Don't be an ass: rational choice and its limits.
If deciding is akin to taking a leap, then deciding rationally is tantamount to ensuring that one leaps over as narrow a chasm of uncertainty as possible. One contemplates a range of possible actions, assigns weights and probabilities to each, and then calculates which is likely to best serve one's elected ends. Such deliberation is often seen as the site of human freedom, but the binding power of rationality does seem to imply that deliberation is, in its own way, a deterministic process. After all, if one knows the starting preferences and circumstances of an agent, then, assuming that the agent is rational and that those preferences and circumstances don't change, one should be in a position to predict what the agent will decide. However, given that an agent could conceivably confront equally attractive alternatives, it is an open question whether rational choice theory can ever eliminate indeterminacy--fully bridging the chasm, as it were.
The clearest support for such a limitation of rationality comes from the "Buridan's ass" scenario, where an agent is confronted with two (or more) equally attractive/unattractive options. The famished ass stuck between equidistant bales of hay is of two minds on the matter. Its eventual action, if any, will of course attest to a singular commitment. But, it seems there must be a prior bottleneck, and it is what transpires in this anteroom of agency that is disputed.
"Choice" can be defined as something that "presupposes alternatives plus a requirement that an outcome be reached in favor of one of them to the exclusion of the other." (1) Call this last clause the univocity assumption. This assumption reveals some common methodological aspirations. Indeed, "the explanatory ideal in science is always to form hypotheses from which a unique observational consequence can be deduced." (2) Buridan's ass cases compromise this desire always to churn out a univocal verdict.
Does rationality by itself have the resources needed to prevent paralysis of action? Differing answers to this question obviously entail differing research programs. One view asks us to note that rational agents are endowed with a power to "just do it," (3) while the other view asks us to collect still finer empirical details about situated exercises of rational agency. If, however, there can remain genuine leeway in the choices of even the most rational agents, this would undermine the prospect of predicting and/or guiding decision-making processes in a totally gap-free way. A modicum of voluntarism must, it seems, always be in the mix.
As we shall see, those who (implicitly or explicitly) adhere to the univocity assumption cannot accept the decisional impotence one finds in Buridan's ass cases, and so devise ways to avoid it. I will catalogue common responses and argue that each are either unwarranted or flawed. Obviously, the philosophical criticisms I will articulate need not challenge the specifics of decision theory--though they might pose a cap on its range of application. My presentation will follow a straightforward structure: I will first pinpoint the problem that concerns me, and will then examine four untenable responses to it.
2. The Problem
In this article, I want to focus solely on what is essential. Mark Balaguer, for instance, distinguishes Buridan's ass scenarios from "torn" decisions. (4) The former involve qualitatively identical but numerically distinct options (e.g., two soup cans of the same brand), whereas the latter involve qualitatively different and numerically distinct options (e.g., a soup can versus a banana). One can certainly make this distinction. However, doing so is needless, since the assignment of an abstract valence like "utility" is, by design, general enough to subsume such features. Hence, it makes no difference what one is stuck between, provided one desires each equally. We simply have to play with the variables until they truly even out, at which point the discussion begins (though to foster clarity, I will nevertheless use qualitatively identical options as my examples).
If we disregard the fact that it involves a non-human animal, the problem of "Buridan's ass" poses no great difficulty for the imagination (at least not obviously so). The "weights" that are balanced are not physical weights, but preferential weights. Hence, a trace element of physical matter added to or withdrawn from a given option does not automatically translate into an increase or decrease in the utility assigned to that option. Moreover, no controversial (say, supernatural) elements are posited. In fact, some theorists like Edna Ullmann-Margalit and Sydney Morgenbesser hold that cases of symmetrical preference abound, and propose that "[s]uper-market shelves supply us with paradigmatic examples of social picking situations proper." (5) Of course, a lot of money is spent by firms for "featuring" their products by placing them in more conspicuous spots. Nevertheless, it does not offend the laws of physics one bit to think that two soup cans (on whatever shelf) might be equidistant from a customer. To the extent that this is correct, Buridan's story might be a ubiquitous part of our daily lives, and we should be intimately familiar with our decisions in such cases.
