seem
a desirable property of such a theory of WRE.
That is, almost tautologically, such a conception should not subvert its own
prevalence as an equilibrium conception if it spreads. If the continuity thesis is
taken seriously then (‘ultimately’) the absorption of such theories could itself be
scrutinized in broadly empirical terms.
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |157
Acknowledgment: Alluding to Robbins 1935 by the title of what started as a con-
tribution to the ‘Ethics and Economics’ conference at the Singapore University
of Technology and Design (SUTD) January, 19–20, 2017 is, of course, intended;
I am grateful to conference participants for their comments, in particular, Dan
Hausman and Zsombor Meder. Extremely helpful were later critical discussions
by Michael Baurmann and Anton Leist who induced me to focus on RCT as an
ideal theory and its relation to science rather than on ‘moral science’ more gener-
ally. They rightly insisted that I be more explicit on the distinction between Max
Weber’s ideal types and Selten type ideal RCT and my ‘old fashioned’ use of the
reflective equilibrium concept as expressive of the ‘methodological’ continuity
of philosophy and science. If I am wrong, I will at least be more clearly wrong
after their intervention. Sarah-Lea Eert provided some very helpful specific dis-
cussion of ideal theories as did the participants of a conference in honor of Alan
Hamlin at the university of Manchester a few years ago. All this made me aware
of my still insucient grasp of an extended discussion and literature concerning
ideal theory to which I here add, if anything, a critical rationalist ‘Albertian’, Al-
bert 1985, account of the perspective of ideal RCT which in a way may be the most
fundamental ideal theory of all. Last but by far most importantly I should like to
mention that my thinking on approximate explanations has been fundamentally
influenced by discussions concerning idealization and approximate explanation
with a younger Albert. See Albert, M./H. Kliemt (2017), Infinite Idealizations and
Approximate Explanations in Economics, MAGKS Working Paper, URL: https://
www.uni-marburg.de/fb02/makro/forschung/magkspapers/paper_2017/26_2017
_albert.pdf.
References
Albert, H. (1985), Treatise on Critical Reason, Princeton
Amadae, S. M. (2016), Prisoners of Reason: Game Theory and Neoliberal Political Economy,
reprint edition, New York
Baurmann, M. (1987), Zweckrationalität und Strafrecht, Opladen
Bentham, J. (1843), Anarchical Fallacies, vol. 2 of Bowring (ed.), works
Berninghaus, S./W. Güth/H. Kliemt (2003), From Teleology to Evolution. Bridging the Gap be-
tween Rationality and Adaptation in Social Explanation, in: Journal of Evolutionary Eco-
nomics 13, 385–410
Binmore, K. (1987/88), Modeling Rational Players I & II, in: Economics and Philosophy 3 & 4,
179–214 & 9–55
Brennan, G./P. Pettit (2005), The Feasibility Issue, in: The Oxford Handbook of Contemporary Phi-
losophy, ed. by Frank Jackson/Michael Smith, Oxford, 258–279
Carnap, R. (1956), Meaning and Necessity, Chicago
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
158 |Hartmut Kliemt
A&K
Gul, F./W. Pesendorfer (2008), The Case for Mindless Economics, in: Caplin, A./A. Schotter (eds.),
The Foundations of Positive and Normative Economics: A Handbook, Oxford–New York, ch. 1
Daniels, N. (1979), Wide Reflective Equilibrium and Theory Acceptance in Ethics, in: The Journal
of Philosophy 76, 265–282
Geuss, R. (2008), Philosophy and Real Politics, Princeton
Goodman, N. (1983), Fact, Fiction, and Forecast, 4th rev. ed., Cambridge/MA
Greene, J. (2013), Moral Tribes: Emotion, Reason, and the Gap Between Us and Them, New York
Güth, W./H. Kliemt/B. Peleg (1999), Co-Evolution of Preferences and Information in Simple Game
of Trust, in: German Economic Review 1, 83–110
— (2000), Evolutionarily Stable Co-Operative Commitments, in: Theory and Decision 49, 197–221
— (2004), Bounded Rationality and Theory Absorption, in: Homo Oeconomicus 22, 521–540
— (2007), The Rationality of Rational Fools, in: Peter, F./H. B. Schmid (eds.), Rationality and Com-
mitment, Oxford, 124–149
—/W. Leininger/G. Stephan (1991), On Supergames and Folk Theorems: A Conceptual Analysis,
in: Selten, R. (ed.), Game Equilibrium Models. Morals, Methods, and Markets, Berlin, vol. 2,
56–70
Hahn, S. (2004), Reflective Equilibrium: Method or Metaphor of Justification?, in: Löer, W./P.
