an essay by Phin Upham

An economic treatment of human behavior generally rests on a structured account of rational choice theory (RCT). Such models often posit utility maximization behavior by all individual agents. In many cases it assumes that all resources can be allocated and evaluated in an orderly way for each individual and that there is a strict utility evaluation function which is strictly monatomic at every point, concave, and twice differentiable, etc.. This formulation is generally highly mathematical and continues to make information and goods commensurable and to allow expectations of other agents rationality to guide utility maximization considerations. This view of human action seems to be far from what we feel going on in our own heads. Even worse, it is hard to imagine that this kind of calculation does go on in our head since it leaves out so many of the important factors that we feel contribute to our decision making procedure. Jealousy, fear, risk-aversion, greed, psychological associations, et al are not incorporated in this model, yet we feel, and psychological data support this, that these do influence our decisions. Finally, it seems, if this is not how we make our decisions, it would be a complex question how a study of such an analysis cannot meaningfully help us to act more rationally or make decisions better. Nevertheless, given sufficient tweaking and enough creative manipulation of the variables, this model might serve as a first-cut approximation of human behavior.

An objection can immediately be lodged that the intuitive feeling that we do not use rational choice theory, even loosely, to make everyday decisions does not carry much weight. Our intuitive feelings about ourselves, however valuable they may be, only carry weight in the realm of what we do consciously. We have little reason to believe that we have strong or reliable intuitions about our unconscious selves (and it would be hard to imagine a way even to verify these intuitions). Thus, a few topics can be posited for further discussion. 1) We consciously use the kind of analysis agents in most RCT models do, or a variation/distortion of it. 2) We unconsciously use the kind of analysis agents in most RCT models do or a variation/distortion of it. 3) Perhaps RCT does not describe the method that we use, consciously or unconsciously, to make decisions in most cases (though we might in some specialized situations). Even if 3) holds true, it is nevertheless conceivable that RCT is able to approximate the results of whatever process we actually do use.

The second point becomes important if we see RCT as making a claim about whether or not RCT is a causal or descriptive process. If it is merely descriptive, that is, if we are merely trying to approximate how individuals act without attributing this as a reason for their actions, then RCT seems to be on firmer ground. Certainly Becker, among others have found good reasons to believe that peoples actions are in keeping with some rational laws.

But this apparently firmer ground may be illusory. If the correlation is simply a matter of historical observations, and not causal, then are there any reasons to expect this correlation to continue? No rigorous Humean skepticism is being marshaled here. No doubt, this correlation between rational expectations of action based on RCT and actual human actions will continue in the future (with its cause/effect considerations and maximizing ends), but when dealing with a subject as complex and misleading as human action, it is worthwhile to know why something holds true not just that it does. In the Cartwrightian analysis a model provides both a causal and predictive explanation of events. Further, there are other explanations, other models, of human action other that RCT. Lastly, it might argued that the connections between RCT and human action are simultaneously not causal[1], and yet reasons not causally generated can be taken to generate the expectation that this connection will continue using an evolutionary perspective.

This discussion attempts to question whether RCT is intended to be truthful or fruitful. Are Rational Choice Theorists such as Gary Becker claiming that people making decisions on some level, consciously or unconsciously, make decisions the way they assume they do? Or, weakening the claim, do they believe that people make decisions on the whole, and mostly, using rational means? Alternatively, rational choice theorists could be making the claim that people do not make decisions rationally (however they are construing the term) but if we assume they do, we can get accurate and revealing predictions (and on some level explain past actions). Further, this viewpoint may be seen as being illuminating in its ability to categorize and organize the massive collection of past behavior.

If we take the view that RCT are being roughly truthful, and simply being too simplistic, we might internalize more factors from the world and make the model more “realistic.” Second-cut descriptive models might therefore incorporate, at the expense of mathematical elegance and simplicity, the positing of “quasi-rational motives for individual actors and make the model dynamic with adaptive learning feed-backs. A third cut descriptive model might try to better understand a richer panoply of human motivations and emotions so that what is deemed ‘quasi-rational’ behavior might indeed be more fully rationalized.” (decision making, p15).

There seem to be two points in which some literature diverges from a classical Rational Choice Theory model with its maximization of ranked preferences and ranked options at its heart and a clear view of the world as its perspective. The first is the claim that there is no clear way to model a rational way to act, rather we can act in keeping with rationally by acting in many different ways. Tversky and Kohneman, Simon, Hernstein, Bell, and recent psychological literature (dealing with the importance of peaks and ends) have all suggested that there are many ways to formulate rational behavior. Tversky and Kohnemon (86), for example, argue that people behave rationally with the stipulation that they are risk adverse with proportionally large sums (to their net worth) and risk hungry with proportionally small sums. Bell includes some psychological dimensions such as regret and disappointment to the calculus of human decision making. With competing understandings of rationality we may find that a number seem to make sense without invalidating the meaning of a rational framework.

