Abstract
When we ask a decision maker to express her preferences, it is typically assumed that we are eliciting a pre-existing set of preferences. However, empirical research has suggested that our preferences are often constructed on the fly for the decision problem at hand. This paper explores the ramifications of this empirical research for our understanding of instrumental rationality. First, I argue that these results pose serious challenges for the traditional decision-theoretic view of instrumental rationality, which demands global coherence amongst all of one’s beliefs and desires. To address these challenges, I first develop a minimal notion of instrumental rationality that issues in localized, goal-relative demands of coherence. This minimal conception of instrumental rationality is then used to offer a more sophisticated account of the global aspects of instrumental rationality. The resulting view abandons all-or-nothing assessments of rationality and allows us to evaluate decision makers as being rational to varying degrees. My aim is to propose a theory that is both psychologically and normatively plausible.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
While it will not affect my discussion, I will for the sake of concreteness take for granted the standard set of structural requirements on preferences articulated in Savage (1972). See Part I of Anand et al. (2009) for a survey of alternative requirements. I will also be adopting the realistic as opposed to formalistic interpretation of decision theory on which an expected utility function is a measure of beliefs and desires rather than an indicator function that offers a “definitional reformulation” of the DM’s preferences (see pp. 144–146 in Hansson 1988). Of course, the DM may possess “brute” preferences that are inconsistent with those entailed by her beliefs and desires. However, to avoid confusion between these two ways of talking about preferences, I will avoid talk of such brute preferences.
I will be assuming the separability of belief and desire. See Levi (1999) and Jeffrey (1965) for opposing sides of this debate.
I use the term ‘judgment’ only to highlight that these attitudes are used to deliberate. I am not assuming that they are conscious, under the agent’s control, or truth-evaluable in the case of desires.
“The overarching goal of normative decision theory is to establish a general standard of rationality for the sort of instrumental (or “practical”) reasoning that people employ when trying to choose means appropriate for achieving ends they desire” (Joyce 1999, p. 9).
Since it is typically assumed that one’s standing mental states ought to be coherent, this thesis is implicitly accepted in discussions of rational coherence. For example, Ralph Wedgwood writes, “Instrumental irrationality crucially consists in this combination of mental states” (Wedgwood 2011, p. 290) No explicit reference to the deliberative context is made.
For example, Maurice Allais writes, “it cannot be too strongly emphasized that there are no criteria for the rationality of ends as such other than the condition of consistency. Ends are completely arbitrary” (Allais 1979, p. 70).
“In conventional economic theory, preferences are taken as primitive. A fundamental assumption is that preferences are independent of the tasks that an agent faces...It is crucial for many standard methods of economic analysis and implicit both in traditional models, such as expected utility theory, and in generalisations of them” (Cubitt et al. 2004, p. 709). Hausman (2012) offers an in-depth argument that the preferences appealed to in formal economic theory should be understood as context-independent all-things-considered judgments.
This study comes from Fischer et al. (1999).
This inference assumes that subjects do not prefer dominated choices. Cubitt et al. (2004) tested this assumption and found that violations of dominance could not explain preference reversals.
See Part II of Lichtenstein and Slovic (2006).
(Shafer (1986), pp. 464–465).
This argument does not rest upon a specific view of rational choice. Cubitt et al. (2004) show that for certain types of preference reversals, the subject’s expressed preferences across different modes of elicitation cannot be represented by any context-independent choice function.
“Practical reason, it might be suggested, is a holistic enterprise, properly concerned not merely with identifying means to the realization of individual ends, but with the coordinated achievement of the totality of an agent’s ends.” To offer just a small sampling, Wallace (2008), Fantl and McGrath (2002), Wedgwood (2007), Gibbard (2009), and Broome (2013) either endorse or appeal to an all-things-considered notion of rationality.