Rational animals would be at a severe disadvantage if they had not evolved ways of wriggling out of decisional paralysis (otherwise their species-specific differentia would be a considerable hindrance). Can we proceed from this to the conclusion that human agents are endowed with a faculty or power that escapes the net of traditional decision matrices? Ullmann-Margalit and Morgenbesser introduce the terms "picking" and "choosing" to describe decisions made in symmetrical and asymmetrical preferential contexts, respectively. "Picking" is non-monotonic, in the sense that one cannot deduce on the basis of an agent's prior commitments and context what the outcome of her rational deliberations will be. Hence, when it comes to picking, the use of the adjective "rational" is moot.
Positing the existence of a faculty like picking would seem to be a perfectly respectable inference to the best explanation. Interestingly, before Buridan and his Latin peers discussed the matter, al-Ghazali had formulated the problem of preferential symmetry using a man stuck between two equally mouth-watering dates. Al-Ghazali concluded that "[e]veryone, therefore, who studies, in the human and the divine, the real working of the act of choice, must necessarily admit a quality the nature of which is to differentiate between two similar things." (6) More often than not, though, this line of abductive reasoning is not carried out, as a certain bias manifests itself.
Philosophy as a distinctive activity is often said to rest on deliberation, so the suggestion that the ambit of those rational powers is limited can naturally be seen as tainting that disciplinary identity. Many philosophers thus assume that, given well-defined circumstances, a theory of rational choice can always tell us what to do. This assumption in turn fosters (or is fostered by?) a general confidence in the exhaustive power of reason. There is a distinctively rationalist flavor to this belief. Rene Descartes, in his Fourth Meditation, offers a canonical statement:
But the indifference I feel when there is no reason pushing me in one direction rather than another is the lowest grade of freedom; it is evidence not of any perfection of freedom, but rather of a defect in knowledge or a kind of negation. For if I always saw clearly what was true and good, I should never have to deliberate about the right judgement or choice; in that case, although I should be wholly free, it would be impossible for me ever to be in a state of indifference. (7)
Pursuing that view, Benedict de Spinoza (8) thought it would be flatly irrational for anyone to regard two options as truly equal, and Gottfried Leibniz (9) thought that decision without preference offends the (in his view, ubiquitous) principle of sufficient reason. Not everyone in the philosophical canon thinks that the principle of sufficient reason has exhaustive coverage, so some (like Arthur Schopenhauer (10) and Jean-Paul Sartre (11)) would find no problem here. Still, theirs is a minority view, so the rationalist stance dominates, informing current debates. (12)
Of course, it is rare to find a theorist or philosopher openly affirming that rational decision knows no bounds. Nevertheless, when confronted with such limits, many write them off as merely apparent. Bruno de Finetti, for example, resorts mainly to (supposedly shared) intuitions to motivate his claims. He characterizes the idea that the world can house cases which have "no feature that would make one preferable to the other" as something that "puts nature in the terribly embarrassing situation of Buridan's ass." (13) By parity of reasoning, one could just as easily say that strong dominance puts nature in the situation of making the choice for the agent.
3. Four Untenable Responses
The world likely admits of a whole range of balances and imbalances, so it is difficult to see why one situation should be deemed more metaphysically absurd than the other. At any rate, the quick fixes encouraged by such a mindset are riddled with difficulties that are more significant than is typically assumed. I will now look at four common strategies.
a. Postulating a neutral valence or state
We can begin with the least sophisticated response. Situations that appear problematic can be made less so by including "indifference" into the calculus. After all, since no recommendation can be inferred from indifference over and above the idea that both options are equally good and could both be picked, this seems like a perfectly sound analysis.
Alas, this response completely dodges the problem. Given that in the end an action will be taken, we have to explain why a specific option was privileged. Clearly, the idea of indifference cannot be of any help here; it is coined to express a state which may perhaps precede an action, but surely cannot prompt or accompany an action. Hence, it is legitimate to ask whether it was a utility or something else that put an end to the indifference.
b. Tipping the utilities
If one endorses a twofold menu of utilities and indifference, then by virtue of a disjunctive syllogism, the observable absence of indifference licenses the inference of utilities. Another response therefore consists in doctoring a single optimum in re for each closed context and attributing all of the remaining rational indeterminacy to a straightforward lack of information on the part of the deciding agent. For de Finetti, equal cases are "only cases that differ in respects that are either unknown or causally unrelated to their happening." (14) Such unknown differences allow one to brush aside challenges to the univocity assumption.
