In order to understand why we accept the principle just mentioned yet continue to feel the force of the intuition that is sensitive to harmful consequences, we need to turn to the evolutionary function of blame judgments.
Though the precise nature and mechanisms of blame remain a matter of dispute, it is widely agreed that the function of blame and punishment is to promote cooperation (Boyd et al. 2003; Frey and Rusch 2012; Cushman 2013). Under a wide variety of circumstances, organisms do better – increase their inclusive fitness – by cooperating with conspecifics (and sometimes with members of other species) than by pursuing more solitary strategies. Cooperative behavior increases inclusive fitness either because the conspecific shares a sufficient proportion of the organism’s genes, such that aiding it increases the likelihood that the organism will have copies of its genes represented in the next generation, or because aid is reciprocated directly or indirectly, or because cooperation enables the exploitation of resources that cannot be exploited by individuals alone.
Punishment enters the picture for familiar game theoretical reasons: populations of altruistic organisms are vulnerable to invasion by organisms playing more selfish strategies. Unconditional cooperators will be taken advantage of by selfish organisms, who will be fitter than them: cooperation will therefore die out and selfish strategies go to fixation. Conditional cooperation – cooperation only with cooperators – is required to avoid this kind of invasion. Blame is a response that protects against free riding and the collapse of cooperation. It may achieve its adaptive ends either by changing the behavior of noncooperators or by excluding them from participation in exchanges (Nakao and Machery 2012).
In human groups in the environment of evolutionary adaptiveness, an individual represents a surprisingly significant investment of time and energy: “the resource debt that individuals acquire as children and adolescents is not paid off until around 50″ (Sterelny 2012: 30). Since individuals represent such a significant investment, the loss of their potential contribution to the group is extremely costly.Footnote 5 Further, the individual is likely to be closely related to other group members, so harming them directly affects other members’ inclusive fitness. These costs can be reduced if blame tracks properties of agents that predict that they are bad investments: that they are likely to continue to defect from cooperative arrangements. These properties are agents’ mental states and capacities. Since deliberate defection is a much better predictor of future defection than is accidental, blame should track agents’ desires and their beliefs about the consequences of their actions. Hence, I suggest, our sensitivity to agents’ quality of will.
We now have an explanation of why blame responses track agents’ mental states. We still need to explain why the harms they cause make an independent contribution to such responses. I suggest that our disposition to blame for harms independently of mental states is a product of the constraints under which blame responses evolved.
Due to the costs or difficulty of gathering accurate information, evolved responses may sometimes utilize proxies for the state it is adaptive to track. Since the mental states of agents are at least partially internal states, mechanisms that generate blame judgments will be required to track proxies to track mental states. They are likely to utilize the outputs of theory of mind modules, which respond to cues ranging from facial expression to gaze direction to explicit utterances. In addition, though, they are likely to be sensitive to consequences because agents often have incentives to conceal their motives from one another. Because our blame responses track mental states, agents have an incentive to attempt to pass off deliberate defection as accidental. It is for this reason, I suggest, that harm contributes independently to blame responses. Because there is a sufficiently high correlation between causing and intending harm in a range of circumstances, we are disposed to take the causation of harm as a sufficient condition for some blame in those circumstances. One effect of this disposition to take harm as a proxy for mental states is that under some conditions we are disposed to blame independently of, and in excess of any contribution to blame judgments of, actors’ actual mental states.
We are now in a position to solve the puzzle of resultant moral luck. The proposal gives us the resources to explain not only (1) why we have a disposition to blame that tracks actors’ mental states and (2) why we nevertheless attribute a degree of blame greater than would be predicted by (1) alone, in cases where the agent causes a harm more severe than they wanted, intended or foresaw; we can also explain why the attribution of blame is not stable: that is, why we experience conflict in our intuitions. Above I promised I would solve the puzzle by showing that agents are implicitly committed to the principle that blame should be commensurate with agents’ mental states. I now have the resources to make good on this promise.
