Frankfurt, Failure, and Finding Fault
V. Alan White
Harry Frankfurt has long argued that examples of overdetermined moral agents prove that reasonable claims of moral responsibility against them do not entail that the agents involved could have acted otherwise (stated as a necessary condition of responsibility, Frankfurt calls this the Principle of Alternative Possibilities, or PAP; [1]). However, recent clarifications of certain of his examples reveal the subtle presence of ceteris paribus assumptions at work in them that, when examined more carefully, either call his entire project into question or at least require a narrower claim for what the examples establish.
In an attempted response to some criticisms by Peter van Inwagen ([4]) to the effect that Frankfurt's arguments do not address questions of the responsibility of failures to act, Frankfurt ([3]) offers the example of an automobile driver Q who fails to drive attentively due to his preference to look left at scenery during a crucial moment. Frankfurt adds overdetermining conditions that counterfactually necessitate Q's looking left at that time. He then remarks: In these circumstances, Q cannot keep his eyes straight ahead. Is he morally responsible for failing to do so? Of course he is! The fact that he cannot avoid failing has no bearing on his moral responsibility for the failure, since it plays no role in leading him to fail. ([3], 292, latter emphasis mine.) Frankfurt believes that this latter claim is justified because while the overdetermining conditions in Q's case stand as redundant sufficient conditions for Q's failure, they are not at all necessary for Q's failing in the actual sequence of events, as opposed to necessary conditions external to an agent that were absent in a consequences-oriented example van Inwagen offered, and thus accounted (in part) for van Inwagen's agent's moral failure. Frankfurt concludes that judgments of moral failures such as that of Q are therefore completely justified without resort to either a PAP-like principle or reference to any existing (but actually inoperative) overdetermining conditions:
Failing to keep one's eyes straight ahead is exclusively a matter of what movements a person makes; it is constituted by what the person himself does, and what the person does is therefore both a necessary and sufficient condition for it. It cannot be said, then, that Q's failure would have occurred no matter what he had done -- i.e., regardless of what bodily movements he made. If he had not moved his eyes to the left at all he would have not failed. ([3], 292-293)
This passage bundles together not only much of the force of Frankfurt's counterexamples against PAP, but the basis of his psychologically-structured compatibilism as well [2]. For here he states quite powerfully what he takes to be the moral sufficiency of agents who act even in overdetermined conditions: [f]ailing to keep one's eyes straight ahead is exclusively a matter of what movements a person makes in such circumstances that do not bring peripheral (i.e., non-agent-related) but actually present necessary conditions of moral action into play. It is the moral purity of the example of Q apart from surrounding circumstances that so effectively fixes our gaze upon Q as the only entity supposedly responsible for the failure.
However, as is the case with all Frankfurt-style examples, the intuitive judgment of Q's responsibility is mainly driven by the apparent irrelevance of all surrounding circumstances, even ones of overdetermination, no matter what their counterfactual significance. The one subtlest factor in all this is that Q's act in the given example is stipulated to be a failure. This begs some critical attention be paid to the fundamental issue of what a failure is, as well as how Q in Frankfurt's example is specifically judged to fail.
Again, intuitively, it would appear that any agent's failure arises because of an absence of some normatively expected act or consequences of an act. Since Frankfurt's example requires that Q's act be a failure in some sense, it should be made clear in what sense that act constitutes an absence of some normatively expected act. In Q's case clearly this is that Q should have kept his eyes fixed on the road ahead during the time period he was actually judged to have failed. Note, however, that the normative expectation here is two-fold, both generally and specifically. Generally we expect that drivers attend to driving, ceteris paribus. Specifically a driver fails to be attentive if this expectation is unmet without qualification to the ceteris paribus specification -- i.e., if there are no circumstances mitigating our normative judgment of failure. If we do discover such mitigating circumstances, then we may find a particular driver absolved of failure, such as when a driver is maliciously drugged or suddenly and unexpectedly attacked by a passenger. The driver may not have been properly attentive to driving in such a case, but we do not attribute a failure to her.(71)
Of course, in the case of Q Frankfurt argues that there are no such mitigating circumstances, and thus we may hold Q responsible for failure. In so arguing Frankfurt draws a distinction between personal and impersonal unavoidable behaviors:
Now there are two ways in which a person's action, or his failure to act, or a consequence of what he has done, may be unavoidable. It may be unavoidable in virtue of making certain movements which the person makes and which he cannot avoid making; or it may be unavoidable because of events or states of affairs that are bound to occur or to obtain no matter what the person himself does. . .I shall refer to the first type of unavoidability as personal and to the second as impersonal. ([3], 293)
Frankfurt argues that Q's unavoidable failure is personal, and thus he is fully responsible for his failure ([3], 292). Why? Though Q's act is overdetermined by external otiose circumstances, he fails due to his own behavior -- not only because of some external condition or situation that requires failure come what may (as in van Inwagen's own imagined impersonal case, involving an apathetic agent unaware that a telephone he should have used was actually broken). Recall that it is Frankfurt's belief that [i]t cannot be said, then, that Q's failure would have occurred no matter what he had done -- i.e., regardless of what bodily movements he made. If he had not moved his eyes to the left at all he would have not failed (emphasis mine). Hence Frankfurt argues that Q, and only Q, is responsible for his failure.
