Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
DIPARTIMENTO DI SCIENZE ECONOMICHE AZIENDALI E STATISTICHE
Via Conservatorio 7 20122 Milano
tel. ++39 02 503 21501 (21522) - fax ++39 02 503 21450 (21505) http://www.economia.unimi.it
E Mail: [email protected]
In pubblicazione su Economics for Real: Uskali Maki and the Place of Truth in Economics, a cura di Aki Lehtinen, Jaakko Kuorikoski e Petri Ylikoski (Routledge, 2012).
ARE PREFERENCES FOR REAL?
CHOICE THEORY, FOLK PSYCHOLOGY,
AND THE HARD CASE FOR COMMONSENSIBLE REALISM
FRANCESCO GUALA
Working Paper n. 2011-18
SETTEMBRE 2011
1
Are Preferences for Real?
Choice Theory, Folk Psychology, and the Hard Case for Commonsensible Realism*
Francesco Guala†
Abstract
According to “realists about commonsensibles” like Dan Hausman and Uskali Mäki, the
existence of the theoretical entities of economics is a fact that no one should seriously doubt.
But commonsensible realism is an unstable philosophical position, with a tendency to
collapse into forms of behaviourism. In fact, behaviourism may turn out to be the only
defensible interpretation of rational choice theory that avoids explicit reference to
unobservable theoretical entities. The price to pay for this return to the old orthodoxy
however is to deny that preferences have a causal role in the explanation of action, and to
severe the economic theory of choice from research in psychology and cognitive science.
JEL Classification: B41, D01, D03
Keywords: preferences, psychology, rational choice theory, realism
* I thank Anna Alexandrova, Dan Hausman, Jaakko Kuorikoski, Caterina Marchionni, Ivan Moscati, and
Michiru Nagatsu for clarifications, comments and suggestions on earlier drafts of this paper. The usual caveats
apply.
† Department of Economics, Business and Statistics, University of Milan, via Conservatorio 7, 20122 Milan,
Italy. Email: [email protected]
2
1. Introduction
“Commonsensibles” are entities – like tables, chairs, flowers, and birds – that belong to our
ordinary view of the world. Most contemporary philosophers – and ordinary people, for that
matter – are realist about commonsensibles: if we know anything at all, surely we must know
that tables and chairs exist. Commonsensibles are epistemically primary, in the sense that they
constitute the basic building blocks upon which all knowledge of the external (and, perhaps,
internal) world is constructed. This does not mean that intuitions about commonsensibles are
incorrigible of course. But the existence of commonsensibles seems to raise epistemic and
ontological problems that are less urgent and complicated than those raised by the strange
unobservable entities posited by the most advanced natural sciences.
Realism about commonsensibles in this broad sense is also the starting point of an argument
aimed at securing a more specific claim in the philosophy of social science. In this second
technical sense, “realism about commonsensibles” – a neologism introduced by Uskali Mäki
– holds that because economic theories refer exclusively to entities that belong to our ordinary
worldview, the existence of the theoretical entities of economics is a fact that no one (except
the radical sceptic) should seriously doubt. An important corollary of this argument is that
problems of scientific realism in economics either do not arise, or take a different shape from
the one that is at stake in the natural sciences.
Realism about commonsensibles has had an easy ride so far. A notable exception is Kevin
Hoover’s (1995) paper (“Is Macroeconomics for Real?”), whose title I gratefully borrow in
this essay. Hoover points out that the variables of macro-economic models do not refer to
“ordinary” economic entities like prices, employees, or firms. They are rather complex
technical constructs that bear only an indirect relation and a vague resemblance to the familiar
entities of “folk” economics. For this reason, Hoover argues, the technical constructs of
economics raise all the standard epistemic and ontological concerns of directly unobservable
entities like electrons, molecules, and black holes.
In this chapter I will push this line of inquiry into the realm of the micro. For realism about
commonsensibles surprisingly has not been challenged where it is intuitively least plausible,
namely, the case of microeconomic posits like preferences and beliefs. As Hausman remarks,
3
It seems absurd to maintain that economics does not refer to unobservables. Surely the
preferences and expectations that explain and predict choices are unobservables. Who
has ever seen or smelt a preference? Who has ever tasted a belief? (Hausman 1998: 196)
And yet there is a sense in which preferences and beliefs belong to our ordinary worldview,
for we constantly use them (and related concepts, like “desire”) to explain our actions and
those of fellow human beings. This idea is not new. Already a century ago, economists like
Ludwig von Mises noticed that the explanatory strategy of economics is on a par with
ordinary explanations of human action. And the source of Donald Davidson’s famous qualms
with decision theory in the 1960s was the latter’s continuity with folk psychology.1
Following a similar line of reasoning in the 1970s Alexander Rosenberg launched a sustained
attack on the scientific pretension of economics (cf. Rosenberg 1976). As Rosenberg has put
it in a later monograph, “the fundamental explanatory strategy of economic theory is of a
piece with that of our ordinary explanations of human action. […] Economics proceeds by
formalizing commonsense explanations of action into a theory of rational choice” (Rosenberg
1992: 118). Like Davidson, Rosenberg used the continuity thesis as a key premise in an
argument aimed at challenging the scientific credentials of economics: because of this
continuity, economic theory inherits many defects and limitations of folk-psychological
explanations of action. In particular, it is impossible to identify the factors that cause
deviations from the predictions of rational choice theory, measure their influence, and
incorporate them into a more accurate theory of choice. According to Rosenberg economic
theory is non-improvable in the way genuine scientific theories are supposed to be.2
At this point, it is tempting to conclude the argument in an eliminativist fashion: being part of
a false and scientifically hopeless theory, it is unlikely that the theoretical terms of economics
refer to anything at all. But although eliminativism is a prominent position in the philosophy
of mind (Churchland 1988), it is rather unpopular in the philosophy of economics. Rosenberg
(1992: 140-148) for example does not deny that the entities of folk psychology (like
1 See the essays in Davidson (1980). Other influential philosophers who endorse the continuity thesis are David
Lewis (1974) and Philip Pettit (1991). 2 For a critical perspective on Rosenberg see Nagatsu (2010).