But we must be careful to lump even this class of dissenters into one group. Some of these, Bell being a clear example, as mentioned above, want to internalize aspects of the human decision making process into rationality theory so that our models can be more realistic and more useful in predicting reality. This view is neatly stated in footnote 6 of Alchian’s seminal essay “Uncertainty and Economic Theory” when he says “Analytical models in all sciences postulate models abstracting from some realities in the belief that derived predictions will still be relevant. Simplifications are necessary, but continued attempts should be made to introduce more realistic assumptions into a workable model with an increase in generality and detail.”[2]

The economists previously mentioned are complicating an overly simplified model with real world intricacies in order to bring it closer to the way things are. This attempt has been made throughout economics with the recent application of stock market analysis being one of the prime drivers. No longer is theoretical beauty and simplicity as precious when accuracy becomes a financially important.

But there is another sort of criticism that in many ways looks very similar to this sort, and which I group in the same class. This criticism is much more dangerous for RTT. It claims that rational choice theory itself cannot be coherently understood or applied to the world. Or rather, that there is not one clear way to apply RTT if one does what is necessary to apply it – that is, slightly weakening at least of its obviously false postulates. This tweaking is, again, done in the service of realism. Since, when so weakened, the theory itself leaves us with a multitude of equally legitimate alternatives of application – all with distinct outcomes – it is hard to make sense of it as it stands. It seems that even if we act according to the precepts of rationality, we might act rationally over a local spectrum of ranked and known options if complete information is not assumed. With the crumbling of the complete information hypothesis it becomes unclear how or when any actual decision can be made. How much information is enough? What about options that are no explicitly offered or obvious but are nevertheless available, are they considered? How long a time range ought one to think about consequences in? A month? A year? A lifetime? Even with the useful tool of present-day value considerations of future benefits, it is not clear if one overwhelming joy in the far future can coherently be thought of as preferable small joys over time.

It is important to realize that this criticism does not necessarily reject RCT, but rather each critic sets up a theory that he/she believes is the real way people apply RCT to a realistic world. While in the first set of critics bring in external factors were introduced into the theory for realism’s sake, the second set of critics, by weakening even the slightest assumption of RCT, fracture it so that it is unclear how, or how usefully, it is applied. Both of these criticisms can be unified under one roof if we consider the first set as a subset of the second. We do this by framing the first set as weakening the assumption that the model contains all the relevant considerations that humans use when making a description (since scientific model believes it is explanatorily and causally complete in generating its specific outcome). These objections do serious damage to RCT. What good is a theory that can be applied coherently only in a fantasy (Cartwright, we have seen, attempts to answer this)?

The second class of critics differs importantly from both of the above discussed sets of critics. These people reject RCT as the process of human thinking altogether. March and Alchian are two good examples. March claims that RTT cannot explain why we did our actions in any meaningful sense. Rather, it is a tool to make future actions. Alchian claims that humans use an evolutionary selection approach to decision making. He appeals to G. Tintner’s claim that since actions and events do not have certain outcomes but instead a spectrum of outcomes, the idea of profit maximization becomes incoherent. How can one compare a narrower but taller utility distribution curve to a wider but shorter distribution curve? If the value of outcomes is not commizerable, RCT crumbles. Alchian continues by constructing a very different way of describing the human decision making process.

So what defenses might our first-cut approximation have as a model for human behavior? Bell, Raiffa, and Tversky (Decision Making) suggest that though virtually no subjects behave coherently enough to satisfy the implied behavior which the assumptions of the model generate, the model might be empirically useful because if individuals deviate too far from the behavior posited by the model then others would be able to exploit such behavior since it is not maximizing.[3] In the long run, therefore, people would remain in an equilibrium state close to the behavior predicted by the model. Thus “the normative character of the model is used as an argument to reinforce its descriptive value.” (DM 15).

But this explanation seems unsatisfying. Even if we grant that this model approximated the way people ought to act, there is no implication that the model is descriptively or causally accurate. We remember from Nancy Cartwright’s analysis that causal accuracy is the heart and soul of a good model. An example might illustrate: a tube half filled with water and half filled with sand can be tipped slowly in one direction and then the other. The resulting pattern of movement of the sand in the water is similar to the movement of the M1 money supply in a well functioning economy. This is because, theory stipulated, the laws which govern sand and water in this constrained situation are similar to the laws which govern money in an economy. A model of this sort is sometimes called an analogical model. There is no implication, however, that sand is like money or that water is like society. Similarly, though models of human rationality can be descriptively accurate there is not yet any reason to believe that we are anything like rational agents or that our society is anything like that posited in the model. Descriptive power does not imply explanatory power.