The inherent shortcomings of measurements provide a ready asylum for the presumed impossibility of preferential symmetries, as many (or most) of the differences that could move an agent to prefer one option over another fall below a threshold of discernment. Given that minuscule differences in the weight of soup cans go undetected at point-of-purchase by a human ceteris paribus, one (fairly unsubtle) way to safeguard the univocity assumption is to interrupt a customer midway and inform her that the selection she was about to make is in point of fact less desirable, say, on account of its slightly lesser weight. However, not only would such an intervention doctor the situation in a question-begging way, it would violate the normal phenomenology of the event and thereby relinquish the claim that what is being modeled is the actual decision-making process of the agent(s).
The point can be stated in a methodological key. Regardless of one's stance on the question of complete preferential symmetries, decision theory can hope to yield verdicts only if the alternatives, expected utilities, and predicted probabilities are kept finite and constant. All parties to the debate agree that the input data must at some point be frozen into place, at least for a given time-slice. To be sure, the social scientist or economist can always leave the observation booth, as it were, and actively intervene in the situation under scrutiny. Doing so, however, would contaminate the results on any gloss. So, while it can certainly be interesting to complicate an experimental design by allowing for a transparent feedback loop that permits agents to revise their forecast of a contingent future in the light of new third-person information about their conduct, (15) the difficulty I am interested in is that which remains once all of these bells and whistles have been added. In other words, a supermarket customer can be informed of the weight of competing soup cans to as many decimal points as an experimenter wishes, but the relevant situation emerges when the quantitative match is perfect or she simply stops caring, whichever comes first.
It could perhaps be replied on behalf of the interventionist strategy that, all other things being equal, an agent would want to know as many decimal points as possible. After all, charity recommends that we try to maximize the rationality of the agent whose antics we are interpreting. (16) According to what may be termed ideal conditions theory, "[i]f we have any reason to think that the agent is operating with partial or misleading information ... then we should not take the choices that they make as revealing their 'genuine' preferences." (17) Isaac Levi expresses this same idea when he writes that "failure to live up to the commitment is excusable insofar as it is due to lack of memory and computational capacity or to emotional disturbance," such that ascriptions of irrationality or incoherence "ought to be reserved for those who persist in violating logical closure even in the absence of such excuses." (18) Hence, on this view, if one happens to choose the lighter soup can, one is not "knowingly" going against the calculus of her utilities.
The difficulty with this seemingly benign gloss, however, is that agents always operate with partial information and imperfect circumstances (as any philosophical skeptic will gladly demonstrate). This is problematic, since there is no standard by which to gauge when to halt the data-gathering. By searching for further facts, descriptive accounts of human decision can thus covertly partake in the intolerance of undecidability.
Faced with this, one might insist that "we should seek to narrow the gap between commitment and performance by improving the technologies which enhance our reasoning and computational capacities." (19) I agree that it is generally laudable to foster one's rational skills. It is debatable, though, whether a lack of preference for two identical soup cans betokens a lapse in reason. To be sure, one can stipulate that preferential symmetries are to be taken as a sign to harvest more informational grist for the rational mill. Yet, "[g]iven a rule or a requirement, we can ask whether you ought to follow it, or whether you have a reason to do so." (20) Even if we accept for the sake of the argument that rationality is "a medal of honor bestowed upon certain decision makers by decision theorists," (21) it is not clear why an agent should receive demerit points for being genuinely perplexed (and/or breaking with her symmetrical preferences "just because").
c. Positing sub-personal influences
Another way to protect univocity is to hold that the minute discrepancies of available options are not consciously accessible but nevertheless exert a "subliminal" force on agents. In essence, one can tip the utilities, unbeknownst to the decider.