4
preferences and beliefs) exist. Although he doubts that they play an important causal role in
human action, he claims that we know them by introspection. Similarly, Mäki claims that
commonsensibles are “experienced” in a broad sense, by means of “introspection, inference,
interpretation, culturally established meanings” (1999: 248). Thus the continuity thesis is used
to support a realist, instead of an eliminativist, argument. The argument goes as follows:
(1) All the entities posited in rational choice theory belong to folk psychology.
(2) The ontology of folk psychology does not include dubious theoretical entities.
(3) Therefore, there is no reason to doubt that the entities posited by rational choice
theory exist.
Versions of this argument have been put forward by Dan Hausman and Uskali Mäki in
various books and articles (e.g. Hausman 1992, 1998; Mäki 1996, 1998), which contain the
canonical formulation of realism about commonsensibles and will constitute the primary
target of this chapter.3 Clearly eliminativists and commonsensible realists agree about premise
1 but disagree about 2 and a fortiori also about 3.
In this paper I will challenge some of Hausman’s and Mäki’s arguments for commonsensible
realism. I will argue that commonsensible realism is an unstable philosophical position, with a
tendency to collapse into forms of behaviourism (such as revealed preference theory) that
deny the continuity thesis. In fact, behaviourism may turn out to be the only defensible
interpretation of rational choice theory that avoids explicit reference to unobservable
theoretical entities. But the price to pay for this return to the old orthodoxy is to deny that
preferences have a causal role in the explanation of action, and to severe the economic theory
of choice from research in psychology and cognitive science.
Here is the game plan: in the next section I outline the main features of realism about
commonsensibles and put it in the context of the classic debate on scientific realism in
philosophy of science. In section 3 I discuss Hausman’s and Mäki’s attempts to replace the
classic distinction between observable and unobservable entities with a distinction based on
3 It should be mentioned at the outset that even though I shall focus mainly on their similarities, Hausman’s and
Mäki’s views differ in a number of respects, many of which concern the issue of realism in economics. A good
overview of these differences can be found in Hausman (1998, 2000a) and Mäki (2000).
5
“old” (commonsensical) vs. “new” (scientific) entities, and reject it on historical and
conceptual grounds. Section 4 includes an analysis of the notion of preference that is implicit
in the formal framework of rational choice theory, showing that preferences are not entities
but either relational or dispositional properties. In section 5 I examine an attempt to
demonstrate that preferences are accessible to direct observation by means of introspection. In
section 6 I argue that the success of this strategy depends on the degree of homogeneity of the
causal bases of preferences, and suggest that there is no ground for optimism. Section 7
concludes with general comments on the status of commonsensible realism and its
relationship with scientific projects in the field of decision theory.
2. Scientific realism and realism about commonsensibles
Commonsensible realism belongs to a family of doctrines that emerged as a reaction to the
decline of logical positivism and of the unity of science approach in the 1970s and 1980s. By
the1980s several philosophers had begun to doubt that “special sciences” like economics raise
exactly the same philosophical problems as physics (the “fundamental science”). One issue
that came under critical scrutiny in the philosophy of economics was the problem of scientific
realism. Realists and anti-realists disagree about the epistemic and ontological status of
theoretical entities in science. Scientific realists hold that we have good reasons to believe that
the basic furniture of the world is approximately as posited by our best scientific theories. An
up-to-date list of “what there is” – or, more precisely, “what we have reason to believe there
is” – should include not only the macro-objects of everyday life (tables, trees, and birds) but
also atoms, electrons, black holes, DNA, fields of force, and all the peculiar entities that are
posited by the most advanced sciences.
Modern anti-realists disagree. While ordinary objects are accessed by direct sensory
experience, the posits of many scientific theories – especially in the realm of physics – can
only be inferred from data generated by complex instruments of observation. Since these
inferences rely on theoretical assumptions, we have no theory-free access to what lies beyond
the realm of ordinary experience. For a variety of reasons (which I am not going to review
here – but see van Fraassen (1980) and Laudan (1981) for a sample of classic arguments, and
6
Psillos (1999) for a critical review) anti-realists claim that we should better be agnostic or
plainly atheists on unobservable posits.4
Scientific realists and anti-realists draw a sharp distinction between phenomena that are
directly observable by unaided sensory experience, and those that are not. The problem of
scientific realism applies to the latter only, and therefore (trivially) cannot emerge in those
disciplines that do not posit theoretical unobservables. The key question then is: does
economic theory posit unobservable theoretical entities? The direct route to establish the
irrelevance of scientific realism for economics is clearly to answer “No”. Some formulations
of commonsensible realism in fact were part of a general strategy aimed at emancipating the
philosophy of economics from the philosophy of physics, cutting a conceptual space where
questions tailored on the specific character of economic science could be debated free from
the conceptual constrains inherited from traditional philosophy of science. If no genuine
unobservable entities are posited by economic theory, then clearly the classic problem of
scientific realism does not arise in economics. If there is an issue of realism in economics – as
many economists and philosophers claim – then it must be a different kind of issue, requiring
a different approach and different solutions from those found in the standard philosophy of
science literature. The next step consisted in shifting the focus of attention from the existence
of unobservables to idealization, abstraction, or – to use another Mäkian neologism – the
“realisticness” of economic theories.
The relevant issue lies elsewhere, namely in how the ordinary objects are represented in
economic theories. Even if the referents of ordinary experience and economic theories
are pretty much the same, the ways in which they are represented often dramatically
diverge. (Mäki 1996: 434)
4 There are alternative ways of drawing the realist/anti-realist distinction, for example focusing on the goals of
science: while realists hold that science aims at discovering the truth about the (mind-independent) structure of
observable and unobservable reality, anti-realists deny that this is possible. The latter formulation has the
advantage of relieving realism from any commitment about factual matters (like the accuracy of current
scientific theories). As a matter of fact however most realists do believe that the goals of science have already
been achieved (to a good degree of approximation) in some areas of science – hence the formulation given in the
main text.
7
In any scientifically relevant sense of ‘direct testing’ or ‘direct observing’, many of the
behavioural postulates of economics are directly testable. The problem with claims such
as ‘People’s preferences are transitive’ or ‘Firms attempt to maximize profits’ is not that
they are untestable, but that they are apparently false. (Hausman 1992: 161)
The observable/unobservable distinction paves a straightforward line of argument towards the
conclusion that scientific realism is irrelevant for economics. Only one step would remain to
be secured, namely the claim that economic theory does not introduce unobservable entities
(assumption (2), in the outline of the argument given earlier). As we shall see, proving this
claim is not so easy, and for this reason perhaps commonsensible realists appear at time to
hesitate. In particular, they sometimes suggest that the real issue in scientific realism is not
observable versus unobservable, but familiar versus unfamiliar, or new versus old entities.