Nancy Cartwright spoke extensively of this problem in her treatment of physics. How do we know a theory is explanatory and how do we know when it is merely descriptive? As in the case of Ptolemaic astronomy, the difference is not always immediately obvious. A look at how this issue was resolved in physics, with its obsession for methodology, might help us find a possible route for solving the problem in economics. As discussed earlier, both disciplines are similar in structure (based on observations and experimentation), and both claim to be sciences in some sense of the word. Thus we have reason to believe that such a structural methodological parallel is appropriate.

For Cartwright, the question is not whether or not the laws of rational choice theory are causally connected to reality but whether or not they are causal connected to the simulacrum – which, in a properly constructed simulacrum based model, they are. Thus the fact that we do not think exactly as rational agents do, but have more complex mental processes is no longer intrinsically troublesome since the simulacrum is explicitly and necessarily not identical to reality. Importantly, not only are rational agents simplified, but the entire context in which we function is similarly simplified and limited. It would be strange to change us and not our world. For example, it is assumed in the world of the simulacrum for RTT that conditions are such that rational thought is always, or at least generally, possible. So no constant clanks or excruciating pains may be present and warp the agents ability to act as the model predicts. In other ways the world of the simulacrum is changed as well. For example, the precepts instilled by natural selection that tend to maximize inclusive fitness seem to, at least in some ways, logically conflict with a RCT perspective. But in the simulacrum the agent is acting according to stricter precepts.

Does this lead to the conclusion that we cannot apply a Beckarian type analysis, which uses simplified agents, in a complex unsimplified reality? If it is indeed true that we are we leaving reality complicated and only simplifying ourselves when both should be simplified. If Becker’s mode of explanation depends on our letting loose simplified agents into an environment substantially different than the simulacrum (in which the laws of their behavior are explanatory) then the model seems to be seriously flawed. On the other hand, if the approximation between the simulacrum and reality is close enough in the right way, then perhaps the application, though flawed, is indeed useful.

Thus the appropriate question is whether or not the assumptions and simplifications of RCT are able to approximate reality in a way that is appropriate to the purpose that we are using the model for. Cartwright never gives us a clear account of how we can know this. Defense of rational choice theory in the light of the leeway provided by her description of laws as not true/realistic in the traditional sense. How many of the attacks on RCT does this deal with? Since rational choice theory is derived from a simpler and more limited reality than the rich reality of existence, there is room within any apparently “irrational” description for it to have been taking into account aspects and dynamics that a rational choice model cannot, i.e. rational choice theory is not the optimal decision mechanism since it is necessarily not totally descriptively accurate. <MORE HERE><Text from Cartwright about how Simulacrum and reality must be similar>

Further, as Cartwright Argues in the Cogito Debates, rationality is only part of the story. In her discussion of physics she pointed out that no one law holds in a vacuum but will instead interact with other laws (gravity and a positive charge, for example). These two laws cannot be isolated but instead make up the infinitely complex nature of reality. Neither acts independently and their effects cannot be separated. Similarly, rationality does not act in a vacuum. Though we can try to isolate rationality in “laboratory settings,” say the business world, we will only be moving closer and closer to a pure view of rationality. There will always be other factors and other motivations in the real world. The assumption of the purely rational actor is wrong at best, and useless at worst. The purpose of this is to recognize the power of the RCT framework, only to tease out methodological

 

Motivating the Issue

An economic treatment of human behavior generally rests on a structured account of rational choice theory (RCT). This understanding is wide-spread within economics. Economists Daniel Hausman and Michael McPherson say, in a joint publication, “at the foundation of both positive and normative economics lies a normative theory of individual rationality.” They continue by defining the conditions in which agents are rational: “agents are rational if and only if their preferences may be represented by ordinal utility functions, and their choices maximize utility.”

In an introduction to a book on the renown rational choice theorist Gary Becker’s work, Mariano Tommasi, et al, make a similar claim: “Economists try to understand and explain the world by assuming that the phenomena they observe are the outcomes of peoples purposeful decisions. Individuals try to achieve their objectives given their limitations – limited time, money, and energy – that is to say they optimize” (italics not mine).

It does not seem strange that economics, commonly refereed to as the study of human action under certain constraints, uses as a basic building block a theory of individual action. If rationality theories are indeed so fundamental to economics, then a close methodological look at the claims of there theories and their relationship to imperial laws is vital. This is becomes doubly certain of one looks at the sorts of criticisms which have been aimed at RCT.

In the last section we raised the question whether economics was a science. Many of the criticism’s – of inaccuracy, simplicity, and incorrect assumptions – are clearly present in theories of human rationality. Rationality modeling is perhaps one of the most challenging for the Simulacrum account to address since it often makes such drastically unrealistic assumptions and deals with a topic so intricately bound up with our actions. Rational Choice Theory models go beyond merely descriptive laws of human psychology when they become simulacrum – that is, when they are asked to generate predictions, serve as causal explanations of past actions, and even, in normative accounts, justify a certain way of acting.

 

Types of Critics to RCT, an overview.