One might object that DMs are capable of grasping complete states and consequences by doing so in a piecemeal fashion. For example, a DM may be capable of considering states that describe both the size and color of an object by independently considering answers to questions about the size and color of the object. However, this way of considering the details of a state or consequence is not sufficient for accomplishing the DM’s deliberative task. Allow me to briefly summarize my response to this objection. For a more detailed version of the argument, see Kim (2012). As the preference reversal phenomenon has shown, the preferences of human agents can vary depending upon how we ask questions. Thus, the DM’s belief about a state may differ depending upon whether she independently considers the potential size and color of an object or considers the size and color together. In the case of belief, the reason for this instability is simple. The DM may believe that there is a correlation between the object’s size and color and so her beliefs may vary having considered these two questions together rather than having considered these two questions independently. Similarly, the DM’s desires may differ depending upon whether she considers answers to questions about what will happen to her independently or all at once. The reasons for the instability of desire come not from correlations between states of affairs but from the complex ways our values interact. For example, it is not obvious how to combine a ranking of jobs in terms of salary with a ranking of jobs in terms of location. To do so, one must consider at once the salary and location of a job, which means that one must be able to have thoughts whose contents are fine-grained enough to describe both salary and location. Therefore, to consider a complete state or consequence, one must be able to have thoughts whose contents are fine-grained enough to describe the answer to every possible question all at once. My claim here is that these contents are too fine-grained to be the contents of human thoughts.
These criticisms and the positive proposal that follows may be situated within the bounded rationality research program first articulated by Herb Simon. “Broadly stated, the task is to replace the global rationality of economic man with a kind of rational behavior that is compatible with the access to information and the computational capacities that are actually possessed by organisms, including man, in the kinds of environments in which such organisms exist” (Simon 1955, p. 99).
I thank an anonymous reviewer for this suggestion.
As I have stated it, the demand seems questionable. Why should we be as coherent as humanly possible? What do we achieve by meeting this demand? In what sense are we doing better as we become coherent across more and more decision problems? And even if we are doing better, are we doing better in the relevant sense? Are we being more rational or are we exhibiting other virtues? One way to make sense of this demand would be to claim that we were getting closer to being all-things-considered rational. However, as I argued in the previous section, the notion of all-things-considered rationality does not apply to human agents.
Quine (1951).
(Hansson (1975), p. 179). Savage famously noted that “it is even utterly beyond our power to plan a picnic or play a game of chess in accordance with [the principles of rational choice] even when the [set] of states and the set of available acts to be envisaged are artificially reduced to the narrowest reasonable limits” (Savage 1972, p. 16). He attempted to deal with this problem by appealing to small worlds, and the following discussion can be seen as an attempt to explicate what small worlds must be. There are other attempts to address this problem of application (cf. Gilboa and Schmeidler 2001). Most notably, Binmore (2008) attempts to offer a minimal extension of Bayesian decision theory to cases in which the DM cannot account for everything of relevance. Nevertheless, Binmore agrees with my conclusion that neither all-things-considered rationality nor all-things-relevant rationality are viable accounts of human rationality. His positive proposal can be seen as an alternative to the one I propose below. However, given the scope of this paper and its focus on incorporating the lessons from the empirical research on preference reversals, I will not discuss his proposal.
In order for a decision theory to be applicable to humans, it need not provide a decision procedure. Instead, it must simply govern deliberative activities that we are capable of engaging in.
DMs must also identify which acts are available to perform. This is an important parameter of a decision problem but I shall ignore it for our discussion.
Two interpretations of this view present themselves. On the strong interpretation, for the multitude of decision problems the DM could face, she ought to possess rational preferences for each. On the weak interpretation, the DM only ought to possess rational preferences for the decision problem of current concern. As my discussion in Sect. 3.2 will show, I think there are various ways of assessing a DM’s rationality. Therefore, being instrumentally rational cannot be reduced to one or the other interpretation.
“There must be such [principles], if rationality is to constrain practical preferences in any way” (Broome 1995, p. 105; Anand 1995; Schick 1991, 1997) also discuss related issues. Bermudez (2009) has a long discussion of individuating consequences but is interested in a different inquiry about whether we can make compatible the internal and external perspectives of decision theory.
Broome thinks that there are “rational requirements of indifference, which limit the fineness of individuation that is allowed” (Broome 1995, p. 104). And “it is rational to have a preference between two alternatives only if they differ in some good or bad respect” (Broome 1995, p. 106). I will appeal to the DM’s values to demarcate consequences but no external criteria of goodness will constrain which values may be relevant.
Broome’s conclusion depends upon the assumption that there is always a way of ensuring the coherence of preferences by simply individuating the relevant acts in a more-fine grained way. However, this assumption is questionable. Suppose that we have acts A’, B’, and C’ that are respectively assumed to be refinements of acts A, B, and C. As Savage (1972) argued, coherence demands that the DM’s preferences over the first set of acts mirror the DM’s preferences over the second set. Therefore, if the DM has a set of transitive preferences over the first, but a set of intransitive preferences over the second, we cannot restore coherence by finely individuating the latter set of acts—so long as we assume that the latter are indeed refinements of the former.