Weingartner (eds.), Wissen und Glauben, Wien, 237–243
—/O. Schlaudt (2016) (eds.), Logic, Morals,Measurement—Origins and Justification of Norms, in:
Analyse & Kritik 38, 311–316
Hamlin, A./Z. Stemplowska (2012), Ideal Theory and the Theory of Ideals, in: Political Studies
Review 10, 48–62
Hart, H. L. A. (1961), The Concept of Law, Oxford
Hausman, D. M. (1992), The Inexact and Separate Science of Economics, Cambridge
Hicks, J. (1979), Causality in Economics, Oxford
Hobbes, T. (1968[1651]), Leviathan, Harmondsworth
— (1990[1682]), Behemoth, Chicago
Hume, D. (1985), Essays. Moral, Political and Literary, Indianapolis
Kahneman, D. (2012), Thinking, Fast and Slow, London
Kant, I. (1977[1798]), Die Metaphysik der Sitten, Vol. 8, Immanuel Kant Werkausgabe, Frankfurt
— (1987), The Reason of Rules and the Rule of Reason, in: Critica 29, 43–86
— (2009), Philosophy and Economics 1, München
— (forthc.), ABC—Austria,Bloomington, Chicago. Political Economcy the Ostrom Way, in: The Aus-
trian and Bloomington Schools of Political Economy. Advances in Austrian Economics, vol.
22, 15–47
Külpe, O. (1897), Introductionto Philosophy a Handbook for Students of Psychology, Logic,Ethics,
Æsthetics and General Philosophy, New York
Leonard, R. (2010), Von Neumann, Morgenstern, and the Creation of Game Theory: From Chess to
Social Science, 1900–1960, Cambridge
McKenzie, R./G. Tullock (1978), The New World of Economics: Exploration into the Human Experi-
ence, Nueva York
Mises, L. von (1966[1949]), Human Action, Chicago
Morgenstern, O. (1928), Wirtschaftsprognose: Eine Untersuchung ihrer Voraussetzungen und
Möglichkeiten, Wien
Parsons, T. (1968), Utilitarianism. Sociological Thought, in: Sils, D./R. K. Merton (eds.), Interna-
tional Encyclopedia of Social Sciences, New York–London, 536–547
Rawls, J. (1951), Outline of a Decision Procedure for Ethics, in: Philosophical Review 60, 177–190
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
A&K
On the Nature and Significance of (Ideal) Rational Choice Theory |159
— (1971), A Theory of Justice, Oxford
Robbins, L. (1935), An Essay on the Nature and Significance of Economic Science, London
Schlick, M. (1986), Die Probleme der Philosophie in ihrem Zusammenhang: Vorlesung aus dem
Wintersemester 1933/34, Frankfurt
Selten, R. (1999), Reply to K. Shepsle, in: Alt, J./M. Levi/E. Ostrom (eds.), Competition and Co-
operation: Conversations with Nobelists about Economics and Political Science, New York,
303–308
Spinoza, B. de (1951[1670]), A Theologico-Political Treatise. A Political Treatise, New York
Steinberg, J. (2013), Spinoza’s Political Philosophy, in: Stanford Philosophical Encyclopedia
Strawson, P. F.( 1962), Freedom and Resentment, in: Proceedings of the British Academy 48, 187–
211
Tetlock, P. E. (2009), Expert Political Judgment: How Good Is It? How Can We Know?, Princeton
— (2015), Superforecasting. The Art and Science of Prediction, New York
Ypi, L. (2010), On the Confusionbetween Ideal and Non-ideal in Recent Debates on Global Justice,
in: Political Studies 58, 536–555
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
Angemeldet | hartmut.kliemt@t-online.de
Heruntergeladen am | 17.06.18 17:01