 

There seem to be two points in which critics diverge from a classical Rational Choice Theory model with its maximization of ranked preferences and ranked options at its heart and a clear fully-informed view of the world as its perspective. Within the first set of critics, I differentiate between sorts of critics.

The first group of critics can be roughly divided into two sub-groups which I will call 1.a and 1.b. Of course, these two have some overlap, but it is valuable nevertheless to differentiate since in general each takes a different route when objecting to RCT. A critic in this first group finds RCT to either leave an important aspects of reality out (1.a), or believes that the assumptions of RCT are too harsh and adjusts one of the conditions to better fit reality (1.b). They can be seen as those who believe that RCT is too loose a glove to fit reality, or that it is too tight a glove to fit reality.

Unlike the first sort of critic, the second sort of critic rejects RCT as a mechanism that exists in reality in any meaningful way. Donald Green and Ian Shapiro exemplify this view when they say: “we do not dispute that theoretical models of immense and increasing sophistication have been produced by practitioners of rational choice theory, but in our view the case has yet to be made that these models have advanced out understanding of how politics [or any other field] works in the real world… [many such models,] on reflection, can only be characterized as banal: they do little more than restate existing knowledge in rational choice terminology.” (Pathologies… 6).

 

The First Sort of critic

The first sort of critic claims that people do not behave consistently with the rules of rationally. Simon, Hernstein, Bell, and recent psychological literature (dealing with the importance of peaks and ends) have all suggested that there are many ways in which agents do not seem to be strictly following the fundamental laws of RCT. They point out inadequacies or inaccuracies in the RCT model derives, usually, through empirical analysis in order to support their model.

David E. Bell (1985) includes some psychological dimensions such as regret and disappointment to the calculus of human decision making. He believes that these are considerations that affect ones actions but are not included in the RCT model. Your reaction to any piece of news, and your approach to any choice, he argues, cannot be understood given only utility expectations.

An example might illustrate: say that you are rewarded a magna cum laude on your senior thesis. Are you happy or sad? Do you feel that you have gotten something of unexpected and great value or been disappointed? If you were expecting a cum laude then you would be overjoyed and you might even dance for joy. If you were expecting a summa cum laude, you might not be overwhelmed, and might be disappointed. How you feel about any reward or outcome will depend on what you expected, not just on the value of your prize. Further, if you find out that all of your friends received higher honors than you did, you might feel worse about an outcome, even if it exceeded your expectations. The greater your expectations, the more disappointment you will feel. Those who are very adverse to disappointment may in fact adopt a pessimistic viewpoint about the future as a proactive defense against feelings of disappointment.

This also might affect your decision making. Bell uses an example from Kahneman and Tversky (1979) in which a prize of $3,000 for sure if preferred by a majority of respondents over an 80% chance at $4,000 (and a 20% chance at nothing), whereas a majority prefer a 20% chance at $4,000 over a 25% chance at $3,000. If expected utility were the only factor considered here (ignoring for the moment the diminishing marginal utility of money) then the agent would be preferring $3,000 over $4,000*.80 or the utility of $3,000 over the utility of $3,200 in the first case and the utility of $4,000*.20 over $3,000*.25 or the utility of $800 over the utility of $750 in the second. Other examples show that the procedure with which one arrives at a choice makes quite some difference to the outcome of your choice. If given a choice to It is clearly undesirable from a RCT viewpoint to make such inconsistent choices, and it is clearly not consistent with a rational choice theory. Bell’s hypothesis is that agents make decisions not only on RCT considerations, but also on other considerations such as the avoidance of disappointment. The sure $3,000 seemed more desirable, Bell argues, since it involved no chance at disappointment even though it netted less money on average. But in the second case, both choices involved significant risk of loss, so the agent chose the one which would maximize expected utility.

What are we to make of such criticisms of RCT given the methodological implications of Nancy Cartwright’s Simulacrum approach to model making? Ought we to be concerned that the RCT model does not correctly predict this outcome and that the RCT model seems to leave out disappointment as a factor in decision making? Cartwright’s view of Simulacrum modeling would have much to say on this subject.

To begin she might point out that it was intrinsic to the nature of a simulacrum model such as that of RCT that the fundamental laws that pertain to it will have inaccuracies and inadequacies when applied to reality. In short, the fundamental laws will be descriptively inaccurate in some situations as a necessary consequence of the unrealistic assumptions and shielding that was necessary in order to get the laws to be causally connected to the simulacrum (as is required to have explanatory laws rather than merely accurately predictive ones). “Really powerful explanatory laws of the sort found in theoretical physics do not state the truth”(Lie,3).

Is it even troubling that these fundamental laws are not completely descriptively accurate? Not necessarily. Phenomenological laws are the ones which are supposed to be most consistent with reality, fundamental laws such as those of RCT are more complex than that. To begin with, the purpose of the model must be taken into account. I discussed earlier the fundamental laws can be applied to different domains and larger domains (at a greater and greater cost of predictive accuracy) depending on the purpose of the modeler. If a small domain were to be covered, then the fundamental laws pertaining to it ought to be very close to predictive, but if a larger domain was to be covered, then the fundamental laws pertaining to it ought to have large gaps and inaccuracies. In the latter case, more unrealistic assumptions and more false constraints would have to be posited in order to make a simulacrum that could support the relevant fundamental laws.