Following Savage, consequences can be understood as states of the agent, and they pick out what the DM’s experiences as a result of her actions. So these properties are instantiated by the DM.
The proposed view of goals, values, and properties is influenced by the discussion in Chap. 2 of Raiffa and Keeney (1976).
For a discussion of desires as values for properties as opposed to states of affairs, see Pettit (1994).
Using constant acts—acts that result in the same consequence regardless of what the world is like—we can talk about preferences for consequences by appealing to our preferences between acts.
While the higher-order desires and meta-preferences discussed respectively by Frankfurt (1971) and Sen (1977) can function as goals, they are not the only attitudes that can play this role. In fact, any attitude, judgment, or combination of attitudes and judgments that can select a set of relevant values and determine one’s preferences over the relevant consequences can function as a goal.
For simplicity, I’ve assumed that a goal entails a unique rationally permissible set of complete preferences over the consequences. This assumption can be weakened.
The adoption of a goal is not always the best means of achieving it. For example, Steve may be his worst enemy, perpetually engaging in self-sabotage. Knowing this about himself, he may know that by aiming to do A, he will most likely not achieve A. Though more discussion is needed, the simplest reply is that self-justifying goals are typically but not always available to rational agents. And there may be something pathological about agents who cannot adopt self-justifying goals.
Of course, it may be difficult to identify what these different goals might be. An interesting case study comes from a published interview (see Lichtenstein and Slovic 2006, pp. 65–68) where the subject recognizes some inconsistency between her preferences but refuses to change them. Given her own reports, it appears that she justifies this inconsistency by taking herself to be involved in different tasks with different aims in the two decision problems.
The appeal to third-order coherence can be used to explain the one situation in which preference reversals clearly disappear. These are cases in which subjects are forced into market arbitrage (Chu and Chu 1990). This simply means that they are put through a money pump. In these cases, if the subjects do not want to lose money, they are forced to engage in the third-order decision problem and must adopt a coherent strategy for evaluating bets across various decisions problems.
While it would go beyond the scope of this paper to offer a detailed discussion, I’d like to make a small remark about the view of preference found in Hausman (2012). Hausman argues that the preferences appealed to in formal economic models are best understood as stable, total comparative evaluations. These are all-things-considered judgments. He notes, however, that the preferences we reveal in our actual choice behavior are typically context-sensitive. Hausman proposes that we can respond in two ways. First, we can separate choice from preference and develop theories of choice that do not appeal to total comparative evaluations. Second, we can retain the connection between choice and preference if we adopt a different view of preference. It is typically thought that by adopting the second way, our theories of actual choice are divorced from theories of rational choice. While I have attempted to take the latter route, I also hope that the proffered view of instrumental rationality will be able to retain some connection between theories of rational choice and theories of actual choice.
For example, Bermudez writes that “we need a way of individuating outcomes that will determine in an absolute sense whether or not the action is rational” (Bermudez 2009, p. 103). I am suggesting that we abandon the view that decision theory provides an absolute standard of rationality.
Chesterton offers a poetic argument against flip-flopping, characterizing flip-floppers as people of decadence whose vice is that they desire to be many different people. What is interesting about Chesterton’s discussion is that even though he argues that flip-floppers fail to experience the joys that can only come from unwavering commitment, he nevertheless admits that the view of the decadent “is a perfectly possible, rational and manly position” (Chesterton 2012, p. 12).
References
Allais, M. (1979). The foundations of a positive theory of choice involving risk and a criticism of the postulates and axioms of the American School (1952). New York: Springer.
Anand, P. (1995). Foundations of rational choice under risk. Oxford: OUP.
Bermudez, J. L. (2009). Decision theory and rationality. Oxford: Oxford University Press.
Binmore, K. (2008). Rational decisions. Princeton: Princeton University Press.
Broome, J. (1995). Weighing goods. Oxford: Wiley-Blackwell.
Broome, J. (2013). Rationality through reasoning. New York: Wiley.
Chesterton, G. K. (2012). The defendant. New York: Dover Publications.
Chu, Y. P., & Chu, R. L. (1990). The subsidence of preference reversals in simplified and marketlike experimental settings: A note. The American Economic Review, 80(4), 902–911.