However, it is instructive to note that in this latter supportive remark that Frankfurt appeals to something like a ceteris paribus case of (some) Q's failure! In Lewisian modal language, the possible Q referred to in this latter statement (the he of the counterfactual antecedent) is a counterpart quite remote from the Q of Frankfurt's example -- a counterpart Q who presumably is not overdetermined to fail (otherwise the consequent of the counterfactual would be false). It is this modally remote Q who may truly be said to have failed ceteris paribus, and that counterfactual reference to failure reveals that at least some of these ceteris paribus assumptions involve conditions surrounding that Q -- namely that he is not overdetermined to fail, and so does not if he does not move his eyes to the left at all. Besides the dubious -- I would say equivocal -- slide from arguing about Frankfurt's Q to appealing to a modally remote counterpart Q, this raises the question of how such ceteris paribus assumptions work -- or are ignored -- in the consideration of Frankfurt's overdetermined Q.
In general, what could be the nature of these assumptions? Appealing to familiar kinds of cases somewhat like Q's, as suggested above, they are of two varieties: one, that the agent involved is of rather ordinary character and behavioral capacity; two, that the agent is not coerced to act or otherwise interfered with in acting. In the case of Frankfurt's Q, both of these, Frankfurt would argue, are intact, and most importantly for Frankfurt's example, in spite of the presence of overdetermining conditions. I would urge, however, that this latter claim overlooks some commonly-held views on what constitutes freedom from interference.
Interferences in another's affairs are of two kinds. One is direct and causal, as in cases of forceful physical or psychological coercion. Obviously Frankfurt's overdetermining conditions for Q are irrelevant here, and assist the plausibility of his example. But another kind of interference is indirect and (at least potentially) more passive, as in cases of clandestine conspiracy. These cases constitute interference not because they are necessarily directly invasive, but because they transgress a basic concept of fairness -- agents should be left completely alone to do as they, and they alone, see fit. Of course, Frankfurt could rightly point out that the overdetermining conditions for Q were, in fact, unneeded -- Q did act as he saw fit. How then could these conditions constitute interference?
I insist to the contrary that our basic moral concept of fairness is not as restricted as that. Consider the case of a gambler who unwittingly agrees to a certain series of bets against a roulette wheel fixed by the house, which would be used near the end of the series of bets to assure that the gambler loses. As it turns out, however, the gambler's luck just happens to be so bad that the means of assured loss are never invoked. If we discover this arrangement afterwards, do we excuse the house from blame completely? I would think most certainly not -- the house conspired against a player, and thus it was not possible for the gambler to win. Our sense of fairness is offended, and we may well argue that the gambler's losses should be returned. Note, moreover, that this sense of fairness is built upon something like the very PAP Frankfurt disdains, though it is not a PAP directly related to questions of personal character. For the gambler, given ordinary conditions, very likely would not have acted otherwise given his proclivity to waste money. Rather, the situation wasn't fair because the wider conditions didn't provide any possibility for the gambler to win, irrespective of his character.
Frankfurt might protest that this example is similar to van Inwagen's in crucial respects, and for that reason is similarly irrelevant as a criticism of his Q example. He could try to argue that the house's conspiratorial action against the gambler constituted matters over which the gambler had no control, and thus necessitated the gambler's losses. Hence, the gambler's failure to win was impersonal, as was van Inwagen's apathetic agent (and for that reason the gambler is not fully responsible for losing his money).(72) However, I would counter that such an argument depends on considering the relevance of overdetermining conditions that Frankfurt himself usually questions by focussing on the actual sequence of events -- namely, that in fact the gambler lost of his own foul luck and the overdetermining conditions were not needed, and thus played no role in the gambler's actually losing. Hence, in the actual sequence of events the gambler's failure to win was personal -- he wanted to gamble and it turned out that he lost his money on fair spins of the roulette wheel. My point specifically is that the fixed roulette wheel was not used because the conspiratorial house got lucky and needed to do nothing, and yet our moral intuitions of fairness cannot exempt the house from responsibility based on the simple fact that the gambler couldn't win in any case. Note that I do not have to claim that the gambler is not at least partially responsible here -- his wantonness about money need not be ignored completely. But the gambler cannot be held fully accountable for losing, which is all I must demonstrate. Our intuitions about Q, I insist, must be parallel. And generally I would say that this situation about the gambler draws out the key defect in all Frankfurt-style scenarios: there are always, according to Frankfurt, unindictable individuals or circumstances that in fact got lucky and needed to do nothing to bring about a certain end result. But, I insist that it is their very indictment that our sense of fairness requires, and that in turn dilutes the attribution of responsibility we apportion to the unlucky evil-doer (as being a sort of unwitting free stooge).
Hence, I would argue that Frankfurt's Q should not be held responsible for failure, or at least not fully responsible for it, as long as we consider that something or someone conspired against him to fail. Generalizing from this point, I would also argue that Frankfurt examples as a whole ignore the role PAP plays in our ordinary ideas of fairness: we require in general that our morally responsible actions are not merely our own, but fairly so, apart from a conspiratorial set of even actually otiose circumstances that would otherwise guarantee a particular kind of outcome. So PAP remains a necessary condition for full moral responsibility in that wider sense, even if Frankfurt's examples do serve to show -- as with a more ordinary ceteris paribus case of our gambler -- that PAP need not apply to agent's characters in order to hold them accountable.(73)
REFERENCES
[1] Frankfurt, Harry G., Alternate Possibilities and Moral Responsibility, Journal of Philosophy, 45 (1969), 829-39.
[2] Frankfurt, Harry G., Freedom of the Will and the Concept of a Person, Journal of Philosophy, 68 (1971), 5-20.
[3] Frankfurt, Harry G., What We Are Morally Responsible For, in John Martin Fischer and Mark Ravizza, eds., Perspectives on Moral Responsibility (Ithaca: Cornell University Press, 1993), 292-93.
[4] Van Inwagen, Peter, Ability and Responsibility, Philosophical Review, 87 (1978), 201-24.
V. Alan White
Professor of Philosophy
University of Wisconsin -- Manitowoc
awhite@uwc.edu