The inaccuracy to a set of fundamental laws such as RCT is the very aspect of a model that allows it to be broadly useful. “The explanatory power of quantum theory comes from its ability to deploy a small number of well understood Hamiltonians to cover a wide range of cases. But this explanatory power has its price. If we limit the number of Hamiltonians, that is going to constrain out abilities to represent situations realistically. This is why our prepared descriptions lie.” (lie 139). Phenomenological laws accurately govern small domains. When we construct fundamental laws we want them to causally explain larger domains. But in order to do this they must abstract away from the details of reality that differentiate the domains or that make the model too complicated. For example, in the law of gravity we do not factor in the changing air pressure (and the resulting difference in air resistance) because we believe it to be too small. Indeed, in most cases we do not even factor in air resistance to the speed of dropping objects. The level of detail we include very much depends on the function or purpose of our model. Are scientists trying to make exact measurements or are we simply trying to calculate the path of a rock? If RCT is expected to apply to a domain as large and complex as human decision making and still be simple enough and accurate enough to be manageable, then inaccuracies, even large ones, are unavoidable.

Some models in a science are not meant to be descriptively accurate at all, but instead serve an organizational function. “many abstract concepts in physics play merely an organizing function.” (lie 18). In these cases, it would not be troubling at all that the concepts did not fit the facts, this was not their purpose.

An even deeper consideration remains. Is it even possible for any model to accurately capture reality? Cartwright claims that it is not. “Nature tends to a wild profusion, which our thinking does not wholly confine.” (lie 19). Nature is either inherently inconsistent, or to modify her rather unconventional view, extremely complicated so that no one set of rules will ever apply to every domain. So RCT rules will always apply better to highly constrained example of a man making routine decisions in the business world than to an experimentee caught by an artificial and rare choice. Cartwright, in the Cogito interviews, says that reality and the results of a model will never mesh perfectly. On can construct artificial laboratory conditions in which the results mesh almost perfectly, but the resulting match will be asymptotic; the predicted results will approach but never exactly mirror reality.

Bell himself acknowledges the substantial claims of this analysis, though likely not for the same reasons, at the end of his critique of the rational choice theory. He says:

“Disappointment, and the related concepts such as regret, have important implications for the study of decision making under uncertainty… While it has taken a study of descriptive behavior to force recognition of the importance of psychological impacts to the decision maker, it is not our intent to revise the normative theory continually until it matches empirical evidence… this chapter may convince decision makers that what is currently omitted from expected utility analysis deserves to be omitted and that a formal analysis may be exactly what is needed to prevent a decision maker’s intuition from forcing economically inefficient decisions.”

From Bell’s analysis we see that the ability to find inadequacies in the predictive power of RCT is not enough to invalidate it as a descriptive or normative theory. Descriptively, it must be able to predict the actions of people in a way that is both not needlessly complex, for it would be ridiculous and worthless to have a different model for every different situation, and that stays close enough to reality so that that the laws that can be supported by the model are useful.

Cartwright seems to have some convincing points when she argues that fundamental laws will never be able to be perfect predictors, nor ought they to be. Her objection seems to be a powerful way to begin to refute, and at very least recast, some of the criticisms of RCT. Certainly, the criticism that there are inaccuracies by itself becomes a self-evident one. The question that would remain is whether these inaccuracies are a result of legitimate simplifications or the result of sloppy modeling. The answer to this question will depend on the domain of the model and the purpose of the model.

Cartwright never gives us a detailed description of how to evaluate and define these two areas. But it might be useful to point out that RCT has a very large domain indeed (all of human action and interaction) and that its function is not intended to be purely descriptive, but also normative (though, it has been argued by Kahneman and Tversky that this confutation of the purposes of RCT is exactly the problem). Further, it ought to be simple enough to be understandable and useful to economists. Though we do not have the tools to dismiss this criticism, it certainly remains in the authors hands to establish more than that the RCT model is inaccurate, but also that this inaccuracy is preventing the model from fulfilling its function or that this inaccuracy is so large that the simulacrum no longer corresponds to reality in a meaningful way.

But her analysis that some disparity between reality and a good model does not mean that all models, no matter how inaccurate, are good. To go this far would be throwing the baby our with the bath-water. We ought to be able to allow for inaccuracies within good models without thereby barring the charge of inaccuracy as being a valid criticism of a bad model. Ptolemaic physics is a good example of a relatively predictive model that is simply too inaccurate. This model is not considered to be a good explanatory one. But the fact that it is inaccurate is, perhaps, secondary to the fact that it posits multiple celestial spheres which have no justification for existing, and even this claim is perhaps less vital than the much better Cartesian model that has been developed to take its place. Since Cartwright has given up the truthfulness of models as they correspond to reality, she can also allow for multiple models that all describe, better and worse, phenomenological events.