Despite criticizing decision theory for being narrower in scope than typically assumed, Jon Elster succumbs to this strategy. There are many moments in Elster's critique of standard accounts where the need for a novel non-rational element suggests itself, yet he opts to pursue explanations that do not challenge a stimulus-response model. He observes, for instance, that even when a situation is such as to present an agent with several equally weighed options that leave no room for rational choice, the agent will nonetheless retain the power "at least to 'pick' one of the options." (22) Since countenancing such a faculty has far-reaching implications, Elster effectively dodges the commitment by introducing more environmental factors like "perceptual salience or some other value-neutral feature of the situation [that] led to one option rather than another being 'picked'." (23) In this way, sub-personal influences are posited in order to avoid the potentially unnerving implications of recognizing a different kind of decision-making power.
Such an appeal misunderstands Ullmann-Margalit and Morgenbesser's original notions, insofar as Elster cannot add a "causal supplement" without adulterating "picking" and transforming it into a "choosing." It is as if, upon noticing the plain fact that the ass in Buridan's example will nevertheless choose one source of food, the philosopher uncomfortable with challenging determinist models of the mind would prefer invoking the time of day that bestowed on one sunlit bowl a perceptual salience, rather than positing some causa sui ("self-caused") capacity. One can always add epicycles to save a theory. It is debatable, though, whether calling on situational minutia to break Buridan's stalemate is more plausible than accepting a supplementary faculty like picking, which most subjects would likely report possessing. (24)
It could be argued that the very idea of unknown utilities is suspect or incoherent, because genuine agency requires an ability to give reasons for what one does. (25) Unless one wants to convert decisions into mere bodily happenings, unknown utilities cannot be unknowable utilities. (26) In any event, the proponent of sub-personal influences is in need of an argument to show why and how something inaccessible to consciousness can nonetheless contribute to tilting the balance of an agent's decision. Such a claim is notoriously difficult to establish (think of the many posits of Freudian psychoanalysis).
This is not to deny the existence and causal efficacy of "subliminal" influences, for which there is undoubtedly experimental support. However, invoking the possible presence of such influences is a plausible strategy only in some situations. If injected into the situation of perfect balance that concerns me, it constitutes a change of topic.
Those worried about manipulative marketers (or well-meaning "nudgers") would do well to consider that, since the problem at hand is a general one, an agent could in principle be frozen before soup cans of the same brand, which would hardly be conducive to purchase. In other words, subliminal pulls could conceivably tug an agent evenly in opposing directions, jointly prompting incompatible courses of action with an equal degree of psychological force.
In a perfect preferential symmetry, the machinery of rationality is brought to a standstill because the set of options it confronts are deprived of any ordinality. In a bid to restore mock preferences, the subliminal retort posits a causal story but plunges it into murky waters, with the convenient assumption that there must be some account to be told in this regard--only we'll never know it. I fail to see how this appeal is more epistemologically responsible or ontologically parsimonious than countenancing an ability simply to "pick" in a subset of cases.
d. Bunching the options
We have now seen a variety of ways to protect the univocity assumption and explain away the troublesome prospect of confronting preferential symmetries. All of those responses have important flaws. Hence, instead of privileging analysis and increasing the pixellation to something more fine-grained, one could privilege synthesis and reduce the pixellation to something coarser-grained. Elster mentions that "one might redefine the choice situation by bunching the top-ranked alternatives into a single option." (27) The idea, in sum, is that "If I am indifferent between a red umbrella and a blue umbrella, but prefer both to a raincoat, the choice becomes determinate once we have bunched the first two options as 'an umbrella. '" (28)
Perhaps, but then this raises the following question: Why bunch these specifically? One might just as easily redefine the choice situation so that umbrellas and raincoats jointly become a single option, say, "Items that protect one from the rain." For the bunching proposal to go through, one would need a principled criterion for why the discrimination of alternatives should be blurred at specific categorization boundaries and not others. It simply won't do to say that the differently colored umbrellas are, as a class, preferred over the raincoat. The predicament is not that an agent stands before undifferentiated umbrellas and is plagued by ("akratic") failure to act in accordance with her preference for them. Rather, the situation is philosophically interesting (and problematic) precisely because she fully notices the distinction in color yet nevertheless remains patently indifferent to it.