Since I do not believe this move to be legitimate, I will spend the next section to analyze and
criticize it. This is a detour with respect to the main line of argument, which will be resumed
in section 4 – so the readers who do not want to be delayed may jump straight there and
ignore section 3.
3. What’s so special with “old” posits?
In some of their writings, Hausman and Mäki present the issue of scientific realism as based
on a distinction between “old” and “new” entities (the entities of commonsense vs. the entities
of science) rather than observable and unobservable entities. Mäki for example states that
preferences and expectations, strictly speaking, are unobservable in the sense that direct
sense experience about them is not possible. On the other hand, they are familiar folk
psychological entities, articulated in a peculiar manner in standard economic theory and
decision theory. (Mäki 1998: 306)
Similarly, according to Hausman,
anti-realists seek to draw a line between the relatively unproblematic claims of everyday
life and the problematic theoretical posits of science. Physics postulates new
unobservables, to whose existence commonsense realism does not commit us. Although
8
economics refers to unobservables, it does not, in contrast to physics, postulate new
ones. Its unobservables – beliefs, preferences, and the like – are venerable. They have
been a part of commonsense understanding of the world for millennia. (Hausman 1998:
197-8)
This way of drawing the distinction between problematic and unproblematic entities,
however, is dubious both for historical and for conceptual reasons. Historically, the relevant
divide in the controversy about scientific realism has never been new versus old, but
observable versus unobservable by unaided perception. I am not sure where commonsensible
realists find inspiration for their restatement. The only possibility that comes to mind are a
few paragraphs in The Scientific Image where Bas van Fraassen – the philosopher who more
than anyone else has contributed to define the coordinates of the contemporary debate on
scientific realism – seems to speaks as if the introduction of “new” scientific entities was the
real issue at stake.5 This way of drawing the boundary between problematic and
unproblematic entities however has been explicitly rejected by van Fraassen in later writings
(cf. Ladyman et al. 1997) so the new/old distinction is groundless from a historiographic point
of view. If this is the problem of realism that does not arise in economics, then the claims of
commonsensible realists are based to a large extent on equivocation.
It is possible, of course, that the old/new distinction is better than the
observable/unobservable one, quite independently from its historical pedigree. But I am
sceptical: what is so special about folk-psychological entities, that makes them exempt from
the anti-realist challenge? Hausman argues that
an antirealist about economics must be a radical skeptic. She must deny that she can
know that her son prefers chocolate ice cream to vanilla or that her aunt believes that
5 See the discussion of ordinary abductive reasoning at pp. 20-22 of The Scientific Image (1981), in the famous
“mice” example. Here van Fraassen seems to suggest that the application of abductive reasoning in ordinary
cases is legitimate because the conclusions derived do not imply a commitment to the existence of a new kind of
entity (e.g. mice), whereas in scientific reasoning abductive inference often leads to claims that refer to
unfamiliar theoretical entities like electrons, DNA, and quarks. See also Psillos (1999: 212-215) for a critical
discussion.
9
airplanes fly. Those who hold that people can know about things like this are realists
about beliefs and wants. (1998: 198)
Hausman concedes that in the context of specific philosophical inquiries one can legitimately
consider mental states and other folk-psychological notions problematic (ibid.). However, he
also points out that this move would have devastating consequences for the legitimacy of
common sense explanations of action – so devastating, indeed, that it would lead to radical
scepticism.
This sort of reductio (the consequences of X are so terrible that a reasonable person cannot
accept X) has been used by other fans of the continuity thesis. According to Jerry Fodor, for
example, the (hypothetical) rejection of commonsense intentional psychology would be “the
greatest intellectual catastrophe in the history of our species” (1987: xii). But of course one
thing is to dislike the implications of a given claim or belief, quite another is to demonstrate
that it is false. So defenders of anti-realism about folk-psychological entities have not been
easily impressed by the “catastrophe argument”.6
A common move in philosophical debates at this point consists in reviewing the historical
record of commonsensical entities that have been expelled from the scientific worldview.
Folk intuitions about the movement of physical objects for example are consistent with the
existence of an internal force or “impetus” that is gradually extinguished as the object changes
location. This intuitive idea is well entrenched in our intuitive interpretation of the physical
world, and was once held even by esteemed scientists like Aristotle and Galileo. The fact that
the impetus theory is “old” or commonsensical however has not prevented its rejection in
light of robust experimental evidence to the contrary.7
6 Stich (1996) includes a good critical discussion of the catastrophe argument. 7 Other parts of folk physics – such as the intuition that an object moving in a circle departs from its orbit
following a curved trajectory, if the centripetal force disappears – have had a similar fate (see McCloskey 1983
for these and other examples). The same moral applies to once familiar but now scientifically discredited entities
like “evil spirits”, “vital forces” or, indeed, the soul. Belief in witchcraft and evil spells is still widespread in both
economically prosperous and developing countries. Yet a would-be scientific theory based on “refinements” of
such concepts would not be considered epistemically privileged merely because it incorporates such “venerable”
entities. (In fact it would lose much of its credibility because of that.)
10
Would the demise of folk action theory be more devastating than the rejection of impetus
theory? On the one hand one may point out that in the case of impetus the revision prompted
by Newtonian physics was circumscribed to a relatively small set of phenomena, whereas
abandoning folk psychology would revolutionize our understanding of ourselves and others,
and undermine many entrenched social practices.8 On the other hand, desire/belief
explanations are not the whole of folk psychology, for we routinely explain behaviour
appealing to factors like emotions, individual character, history, or social context, which are
not easily accommodated in the belief/desire scheme. So although abandoning preference-
based explanations would certainly have important moral consequences (think about ordinary
attributions of responsibility and autonomy), it does not necessarily imply the invention of a
new Weltanschauung.
Not only it seems perfectly possible (logically and psychologically) to abandon some
commonsensible beliefs concerning the human mind; the evidence suggests that this has
already happened. Historians and anthropologists tell us that folk-psychological theories can
vary across cultures, and change through time. Scholars of ancient Greece like Bruno Snell
(1953) for example have argued persuasively that Homeric poets lacked a unified concept of
the mind such as the one that later became prevalent in Western culture. If Snell is right, then
it is possible for people living in different cultural environments to endorse different
conceptions of the aetiology of action, while sharing other ontological beliefs about the world.