Cubitt, R. P., Munro, A., & Starmer, C. (2004). Testing explanations of preference reversal. Economic Journal, 114, 709–726.
Fantl, J., & McGrath, M. (2002). Evidence, pragmatics, and justification. The Philosophical Review, 111(1), 67–94.
Fischer, G. W. et al. (1999). Goal-based construction of preferences: Task goals and the prominence effect. Management Science, 45(8), 1057–1075.
Frankfurt, H. G. (1971). Freedom of the will and the concept of a person. Journal of Philosophy, 68(1), 5–20.
Gibbard, A. (2009). Thinking how to live. Cambridge: Harvard University Press.
Gilboa, I., & Schmeidler, D. (2001). A theory of case-based decisions. Cambridge: Cambridge University Press.
Hansson, B. (1975). The appropriateness of the expected utility model. Erkenntnis, 9, 175–193.
Hausman, D. M. (2012). Preference, value, choice, and welfare. Cambridge: Cambridge University Press.
Jeffrey, R. (1965). The logic of decision. Chicago: University of Chicago Press.
Joyce, J. M. (1999). The foundations of causal decision theory. Cambridge: Cambridge Univ Press.
Kim, B. (2012). The context-sensitivity of rationality and knowledge. PhD Thesis. Columbia University.
Lichtenstein, S., & Slovic, P. (1968). Relative importance of probabilities and payoffs in risk taking. Journal of Experimental Psychology, 78(3), 1–18.
Lichtenstein, S., & Slovic, P. (1971). Reversals of preference between bids and choices in gambling decisions. Journal of Experimental Psychology, 89(1), 46–55.
Lichtenstein, S., & Slovic, P. (2006). The construction of preference. Cambridge: Cambridge University Press.
Lin, H. (2014). On the regress problem of deciding how to decide. Synthese, 191(4), 661–670.
Pettit, P. (1994). Decision theory and folk psychology. In M. Bacharach & S. L. Hurley (Eds.), Essays in the foundations of decision theory (Chap. 4, pp. 147–175). Oxford: Blackwell.
Quine, W. V. (1951). Main trends in recent philosophy: Two dogmas of empiricism. The Philosophical Review, 60(1), 20–43.
Raiffa, H., & Keeney, R. (1976). Decisions with multiple objectives: Preferences and value tradeoffs. New York: Wiley.
Schick, F. (1991). Understanding action: An essay on reasons. Cambridge: Cambridge University Press.
Schick, F. (1997). Making choices: A recasting of decision theory. Cambridge: Cambridge University Press.
Sen, A. K. (1977). Rational fools: A critique of the behavioral foundations of economic theory. Philosophy & Public Affairs, 6(4), 317–344.
Shafer, G. (1986). Savage Revisited. Statistical Science, 1(4), 463–485.
Shafer, G., & Tversky, A. (1985). Languages and designs for probability judgment. Cognitive Science, 9(3), 309–339.
Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99–118. ISSN 0033-5533.
Smith, H. (1991). Deciding how to decide: Is there a regress problem? In M. Bacharach & S. Hurley (Eds.), Essays in the foundations of decision theory. Oxford: Blackwell.
Velleman, J. D. (1997). Deciding how to decide. In G. Culitty & B. Gaut (Eds.), Ethics and practical reason (pp. 29–52). Oxford: Oxford University Press.
Wallace, J. (2008). Practical reason. http://plato.stanford.edu/entries/practical-reason.
Wedgwood, R. (2007). The nature of normativity. Oxford: Oxford University Press.
Wedgwood, R. (2011). Instrumental rationality. Oxford Studies in Metaethics, 6, 280–309.
Acknowledgments
For their thoughtful comments and objections, I would like to thank Achille Varzi, Anubav Vasudevan, Dana Howard, Don Hubin, Glenn Ross, Guillermo Del Pinal, John Brunero, John Collins, Katie Gasdaglis, Mark Alfano, Sigrun Svavasdottir, and two anonymous referees for Synthese. I also thank the audiences at Kansas State, Columbia University, the Decisions, Games & Logic Workshop, and the Canadian Society for Epistemology for their comments on earlier versions of the paper
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Kim, B. The locality and globality of instrumental rationality: the normative significance of preference reversals. Synthese 191, 4353–4376 (2014). https://doi.org/10.1007/s11229-014-0529-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-014-0529-8