On the account of blame attribution I just sketched, blame responses that are caused by perceiving that an actor has caused a harm are not alternatives to blame responses that track mental states; the former responses are themselves mediated by the attribution of mental states. Recent work provides independent evidence for this claim: Lench et al. (forthcoming) showed that the direct effect of the moral luck beliefs of their subjects on perceptions of blameworthiness was reduced when the intention of agent blamed was included as a mediator in a mediation analysis, suggesting that people who believe that moral luck can make an agent more blameworthy hold that belief because they implicitly think that a worse harm reflects the agent’s intentions.Footnote 6 Because of the difficulties of tracking mental states directly, we take harms as proxies for mental states. We do not have two rival implicit theories of conditions sufficient for blame, one mental state-based and one harm-based; we have one implicit theory of sufficient conditions for blame, which is mental state-based.
We can now solve the puzzle of resultant moral luck. In a resultant moral luck case, we are torn because we are implicitly disposed to accept the following claims:
-
1.
Agents deserve blame that is commensurate with the quality of their will;
-
2.
Actions that cause a harm stem from a quality of will that matches the severity of the harm caused;
-
3.
The agent who caused the more severe harm has no worse a quality of will than the agent who caused the less severe.
We are implicitly disposed to accept claim 1 because we are disposed to attribute blame based on the quality of agents’ wills (recall that our disposition to blame based on harms normally supports, rather than conflicts with, this disposition). The evidence that people attribute more blame to those who cause more severe harms, independent of what they know about the actor’s mental states, together with my hypothesis that this blame is mediated by an implicit – and nonconscious – attribution of a commensurate mental state to the agent, entails claim 2. In the philosophical context, we are disposed to accept claim 3 because it is stipulated (outside that context, our disposition to attribute mental states to agents on the basis of the harms they cause restricts the frequency with which we encounter the puzzle, but of course there is sometimes good evidence that someone has caused a harm in excess of anything they desired, foresaw or intended).
We experience conflict in cases in which claims 2 and 3 commit us to attributing different qualities of will to an agent. When three is combined with one, we are disposed to think that the agent deserves no more blame than another who did not cause as serious a harm, but when two is combined with one we are disposed to think that that agent deserves more blame than his counterpart. Since we are implicitly committed to all three premises, we are implicitly committed to both judgments. Hence the conflict.Footnote 7
The resulting account of the intuitions generated by resultant moral luck cases is plainly a variety of epistemic reductionism: it explains the intuitions we experience by reference to facts about us and the manner in which we attribute blame to others, rather than facts about the agents who feature in these cases. Unlike other epistemic accounts of moral luck, however, this account predicts the intuitions that have sometimes been wielded against epistemic reductionism. As Adler (1987) and Schinkel (2009) note, contemplating the fact that had the agent been luckier they would not have caused a particular harm does not in fact cause the intuition that they deserve more blame to dissipate; this fact, they believe, counts against the claim that our judgments are a function of our epistemic situation. The account offered actually predicts that our intuitions will be recalcitrant in just the way Adler and Schinkel each highlight. The kinds of psychological processes that underlie our judgments are, like many others we have as a consequence of our evolutionary history, likely to be encapsulated in the sense Fodor (1983) made famous: that is, insensitive to personal-level information. Hence our knowledge that the judgments are generated by two different and conflicting processes will not affect the intuitions generated by each, which will continue to be phenomenologically insistent.Footnote 8
Attention to the same facts can also help to motivate a response to Domsky’s (2005) objections to epistemic reductionism. Domsky argues that the problem of moral luck can arise as insistently in the self-regarding case as in the other-regarding case; since we have direct access to the content of our mental states, the harm cannot serve as evidence for these mental states. Indeed, no harm needs actually to occur: as Domsky also points out, the problem can arise prior to discovering whether our actions have any consequences at all. We need only to consider how things might unfold to have the intuition that if things go badly we are much more blameworthy than otherwise. Again, though, the hypothesis concerns an encapsulated module which generates a nonconceptual output. There is no more reason to think that it is sensitive to information concerning our own mental states than it is to information concerning the mental states of other agents: it produces its characteristic output, given an appropriate stimulus, independent of what other beliefs the person maintains.