Surely the moral principles that governed Homeric society (e.g., attributions of moral
responsibility) were very different from ours. And yet, moral relativism does not imply
epistemic relativism on a grand scale. Presumably the Homeric poets believed in trees and
birds just like us, so rejecting a folk-psychology does not imply universal scepticism about
ordinary ontology either.9
So, all things considered, the direct route to commonsensible realism seems preferable. The
claim that economic theory does not posit unobservables, if true, would establish the
irrelevance of scientific realism. But is it true? In the next section I begin to examine that
claim, starting from a detailed analysis of the posits of rational choice theory.
8 Dan Hausman, in private correspondence. 9 For other examples of cultural variation in folk psychology, as well as philosophical discussion, see Kusch
(1999).
11
4. What preferences are (not)
The theory of rational choice is the basic building block of microeconomic theory and, to the
extent that it is used in macroeconomic models, of neoclassical economics in general. An
idealized rational decision maker is defined by a set of axioms that impose consistency
requirements on its preferences and beliefs. Preferences must satisfy transitivity,
independence, and completeness conditions, for example, while beliefs ought to satisfy
Kolmogorov’s axioms of probability. While there is general agreement on the formal
properties of this idealized model, its interpretation – as it often happens in science – is a lot
more controversial.
One murky issue is the nature of preferences.10 Some textbook presentations of choice theory
still reflect the behaviourism that became fashionable in economic and psychology during the
1930s and 40s (when choice theory was systematized). The seminal “revealed preference”
interpretation of choice theory proposed by Paul Samuelson (1938) was originally meant to
do away with all psychological concepts, and identified preferences with consistent patterns
of observable choices. But the decline of behaviourism, together with sustained conceptual
criticism, have led over the years to temper this radical interpretation,11 to the point that
nowadays “revealed preference theory” for most economists is no more than an empiricist
commitment to derive demand curves from choice data, or to construct models that can be
tested against observable behaviour.
This is entirely compatible with a psychologistic interpretation of preferences, as stable
entities or properties that cause and explain the choices of economic agents, but are not
identical with such choices. Consider an ordinary expression like “I’ve chosen salmon rather
than fois gras because I prefer fish to meat”. If we were to translate it into “I’ve chosen
salmon rather than fois gras because I choose fish rather than meat”, we would clearly betray
10 For simplicity, I will focus on preferences throughout the paper, but most of what I say can be extended to
beliefs. 11 See e.g. Sen (1993), Hausman (2000b, 2011). Although the post-behaviourist rehabilitation of psychological
preferences has become widely accepted in philosophy, there are also dissenters who defend the superiority of
the revealed-preference interpretation of rational choice theory (e.g. Ross 2005).
12
its meaning. The original sentence is offered as an explanation of my behaviour: my current
beliefs (that duck liver is the main ingredient of fois gras, say, and that salmon is fish) and my
tastes (I don’t like meat) prompted my choice of salmon from the options available in the
menu. Beliefs and desires under this reading are the psychological causes of behaviour
without being themselves identical with behaviour.
Commonsensible realism is consistent with the post-positivist move toward psychologism in
rational choice theory, for it starts from the premise that there is no ontological gap between
folk psychology and rational choice theory. Preferences are formal refinements of desires;
since desires are psychological (mental) entities, realism about commonsensibles seems at
odds with a strict behaviourist interpretation of choice theory. Overcoming the behaviourist
ban on psychology however may create more problems than it solves. What are these
“psychological preferences” exactly? How do we know what they are (i.e. their properties)?
And how do we know that they exist?
Commonsensible realists – and many others, for that matter – often speak of preferences as if
they were entities. The very distinction between the existence of preferences and their
theoretical description (in the axioms of rational choice) suggests that there is some thing (an
entity or set of entities) that the axioms of the theory (transitivity, independence, etc.) describe
somewhat inaccurately. But this interpretation is problematic. Of course we do say things like
“I have a preference for fish over meat”, or “she has weird preferences” which suggest a
commitment to the existence of entities called preferences. But philosophers have known for a
long time that literal interpretations of ordinary language expressions are perilous. Consider
the following analogy: a sentence like “Paul gave Helen a kiss”, seems to posit the existence
of Paul, Helen, plus an entity (a kiss) that somehow “travels” between Paul and Helen. As
Bertrand Russell taught us, however, a little bit of paraphrasis can eliminate the implausible
commitment to the existence of kisses. “Paul kissed Helen”, for example, is a synonymous
but ontologically more parsimonious expression that only commits us to the existence of two
individuals engaged in a certain activity (kissing). So “I have a preference for chocolate” does
not necessarily imply a commitment to the existence of a set of entities called preferences,
any more than “Paul gave Helen a kiss” implies a commitment to the existence of kisses.
13
The refinement of folk psychology into rational choice theory involves, among other things,
paraphrases of this kind. In the theory of rational choice, preferences are represented as
relations between the objects of choice (options, states of the world, consequences, or bundles
of goods). The axioms of rational choice impose restrictions on the range of legitimate
preference relations, by stating for example that x must be preferred to z, if x is preferred to y
and y is preferred to z:
(x > y) & (y > z) → (x > z) [transitivity axiom]
What sorts of entities are posited by propositions of this kind? If we make the reference to the
decision-maker explicit by means of an index, it emerges clearly that all the relata of the
preference relation are observable entities: x >i y means that individual i prefers the bundle (or
state of the world) x to the bundle y. The preference relation, rather than the related entities, is
the only potentially unobservable term here.
So what is the preference relation? One possibility is to stop the analysis at this level and say
that preferences are mere correlations among choices, or accounting devices that we use to
keep track of patterns of behaviour. This is basically the behaviourist interpretation of choice
theory, which however, as we have seen, seems to be at odds with the spirit of realism about
commonsensibles. Alternatively, one can opt for a more substantial interpretation of the
preference relation: to prefer x to y is to have a propensity or disposition to choose x, if and
when certain enabling conditions occur. (Among these conditions, for example, x and y must
be available as objects of choice, the agent must know that they are available, and she must
not believe that there is a third option z that is preferred to both x and y.)