It seems Cartwright intends there to be two sorts of criteria for a good model. “The success of an explanatory model depends on how well the derived laws approximate the phenomenological laws and the specific causal principles which are true of the objects modeled.” (lie 17). So the model must be as close to reality as it can be while still being far enough from reality in order to be able to abstract away enough irrelevant and unnecessary elements so that it can be causally related to the events it posits. Though Cartwright’s analysis does provide insight into these sorts of criticisms of RCT, showing that establishing inaccuracies is not enough, it also fails to provide a way to establish whether these inaccuracies are acceptable.

 

Critics of RCT 2.b

But we must be careful to lump even this first class of dissenters into one group. Some of these, Bell being a clear example, as mentioned above, want to internalize aspects of the human decision making process into rationality theory so that our models can be more realistic and more useful in predicting reality. This view is neatly stated in footnote 6 of Alchian’s seminal essay “Uncertainty and Economic Theory” when he says “Analytical models in all sciences postulate models abstracting from some realities in the belief that derived predictions will still be relevant. Simplifications are necessary, but continued attempts should be made to introduce more realistic assumptions into a workable model with an increase in generality and detail.”[4]

There is another sort of criticism that in many ways looks very similar to this sort, and which I group in the same class, but which also differs t a large extent. It claims that rational choice theory as it stands cannot be coherently understood or applied to the world. Rather one must slightly weakening at least some of its more obviously false postulates in order for it to correspond to reality. This tweaking is, again, done in the service of realism. Some RCT’s, for example, demand that the agent have perfect information – some objectors to these theories suggest that this is too unrealistic an assumption to allow to stand and still yield accurate results. Other critics suggest that agents do not have perfectly ordered and complete preferences (Levi, 1980).

Chu and Chu (1990), for example, were able to show that agents preferences were not complete or transitive by selling K bets to experimenters, for the price they claimed they were worth, exchange K bets for J bets, and then having the experimenters buy back the J bets for the lower prices subjects claimed they were worth. This money pumping cycle, of course, left the subjects poorer after each round. If the agents preferences were transitive, then if they preferred A to B and B to C, they would also prefer A to C. In the Chu and Chu example, though the agents showed intransitive preferences, they quickly adjusted their preferences when the experimenters began to take advantage of their preference ordering. Bell, Raiffa, and Tversky (Decision Making) suggest that though virtually no subjects behave coherently enough to satisfy the implied behavior which the assumptions of the model generate, the model might be empirically useful because if individuals deviate too far from the behavior posited by the model then others would be able to exploit such behavior through arbitrage since the behavior is not maximizing.[5] In the long run, therefore, people would remain in an equilibrium state close to the behavior predicted by the model. Thus “the normative character of the model is used as an argument to reinforce its descriptive value.” (DM 15). This is exactly what we saw with the Chu and Chu example. (****Should I use the race track counter example?****).

When viewed through a Simulacrum account, false or overly strict rules take on a new meaning. The rules governing a set of laws must be more constraining than the rules governing reality in order to be able to allow the Simulacrum to be causally connected to the simulacrum. In the field of Economics, Carwright makes the special proviso that in economics, there are an enormous range of applications inherent to the concrete concepts, but in order to get deductive inferences we have to place them in special models. These assumptions, though unrealistic, are what allows us to take highly general points and force them to pertain to specific situations in a useful and meaningful way. Left unmediated, the idea that agents act to maximize their goals would be useless. Only when constrained and abstracted with assumptions about information and evaluative structures of the agent can a theory that derives useful results be generated.

Perhaps these assumptions are necessary for a causal connection between the model and the fundamental laws, perhaps they are necessary in order to shield the model from the infinite other considerations a true decision makes must sort through. In the end, these assumptions, as Milton Friendman in his previously discussed essay “The Methodology of Positive Economics” says, “the particular ‘assumptions’ termed ‘crucial’ are selected on grounds of their convenience in some such respects as simplicity or economy in describing the model, intuitive plausibility, or capacity to suggest, if only by implication, some of the considerations that are relevant in judging or applying the model.” (Rinthe PSS, 525). Though Friendman differs significantly in his theory from Cartwright, taking a more pragmatic view, it is nevertheless clear from this passage that he considers assumptions to be held to very different criteria than verisimilitude. The critics who point to inaccuracies in RCT’s assumed laws and propose to change them need to do more than establish a divergence from the truth, they also need to establish why this divergence is an unreasonable one. Alternately, a critic might argue, it is on the shoulders of the Rational Choice theorists shoulders to prove that each and every assumption is a justified one and is necessary for the purpose and causal integrity of his model.