In a sense, the bunching strategy attempts to treat macroscopic differences as the equivalent of unnoticed differences. Minute differences are indeed bunched, but in such cases the preferential indifference stems from a straightforward cognitive ignorance. It is platitudinous to say that one does not care about the complementary class of things one knows nothing of. (29) Hence, for the bunching strategy to succeed, it has to accord with the actual experiential situation of the agent, and therefore must acknowledge that the agent is indifferent with regard to two options, not one. Otherwise, one could just as well "bunch" a grocery store as a whole, since it is clearly preferred over starvation. Needless to say, that would not be very helpful. A rewrite into one "option" will thus remain ineffectual unless it can be shown to be more than merely ad hoc.
Factually, paralysis of action is not a pervasive phenomenon. This is either because (i) the utilities one assigns to two or more options can never be balanced or because (ii) thanks to some non-rational faculty (say, the will), we would not be stuck even if those utilities were perfectly counterpoised. Having looked at four untenable responses, it becomes apparent that (i) is often just a dogma and (ii) is by no means a silly position.
A table of alternatives and probabilities, whatever the complexity of the resultant grid, is a fairly benign construction. Moreover, standard accounts of rational choice openly omit to discuss where the input utilities come from. Hence, Ullmann-Margalit and Morgenbesser conclude that "we pick or we choose as the case may be; but as to our utilities or values themselves, to the extent they can be thought to be selected at all, they can only be picked." (30) If it does not matter where the agent's elected preferences come from, I do not think it should create a kerfuffle to acknowledge that, once refined in accordance with the canons of rationality, these utilities still allow for causa sui actions. Endorsing a threefold menu of utilities, indifference, and will, the observable absence of indifference no longer licenses the inference of tacit utilities. (31)
University of Helsinki
(1) James I. McAdam, "Choosing Flippantly or Non-Rational Choice," Analysis 25, suppl. no. 3 (1965), p. 133.
(2) Jon Elster, "The Nature and Scope of Rational-Choice Explanations," in Actions and Events: Perspectives on the Philosophy of Donald Davidson, ed. Ernest LePore and Brian P. McLaughlin (Oxford: Blackwell, 1985), p. 65 (emphasis added).
(3) See Marc Champagne, "Just Do It: Schopenhauer and Peirce on the Immediacy of Agency," Symposium: Canadian Journal of Continental Philosophy 18, no. 2 (Fall 2014), pp. 209-32.
(4) Mark Balaguer, Free Will as an Open Scientific Problem (Cambridge, MA: MIT Press, 2010), pp. 72-73.
(5) Edna Ullmann-Margalit and Sydney Morgenbesser, "Picking and Choosing," Social Research 44, no. 4 (1977), p. 761.
(6) Quoted in Nicholas Rescher, "Choice without Preference: A Study of the History and of the Logic of the Problem of 'Buridan's Ass,'" Kant-Studien 51, no. 2 (1959-60), p. 148.
(7) Rene Descartes, The Philosophical Writings of Descartes, vol. 2, trans. John Cottingham, Robert Stoothoff, and Dugald Murdoch (Cambridge, UK: Cambridge University Press, 1984), p. 40. For a similar view, see Brand Blanshard, Reason and Analysis (Chicago, IL: Open Court, 1991), p. 493.
(8) Spinoza doubts whether a person frozen in equilibrium would still count as a person. The topic of balanced utilities, though motivated by formal considerations, thus becomes the province of "babies, fools, and madmen." See Benedict de Spinoza, Complete Works, trans. Samuel Shirley (Indianapolis, IN: Hackett, 2002), p. 276.
(9) According to Leibniz, the scenario of Buridan's ass is typical of the medieval "Schoolmen, whose ideas"--unlike, say, his windowless monads--"tend towards the chimerical." Thanks to the activity of those Leibnizian monads, "small perceptions" (somehow) intervene to ensure that "|t]here is always a prevailing reason which prompts the will to its choice." See Gottfried Leibniz, Theodicy, trans. E. M. Huggard (LaSalle, IL: Open Court, 1985), p. 148. As we are about to see in the following section, this is a prime example of response (c), which gratuitously posits subpersonal influences.
(10) For Schopenhauer, determinism applies to everything we perceive, but there is no reason to think that everything we perceive exhausts everything there is. See Arthur Schopenhauer, Essay on the Freedom of the Will, trans. Konstantin Kolenda (Mineola, NY: Dover, 2005). See also Champagne, "Just Do It: Schopenhauer and Peirce on the Immediacy of Agency."