Notice again that this propensity is the property of an individual (the decision maker) so this
interpretation does not introduce any unobservable entities yet. Does it introduce an
unobservable property? Dispositional properties are tricky entities, raising a host of
metaphysical puzzles. If a disposition only referred to the fact that a certain state or event (i
choosing x, say) could become actual in some counterfactual circumstance, then clearly the
dispositional property would be observable. To say that Ann prefers tea to coffee would be no
more mysterious than to say that salt is soluble or that glass is breakable: it is directly testable
14
by creating circumstances in which the disposition can become manifest, as experimental
economists and psychologists do all the time.
However this would challenge the continuity thesis: preferences-as-propensities would not be
the causes of choice as in folk-psychological explanation. And similarly the fragility of a
glass would not be among the causes of its breaking if I drop it on the floor (which is
counterintuitive). The only way to save the explanatory power of preference-based
explanations and the continuity with folk psychology is to identify preferences-as-propensities
with whatever it is that makes i choose x in a certain range of circumstances. Presumably an
individual has the propensity to choose x instead of y across various circumstances in virtue of
the existence of a set of mechanisms that – once triggered – cause her to choose x. (Similarly,
salt dissolves in water in virtue of a series of chemical processes triggered by mixing NaCl
with H2O molecules, that cause the dissociation of NaCl into ions.). Reference to a
dispositional property then must stand for a more complicated story involving entities,
properties, and causal mechanisms that we know only imperfectly, or the details of which are
not important for our present explanatory purposes.
These entities, properties, and mechanisms are known in philosophical jargon as the “causal
bases” of dispositions. The identity of dispositions and their causal bases is a contested thesis
in metaphysics (see e.g. Mumford 1998), but for our purposes it can be established by means
of the following argument: if preferences are dispositions, then either these dispositions are
identical with their causal basis, or preferences do not cause behaviour. But preferences do
cause behaviour (according to folk psychology), thus as dispositional properties they must be
identical with their causal bases (if the continuity thesis is to be preserved).
5. The role of introspection
If dispositional properties are identical with their causal bases, the issue of scientific realism
in economics depends crucially on the possibility to directly observe such mechanisms by
means of unaided sensory experience. But the identity thesis puts us in a difficult position: the
causal basis of preferences is constituted by cognitive processes that lie hidden in the human
mind (or brain, depending on one’s taste). A tempting solution is to appeal to a special type of
direct observation that is only available in the case of mental states: introspection. Before the
15
heyday of behaviourism introspection was generally considered a legitimate source of data for
psychology, and was even considered for a long time the chief method of validation for
economic theory. 12 The most famous and influential attempt to put introspection at the
foundations of economic science can be found in the writings of John Stuart Mill (1836).
Since Mill’s methodological views have been recently revamped in the philosophy of
economics – notably by Hausman (1992) among others – it is natural to ask whether
introspection can be used to rescue commonsensible realism. Moreover, as we have seen in
section 1, both Rosenberg and Mäki invoke it explicitly as a source of knowledge concerning
preferences and beliefs.
That in many circumstances we have direct experience of the causes of our own behaviour is,
I suppose, beyond doubt. Consider cravings and revulsions. If I see a tasty sausage, I salivate
and feel a strong urge to eat it. But when I see a cooked snail, in contrast, I have reaction of
disgust. Offered a choice between a snail and a piece of sausage, I politely say that I prefer the
latter. In a case like this my craving for sausages and my aversion to snails are made vivid by
two bodily reactions that are directly accessible and that – barring crazy scientists and “brain-
in-a-vat” scenarios – I do not seriously doubt. So if my disposition to choose sausages rather
than snails involves cravings and revulsions like these, it is correct to say that I have direct
access to this kind of preference.
Notice that this does not mean that every aspect of the causal basis of my preference for
sausages is directly accessible. There are obviously several aspects of my cravings,
revulsions, feelings, emotions, etc. that are not accessible by introspection. Suppose for
example that, as the latest neuroscientific evidence suggests, the insular cortex – an area of the
limbic brain – is implicated in the experience of disgust (Small et al. 2003). Clearly this is not
something that we know by introspection, but that can only be discovered by means of
scientific investigation. Since the discovery requires sophisticated inferences from fMRI data
to unobservable neural processes and mechanisms, this particular aspect of the causal basis
must count as unobservable.
12 On introspection in psychology see e.g. Daston (1978) and, for economics, Maas (2005).
16
But this is not a problem for the realist doctrine as standardly conceived. Realism about a
certain type of entity does not require that all properties of that type of entity be directly
observable by means of sensory experience. The moon, to take a simple case, is observable to
the naked eye. And yet not all of its properties are accessible in the same manner: its mass, or
the chemical composition of its atmosphere, for example, are unobservable properties that
have been inferred by scientific means. And yet the moon counts as observable: even though
we can be wrong about some of its properties, the moon will always belong to our ontology.
Back to preferences, we can now try to generalize the argument from introspection beyond the
case of bodily states like feelings and emotions. Consider the case where an impulsive bodily
reaction is overridden by a higher cognitive process: although I find snails disgusting, I also
believe that they are very good for my health (or so my alternative-medicine friend says). So
on reflection I prefer to eat a snail rather than a very tasty but unhealthy sausage – and
swallow it. As examples like these demonstrate, preferences are all-things-considered
dispositions, or bundles of heterogeneous causal factors that, collectively, bring about (or
prevent) a certain action.13
Now, do we have access to all of these causal factors? The answer is almost certainly
negative. In the example of the previous paragraph the decision to swallow the snail is
reached by reasoning about the consequences of different dietary regimes.14 Being conscious,
the reasoning process is directly accessible by introspection. But there are plenty of cases
where the causal mechanisms behind a decision are invisible even to the agent. These include
all cases of subliminal, off-line processes that cannot be accessed by introspection, and the
existence of which we only know thanks to research in cognitive and neuroscience.
13 Hausman (2011) usefully distinguishes two meanings of preference in ordinary language: preferences as
likings, and preferences as comparative evaluations. Upon analysis, he concludes that preferences as modelled in
economic theory fit best with the notion of total comparative evaluations. Since Hausman also holds that
preferences can determine choices, his notion is very similar to what I call “all-things-considered dispositions”. 14 I assume here that the reasoning caused the action. This may turn out to be false, as some studies in cognitive
science purport to show (see e.g. Wilson and Dunn 2004), which again would invalidate the continuity thesis.