On a trivial level, not only are rational agents simplified as previously discussed, but the entire context in which we function is similarly simplified and limited. It would be strange to change us and not our world. For example, it is assumed in the world of the simulacrum for RCT that conditions are such that rational thought is always, or at least generally, possible. So no constant clanks or excruciating pains may be present and warp the agents ability to act as the model predicts. In other ways the world of the simulacrum is changed as well.

So a theory that takes fundamental laws and expects them to pertain in a neat way to reality will likely fall short. This may lead to the conclusion that we cannot apply a Beckerian type analysis, which uses simplified agents, in a complex unsimplified reality? Cartwright certainly does claim that the fundamental laws apply to the world of the simulacrum and not to the more complex and messy one of reality. This has some force if it is indeed true that we are we are leaving reality complicated and only simplifying ourselves when both should be simplified. If Becker’s mode of explanation depends on our letting loose simplified agents into an environment substantially different than the simulacrum (in which the laws of their behavior are explanatory) then the model seems to be seriously flawed. On the other hand, if the approximation between the simulacrum and reality is close enough in the right way, then perhaps the application, though flawed, is indeed useful.

The false or overly generalized laws of a simulacrum are not meant to be absolutely perfect approximations of reality. As the critics in both 1.a and 1.b have correctly pointed out, no laws or simple set of laws can perfectly predict or describe the complex and messy minds of agents. The domain is too large, and the complications, whether instilled evolutionarily or psychologically, are too complex. Philosophy of science has seriously questioned the validity of the covering-law model which states that laws pertain to reality. It is forcefully argued by Ronald Giere in Science Without Laws (p***), Nancy Cartwright, and a rising branch of philosophy of science authors that different models have different functions and that no one model can meaningfully even begin to perfectly correspond to reality. A respected example of this is Nancy Cartwright’s Simulacrum account.

The economists previously mentioned are either complicating an overly simplified model with real world intricacies or weakening some assumptions of RCT all in order to bring it closer to the way reality appears to be. Both these attempt has been made throughout economics with the recent application of stock market analysis being one of the prime drivers. No longer is theoretical beauty and simplicity as precious when accuracy becomes financially important.

This criticism doe not necessarily reject RCT, but rather each critic sets up a theory that he/she believes is the real way people apply RCT to a realistic world. Both of these criticisms can be unified under one roof if we consider the first set as a subset of the second. We do this by framing the first set as weakening the implicit assumption that the model contains all the relevant considerations that humans use when making a decision (since scientific model believes it is explanatorily and causally complete in generating its specific outcome). These objections do serious damage to the conceptions of RCT they criticize.

 

Second Class of Critics

The second class of critics differs importantly from both of the above discussed sets of critics. These people reject RCT as the process of human thinking altogether. March and Alchian are two good examples. March (1973) claims that RCT cannot explain why we did our actions in any meaningful sense. Intentionality, which is at the heart of rationality, he argues, is an interpretation of action, rather than a prior position. Rather, RCT can at best be seen as a tool to make future actions, but not an explanatory mechanism.

Alchian claims that humans use an evolutionary selection approach to decision making. He appeals to G. Tintner’s claim that since actions and events do not have certain outcomes but instead a spectrum of outcomes, the idea of maximization becomes incoherent. How can one compare a narrower but taller utility distribution curve to a wider but shorter distribution curve? If the value of outcomes is not commensurable, RCT crumbles. Alchian continues by constructing a very different way of describing the human decision making process. More recent psychological literature, such as that of Donald Green and Ian Shapiro, has questioned the empirical underpinnings of RCT, claiming that it has serious methodological flaws that are revealed when good empirical experimentation shows that it is hopelessly out of touch with reality. Tversky and Kahnemon (86), for example, argue that RCT is so seriously flawed that he constructs a new theory of rationality. We argue that the deviations of actual behavior from the normative model are too widespread to be ignored, too systematic to e dismissed as random error, and too fundamental to be accommodated by relaxing the normative system.”(167) His argument begin with the claim that the “modern theory of decision making under risk emerged from a logical analysis of games of chance rather than a psychological analysis of risk and value.” (167). Once they have established a basis for the lack of descriptive adequacy for RCT, they continue to show numerous ways that agents violate RCT laws when acting. For example, they are risk adverse with proportionally large sums (to their net worth) and risk hungry with proportionally small sums.

They also point out that the psychological question of the framing of a question, irrelevant form the perspective of a simple rationality model, makes a very large difference. They give an example taken from McNeil et al, 1982. Given the choice between:

Problem 1 (survival frame)

Surgery: Of 100 people having surgery 90 live through the post-operative period, 68 are alive at the end of the first year, and 34 are alive at the end of five years.

Radiation Therapy: Of 100 people having radiation therapy all live through the treatment, 77 are alive at the end of one year, and 22 are alive at the end if five years.