(11) For Sartre, consciousness always effaces itself before whatever intentional object it has. Since the conscious ego is nothing (literally, no thing), it cannot be subject to any law or causal force. To maintain otherwise would be to craft a cowardly excuse for one's freely elected stance/attitude toward the world. See Jean-Paul Sartre, Being and Nothingness, trans. Hazel E. Barnes (New York: Random House, 1994). This resembles the claim, "[i]n the Objectivist view, [that] no antecedent, deterministic factors can explain why people choose or do not choose to focus." See Chris Matthew Sciabarra, Ayn Rand: The Russian Radical, 2d. ed. (University Park, PA: Pennsylvania State University Press, 2013), p. 153. For a comparative study of existentialist and objectivist views on free will and commitment, see Marc Champagne and Mimi Reisel Gladstein, "Beauvoir and Rand: Asphyxiating People, Having Sex, and Pursuing a Career," The Journal of Ayn Rand Studies (forthcoming).
(12) It could be argued that, by moving from a human to an ass, the Western tradition (re)cast the thought-experiment in terms that rhetorically poison the well.
(13) Bruno de Finetti, "Probabilism: A Critical Essay on the Theory of Probability and on the Value of Science," Erkenntnis 31, nos. 2-3 (1989), p. 178.
(14) Ibid., p. 177.
(15) For an account of rational choice in the face of changing expected utilities, see Edward F. McClennen, Rationality and Dynamic Choice: Foundational Explorations (Cambridge, MA: Cambridge University Press, 1990).
(16) Dagfinn F0llesdal, "The Status of Rationality Assumptions in Interpretation and in the Explanation of Action," Dialectica 36, no. 4 (1982), pp. 301-16.
(17) Jose Luis Bermudez, Decision Theory and Rationality (Oxford: Oxford University Press, 2009), p. 57.
(18) Isaac Levi, "Possibility and Probability," Erkenntnis 31, nos. 2-3 (1989), p. 366.
(19) Ibid., p. 367.
(20) John Broome, "Is Rationality Normative?" Disputatio 2, no. 23 (2007), p. 162.
(21) Itzhak Gilboa, Rational Choice (Cambridge, MA: MIT Press, 2010), p. 6.
(22) Jon Elster, "The Nature and Scope of Rational-Choice Explanations," in Actions and Events: Perspectives on the Philosophy of Donald Davidson, ed. LePore and McLaughlin, pp. 65-66.
(23) Ibid., p. 66.
(24) Balaguer, Free Will as an Open Scientific Problem, pp. 89-91.
(25) See, for example, the cluster of views presented by Chauncey Maher in The Pittsburgh School of Philosophy: Sellars, McDowell, Brandom (New York: Routledge, 2012).
(26) Interestingly, when they address the topic of subliminal influences, Richard H. Thaler and Cass R. Sunstein require that institutional "choice architects" be prepared to give (presumably persuasive) reasons for the selections they privilege. See their Nudge: Improving Decisions About Health, Wealth, and Happiness (New Haven, CT: Yale University Press, 2008), pp. 244-45.
(27) Elster, "The Nature and Scope of Rational-Choice Explanations," p. 66.
(28) Ibid. For an example of this maneuver, see Hans-Hermann Hoppe, "A Note on Preference and Indifference in Economic Analysis," The Quarterly Journal of Austrian Economics 8, no. 4 (2005), pp. 90-91.
(29) See the discussion of neutral valences in Marc Champagne, "Axiomatizing Umwelt Normativity," Sign Systems Studies 39, no. 1 (2011), pp. 45-47.
(30) Ullmann-Margalit and Morgenbesser, "Picking and Choosing," p. 783.
(31) I would like to thank Muhammad Ali Khalidi, Mathieu Doucet, Susan Dimock, and audience members at the 2012 meeting of the Canadian Philosophical Association for feedback on an earlier version of this article.
|Printer friendly Cite/link Email Feedback|
|Date:||Mar 22, 2015|
|Previous Article:||Portraits of egoism in classic Cinema II: negative portrayals.|
|Next Article:||Philip Booth's (ed.) ... and the Pursuit of Happiness: Wellbeing and the Role of Government.|