Since I’m interested here in scenarios that are friendly to knowledge by introspection, I will ignore this
possibility for the sake of the argument. I should however put on record that the empirical evidence invites
scepticism about introspection in my view, and that I find “fictionalist” accounts of preference/desire
explanations quite convincing.
17
Consider the “mere-exposure” effect: a large body of research demonstrates that repeated
exposure to a benign stimulus increases the probability of choosing the stimulus in a
subsequent decision task, even though the subject is unaware of the exposure (e.g. Elliott and
Dolan 1998, Zajonc 2001). Subliminal conditioning generates a reliable disposition to choose,
which therefore counts as a genuine preference without conscious awareness of the causal
basis and, in some cases, even of the preference itself. Another familiar example is preference
for brands, which most people rationalise away by inventing post-hoc stories about how good
a certain drink or food tastes, even though the evidence shows that the brand itself determined
the choice subconsciously (McClure et al. 2004).
A friend of commonsensible realism at this point may reply that the above examples suggest,
at best, that some dispositions are not directly accessible by introspection, while others clearly
are. But this does not amount to a demonstration that preferences (or their causal bases) are
unobservable in general, because the issue of observability does not apply to token entities,
but to types. What the anti-realist denies – and the realist accepts – is that we can have
knowledge of entities that are never observable, in principle, by unaided sensory experience.
Or, in other words, the scientific realist and anti-realist disagree on the introduction of new
unobservable categories of objects, not of individual objects whose categories feature already
observed tokens.
The following analogy may help clarify: some stars are directly observable, while others are
not. The issue of realism does not apply to these far-away stars, even though we can observe
them only using sophisticated telescopes. Irrespective of whether the telescopes are reliable or
not, the existence of the far-away stars is not metaphysically inflationary. Since we already
know that some stars exist, the addition of a few thousand extra stars does not increase the
types of entities belonging to our worldview. The addition of a single black hole, in contrast,
makes an enormous difference, for we have to add an entirely new category of objects that
previously (before the invention of x-ray detectors) was not part of our ontology.
The commonsensible realist then could argue that preferences are like stars. Scientific anti-
realists are concerned about the unnecessary inflation that would be prompted by the
introduction in our worldview of a new theoretical category, but would not mind about a few
18
extra preferences if the category is already there. Even though not all preferences are directly
observable (because they are hidden “too far inside” ourselves), some of them are accessible
by introspection. The discoveries of cognitive and neuroscience are not ontologically
inflationary: they just add more tokens of the same type.
6. From preference to reference
So it looks like introspection may provide commonsensible realists with a solid empirical
ground. But does it, really? In this section I will show that the appeal to introspection raises
some difficult problems concerning reference. Although these problems may not be insoluble,
no solution can be found in the writings of commonsensible realists at present. So minimally I
set out to articulate a series of objections that need to be addressed if commonsensible realism
is to be defended along the lines explored so far.
Philosophical theories of reference try to explicate the way in which the extension of terms is
fixed in everyday and scientific discourse. The debate during the second half of the last
century has revolved mainly around two alternative accounts, known as the “description
theory” and the “causal theory” of reference. Although commonsensible realists do not
commit explicitly to either account, it is clear that the position they want to defend requires to
endorse some elements of one theory and to reject some features of the other. Mäki (1999:
312-4) for example cites approvingly Keith Donnellan’s distinction between “attributive” and
“referential” uses of a description, and claims that failure to distinguish between them is at the
source of several mistakes in the realism debate. But the philosophical account that fails to
draw such a distinction is the so-called description theory of reference, according to which the
sentences of a theory simultaneously determine what we are talking about (the theory is used
referentially) and attribute certain properties to the referent (the theory is used attributively).
One of the attractions of the causal account in contrast is the way in which it makes sense of
those cases where we seem to be able to speak of a thing in false or contradictory terms. My
friend Paul and I, for example, can talk about whales even though Paul believes that whales
are mammals, and I believe they are fish. The latter belief happens to be false, but as long as
the term “whale” is firmly attached to a specific class of entities, this falsity does not matter
for referential purposes. By investigating the properties of those entities I may even come to
correct my mistake and learn something that I did not previously know about whales.
19
This is clearly a great advantage of causal compared to description accounts of reference,
according to which a theoretical term refers only to those entities that satisfy or make the
description true. In the above case, Paul and I would not be talking about the same animals –
in fact he would be talking about whales and I would not be talking about anything at all. In
the case of scientific theories a description account of reference implies that every time a
theory is modified (which happens rather often in science) its reference changes too. That
scientists subscribing to different theoretical paradigms do not talk about the same things (one
of Kuhn’s senses of incommensurability). And since even the best theories are only
approximately true, we also carry the risk of concluding that no scientific theory refers.15
The distinction between problems of realisticness (how preferences are described in rational
choice theory) and realism (do preferences exist at all) that commonsensible realists want to
draw clearly fits well with a causal account of reference. So even though commonsensible
realists are right to say that we currently lack a well worked-out theory of reference for social
objects (Mäki 1999: 315), there are many features of the causal account that such a theory
will have to be retained for it to be of any use in a defence of commonsensible realism.
According to classic causal accounts (à la Kripke-Putnam) reference is not fixed by
description but by an original act of “dubbing” during which a label is attached to an
exemplar of a certain kind of entity. The label then remains “attached” to that kind by
convention and is causally transmitted through a chain of social interactions among speakers
(hence the term “causal theory”), as these learn more about the true properties of the kind in
question. In the case of preferences, the original dubbing may take place by introspection, for
example when we experience strong feelings like cravings and disgusts.16
15 To avoid this paradoxical conclusion, some description theorists speak of quasi-satisfaction of theories by their
referents (e.g. Lewis 1972). But even advocates of realism acknowledge that it is difficult to go beyond vague
intuitions and develop a precise theory of approximate satisfaction (cf. Psillos 1999, Chs. 11-12). 16 This contrasts with description accounts à la Sellars, according to which preferences and beliefs are first
postulated by a theory aimed at predicting the behaviour of other agents, and then are identified with internal
mental states (as outlined in the famous “Myth of Jones”, cf. Sellars 1956). Other philosophers who endorse
similar views are Davidson and Dennett.