Problem 1 (mortality frame)

Surgery: Of 100 people having surgery 10 die during surgery o the post operative period, 32 die by the end of the first year, and 66 die by the end of the five years.

Radiation therapy: Of 100 people having radiation therapy, none die during treatment, 23 die by the end of one year, and 78 die by the end of five years.

The difference in presentation produced markedly different responses. Those choosing radiation therapy rose form 18% in the survival frame (N=247) to 44% in the mortality frame (N=336). This result held steady whether the question was asked to clinical patients or physicians. Yet the two frames give the same substantive information.

Tversky and Kahnemon’s claim is simple: the theory is very inaccurate, there are theories that seem more descriptively accurate in many ways, such as prospect theory. Unfortunately, Cartwright would have little to say to Tversky and Kahnemon. They have pointed out inaccuracies in the theory, which is completely consistent in a Cartwrightian understanding of model building. They have pushed for research into other models which would fit the picture better. Nancy Cartwright would not object to such attempts to find a better model with which to explain human action, as long as it was causally accurate. The rejection of these critics of RCT makes it difficult to deal with. But using the necessary consequences of Cartwright’s methodological approach, whatever explanations that these critics embrace will also have the characteristics that they originally criticize (though, arguably, less of them) – inaccuracy and invalid assumptions. “Fundamental laws are meant to explain, and paradoxically enough the cost of explanatory power is descriptive adequacy. Really powerful explanatory laws of the sort found in theoretical physics do not state the truth.” (3)

Nancy Cartwright’s methodological approach to science has had a large impact on how she views physics. I have focus on the following question in my exploration of RCT with Cartwright’s approach in mind: how does the idea that a simulacrum mediate the application of scientific laws to reality answer certain objections to Rational Choice Theory or at least recast the issue? RCT seems a particularly fertile simulacrum to apply this analysis to. Its critics have complained primarily of its inaccuracy in prediction and its false assumptions. I have split RCT’s critics into two groups, the first having two sub groups. Within the first groups are those who complain of the theories unrealistic predications and often attempt to modify the theory using empirical psychological principles and those who complain about the theories unrealistic assumptions. In the second are those who reject the theory, either in principle or due to what they consider an unacceptable lack of explanatory power.

Cartwright’s simulacrum framework gives some room for inaccuracy and false assumptions within a model. In fact, these are required of any truly good causal models. She never adequately describes what is an acceptable and unacceptable level of inaccuracy and false assumptions, clearly a relevant issue. The closest she comes to this is when asking is abstract anti-particles matter exists as posited in quantum theory. She suggests we “focus on the causal roles which the theory gives to these strange objects: exactly how are they supposed to bring about the effects which are attributed to them, and exactly how good is our evidence that they do so? The general success of a theory at producing accurate predictions , or at unifying what before had been disparate, is not help here.” (8) Further, “we construct different models for different purposes… no single model serves all purposes best.” (11). Perhaps one model is rough predictively than another, but uses simpler equations that are easier to solve. It is not clear which is preferred, or, for that matter, even which is more causally accurate.

Nevertheless, Cartwright’s simulacrum account, if accepted, forces critics to take an extra step. No longer is descriptive inaccuracy sufficient to throw out RCT, now an account of the function of RCT is necessary to provide some evidence that RCT yields unacceptable error or inappropriate false assumptions. A project to explore the function and the domain of RCT would be the only way to provide an answer to these questions.

If one believes Cartwright’s analysis, the sciences have been unified. The difference between “hard” sciences such as physics of biology and “soft” social sciences such as economics and sociology has been reduces to token ones, differences of degree, rather than type ones, differences in kind. Phenomenological laws exist as the best descriptive and predictive accounts of very local behaviors. Fundamental laws are causally connected to the simulacrum, and are true of the simulacrum. But they are not causally connected to, nor true of, reality. The purpose of the simulacrum, of course, is to approximate reality in a useful way by drawing out the relevant aspects of reality and modifying them. This is as true of the theory of gravity as it is of rational choice theory. For critics to point to universal aspects of a science in RCT, without marshaling further arguments, is, for under Cartwright’s analysis, inadequate. Perhaps a closer methodological analysis if RCT will reveal whether a scientific approach is appropriate for this discipline, and whether RCT is the best, or even an adequate, model in this domain. But before such a project takes off, the preceding analysis might suggest that a closer study of methodology might benefit economists – both theory builders and critics alike.

The proceeding study does illustrate one thing; methodology counts. Cartwright’s conception of a simulacrum in science profoundly affected the criticisms leveled against RCT, and also, perhaps, affected the perception of the rational choice theory model itself. Further, even such questions as the status of economics as a science were shown to be potentially affected by this sort of analysis. Such considerations as the epistemological status of fundamental laws, and the relationship between fundamental laws and reality, though they might seem specious and academic to some economists, are vital to the understanding of any science, and profoundly affect our understanding of it as we have seen in the case of economics as a whole as well as rational choice theory more specifically.