20
Once a term has been attached to an exemplar, its extension is determined by the way the
world is, rather than by how we believe it is. What preferences are, and how many preferences
there are, then depends on the structure of our cognition and is something to be discovered by
empirical investigation. So the causal account of reference depends crucially on the
assumption that the world already comes sorted into natural kinds, for otherwise the whole
idea of learning about the extension of terms would make little sense.
Now why should we assume that all preferences (dispositions to choose x instead of y) belong
to the same natural kind? The causes of behaviour may well be a heterogeneous, diverse lot,
with the term “preference” used as a placeholder to bunch them together, irrespective of their
differences or similarities. But in this case the question “are preferences observable” may
have no unique answer. Some of them are, while others perhaps are not.
This is not a mere speculation. There is no doubt that what we nowadays know about the
causal bases of preferences points towards a considerable heterogeneity. Some cognitive
mechanisms are conscious, others operate off-line. Some are located in the neo-cortex, others
in the limbic brain. Some are automatic, others are under cognitive control. Some are
regulated by dopamine, others by serotonin. And so on and so forth. The more we know about
the causal bases of preferences, the less the term preference itself looks like a genuine natural
kind.
Which is not so peculiar, after all. The male parasitic wasp is the smallest animal on earth;17 it
is only about 130 micrometers long, and a micrometer is one thousandth of a millimetre. Now
let us ask: are animals observable? Surely elephants are, but parasitic wasps are not. The
problem is that the term “animal” is so generic that asking questions of realism makes little
sense at this coarse level. Similarly, it makes little sense to ask whether preferences in general
are observable or not. If the problem of realism applies to some animal species, then surely it
applies to (some) preferences too.
A traditionally minded economist who is fond of her discipline’s autonomy from psychology
may point out at this stage that a term like “preference” was introduced precisely to ignore the
17 There is some controversy concerning amoebae, which under some definitions do not belong to the animal
kingdom though.
21
complications caused by the heterogeneity of psychological kinds, and focus on the properties
shared by a wide range of behaviours irrespective of the diversity of mechanisms that
generate them. This is a legitimate stance, and the economists who take it cannot be charged
of inconsistency. But if one is not interested in the causal bases of preferences, then one
should better be a behaviourist. This option is not open to the realist about commonsensibles,
as I have repeatedly argued, because the continuity of rational choice theory with folk
psychology (and especially the causal efficacy the latter attributes to preferences) would be
lost.
For many contemporary theorists the point of embracing a folk-psychological or cognitive-
science interpretation of rational choice theory is that it will enable us to improve the
explanatory and predictive power of economics. These theorists claim that it is very unlikely
to obtain such improvements – i.e. to find better causal generalizations – if we stick to
“generic” preferences and ignore the heterogeneous processes that govern human behaviour.
A quick look at the best alternative models of individual choice that have been proposed by
economists and cognitive scientists over the past three decades shows that they typically try to
capture the functioning of one or more specific cognitive mechanisms that cannot be captured
by the classic theory of expected utility. Some of these “generalized expected utility” models,
like Regret Theory (Loomes and Sugden 1982), are even named after a specific process.
Others, like Prospect Theory (Kahneman and Tversky 1979), capture more than one cognitive
mechanism or “bias”, such as probability weighing and framing.
These models are not mutually incompatible: although they sometimes compete for the
explanation of the same experimental phenomenon, they can also peacefully co-exist by
division of labour: each focusing on a specific mechanism that may or may not be triggered in
a specific context. Thanks to the availability of neuroimaging techniques, cognitive scientists
and neuroeconomists have made some progress identifying the substrate of these mechanisms
and locating them in specific areas of the human brain (see e.g. Coricelli et al 2005 on regret,
de Martino et al 2006 on framing).
So the latest science of decision-making tells us that the relevant scientific kinds are narrower
and more specialised than the generic use of the term “preference” would suggest. Those that
are accessible by introspection (such as regret) are probably only a subset of all the relevant
22
kinds. (For example, we are mostly unaware of the causes of framing effects.) The subliminal
ones are posited theoretically, discovered scientifically, and surely count as unobservable in
the context of the debate on scientific realism.
7. Conclusion
What makes realism about commonsensibles unstable as a philosophical position can be
summarized in the following dilemma: either we interpret the claims of rational choice theory
“thinly”, in line with old-fashioned behaviourist interpretations; or we give them a “thick”
psychological interpretation, to try to preserve all the power of folk explanations of action. If
we follow the first route, rational choice explanations do not refer to any unobservable entity
or property, as commonsensible realists claim. But this victory comes at the cost of severing
the continuity with psychological explanations of action that seems to be a cornerstone of
commonsensible realism. If we follow the second route, we preserve the continuity with
psychology, but at the cost of introducing unobservable entities and properties (mental or
neural mechanisms, depending on one’s tastes regarding monism, identity, and the reduction
of mental to physical states). The claim that problems of scientific realism do not arise in
economics would have to be abandoned.
So once behaviourism is rejected, scientific realism becomes a relevant issue in economics.
This makes sense, considering that realism about commonsensibles was not born in a vacuum.
The challenge to the behaviourist interpretation of the theory of choice in the writings of
authors like Amartya Sen (1987) was driven by strong revisionist motives. Kahneman,
Tversky and their fellow “behavioural economists” similarly believed that the return to a
psychologistic interpretation of preferences would help improve predictive and explanatory
power, with the help of the most advanced cognitive sciences. But the more we learn about
the cognitive and neural mechanisms that constitute the causal bases of preferences, the
further away we travel from the realm of the commonsensical and its methods of
investigation. We learn that generic “umbrella-kinds” like “preference” are unable to account
for the heterogeneous mechanisms that govern human behaviour, and we are pushed to revise
our ontology accordingly. New kinds and mechanisms are introduced by theoretical
hypothesis, and the problem of scientific realism resurges in its classic guise. This is, I think,
a rather pleasant conclusion for those who believe that the most interesting philosophical
23
problems have a substantial empirical element, and that our path towards finding a solution is
strictly linked with the progress of science.
References
Churchland, P.M. (1988) Matter and Consciousness. Cambridge, Mass.: MIT Press (2nd edition)
Coricelli, G., Critchley, H.D., Joffily, M., O'Doherty, J.P., Sirigu, A. and Dolan; R.J. (2005) “Regret and Its
Avoidance: A Neuroimaging Study of Choice Behavior”, Nature Neuroscience 8: 1255-1262.
Daston, L. (1978) “British Responses to Psycho-Physiology, 1860-1900”, Isis 69: 192-208.
Davidson, D. (1980) Essays on Actions and Events. Oxford: Oxford University Press.
De Martino, B., Kumaran, D., Seymour, B. and Dolan, R.J. (2008) “Frames, Biases, and Rational Decision-
Making in the Human Brain”, Science 313: 684-687.
Elliott, R. and Dolan, R.J. (1998) “Neural Response during Preference and Memory Judgments for Subliminally
Presented Stimuli: A Functional Neuroimaging Study”, Journal of Neuroscience 18: 4697-4704.
Fodor, J.A. (1987) Psychosemantics. Cambridge, Mass.: MIT Press.
Hausman, D.M. (1992) The Inexact and Separate Science of Economics. New York: Cambridge University
Press.
Hausman, D.M. (1998) “Problems with Realism in Economics”, Economics and Philosophy 14: 185-213.
Hausman, D.M. (2000a) “Realist Philosophy and Methodology of Economics: What Is It?” Journal of Economic
Methodology 7: 127-33.
Hausman, D.M. (2000b) “Revealed Preference, Belief, and Game Theory”, Economics and Philosophy 16: 99-
115.
Hausman, D.M. (2011) “Mistakes about Preferences in the Social Sciences”, Philosophy of the Social Sciences
41: 3-25.
24
Hoover, K. (1995) “Is Macroeconomics for Real?”, The Monist 78: 235-257; reprinted in The Economic
Worldview, edited by U. Mäki. Cambridge: Cambridge University Press, 2001.
Kahneman, D. and A. Tversky (1979) “Prospect Theory: An Analysis of Decision under Risk”, Econometrica
47: 263-91.
Kusch, M. (1999) Psychological Knowledge. London: Routledge.
Ladyman, J., Douven, I., Horsten, L., and van Fraassen, B. (1997) “A Defence of van Fraassen’s Critique of
Abductive Reasoning: Reply to Psillos”, The Philosophical Quarterly 47: 305-321.
Lewis, D.K. (1972) “Psychophysical and Theoretical Identifications”, Australasian Journal of Philosophy 50:
249-58. Reprinted in Papers in Metaphysics and Epistemology. Cambridge: Cambridge University Press, 1999.
Lewis, D.K. (1974) “Radical Interpretation”, Synthese 27: 331-344. Reprinted in Philosophical Papers Vol. 1.
Oxford: Oxford University Press, 1983.
Loomes, G. and R. Sugden (1982) “Regret Theory: An Alternative Theory of Rational Choice under
Uncertainty”, Economic Journal 92: 805-24.
Maas, H. (2005) “Jevons, Mill and the Private Laboratory of the Mind”, The Manchester School 73: 620-649.
McCloskey, M. (1983) “Intuitive Physics”, Scientific American 248: 122-129.
McClure, S.M., J. Li, D. Tomlin, K.S. Cypert, L.M. Montague, and P.R. Montague (2004)
“Neural Correlates of Behavioral Preference for Culturally Familiar Drinks”, Neuron 44: 379-387.
Mäki, U. (1996) “Scientific Realism and Some Peculiarities of Economics”, in R.S. Cohen, R. Helpinen and Q.
Renzong (eds.) Realism and Anti-Realism in the Philosophy of Science. Dordrecht: Kluwer.
Mäki, U. (1998) “Aspects of Realism about Economics”, Theoria 13: 301-319.
Mäki, U. (1999) “Representation Repressed: Two Types of Semantic Scepticism in Economics”, in
Incommensurability and Translation, edited by R. Rossini Favretti, G. Sandri and R. Scazzieri. Cheltenham:
Edward Elgar, pp. 307-321.
Mäki, U. (2000) “Reclaiming Relevant Realism”, Journal of Economic Methodology 7: 109-125.
Mill, J.S. (1836) “On the Definition of Political Economy ad the Method of Investigation Proper to It”, in
Collected Works of John Stuart Mill, Vol. 4. Toronto: University of Toronto Press, 1967
25
Mumford, S. (1998) Dispositions. Oxford: Oxford University Press.
Nagatsu, M. (2010) “Beyond Circularity and Normativity: Measurement and Progress in Behavioural
Economics”, Philosophy of the Social Sciences 40: 265-290.
Pettit, P. (1991) “Decision Theory and Folk Psychology”, in Foundations of Decision Theory edited by M.
Bacharach and S. Hurley. Oxford: Blackwell.
Psillos, S. (1999) Scientific Realism: How Science Tracks the Truth. London: Routledge.
Rosenberg, A. (1976) Microeconomic Laws. Pittsburgh: University of Pittsburgh Press.
Rosenberg, A. (1992) Economics – Mathematical Politics or Science of Diminishing Returns? Chicago:
University of Chicago Press.
Ross, D. (2005) Economic Theory and Cognitive Science: Microfoundations. Cambridge, Mass.: MIT Press.
Samuelson, P. (1938) “A Note on the Pure Theory of Consumer’s Behavior,” Economica 5: 61-71.
Sellars, W. (1956) Empiricism and the Philosophy of Mind. Cambridge, Mass.: Harvard University Press.
Sen, A. (1987) On Ethics and Economics. Oxford: Blackwell.
Sen, A. (1993) “Internal Consistency of Choice,” Econometrica 61: 495-521.
Small, D.M., Gregory, M.D., Mak, Y.E., Gitelman, D., Mesulam, M.M. and Parrish, T. (2003) “Dissociation of
Neural Representation of Intensity and Affective Valuation in Human Gustation”, Neuron 39: 701-711.
Snell, B. (1953) The Discovery of the Mind. Oxford: Blackwell.
Stich, S.P. (1996) Deconstructing the Mind. Oxford: Oxford University Press.
van Fraassen, B. (1980) The Scientific Image. Oxford: Clarendon Press.
Wilson, T.D., and E.W. Dunn (2004) “Self-Knowledge: Its Limits, Value, and Potential for
Improvement”, Annual Review of Psychology 55: 493-518.
Zajonc, R.B. (2001) “Mere Exposure: A Gateway to the Subliminal”, Current Directions in Psychological
Science 10: 224-228.