Pitcovski’s explanation-based account of harm

Abstract

In a recent article in this journal, Eli Pitcovski puts forward a novel, explanation-based account of harm. We seek to show that Pitcovski’s account, and his arguments in favor of it, can be substantially improved. However, we also argue that, even thus improved, the account faces a dilemma. The dilemma concerns the question of what it takes for an event, E, to explain why a state, P, does not obtain. Does this require that P would have obtained if E had not occurred? Pitcovski’s theory faces problems no matter how one answers that question.

1 Introduction

The two main accounts of harm are the counterfactual comparative account and the causation-based account. Eli Pitcovski has recently argued that both of these accounts yield unsatisfactory results concerning what factors help to determine the degree of harm (Pitcovski, 2023). Whereas the counterfactual comparative account is too permissive in this regard, Pitcovski claims, the causation-based account is too restrictive. To overcome these problems, he puts forward a novel, explanation-based account.

Pitcovski formulates the counterfactual comparative account as follows:

COMPARE: An event E is harmful for some subject S iff the actual world (in which E occurs) is overall worse for S than the nearest possible world in which E does not occur. The degree to which an event E is harmful for S is the degree to which the actual world is worse for S than the nearest possible world in which E does not occur. (Pitcovski, 2023: 510)

Reflection on cases involving counterfactual backtracking, Pitcovski argues, shows this account to be inadequate. Here is his main case:

JOSEPH: Joseph’s town is captured by a cruel and powerful army. The people of the town are given the following choice: either they execute one person and leave the body outside the walls by midnight, or the whole town will be subjected to a lifetime of torture. Since the commander who gave them the choice is reliable and has a reputation for carrying out his threats, the people of the town decide on a fair lottery that will determine which of them will be executed. By bad luck, Joseph is chosen. He is injected with a dose of poison strong enough to kill an elephant and then put in a cage outside the walls. The poison causes him pain that is worse than any pain he has ever experienced and by midnight, it causes his death. The rest of his town is not tortured. (Pitcovski, 2023: 511)

What would have happened if Joseph’s death had not occurred? According to Pitcovski, the most natural answer is that the past, relative to the time of Joseph’s death, would have been substantially different. In particular, Pitcovski contends, Joseph would not have been poisoned, as someone else would have been chosen by the lottery. Hence, if COMPARE is true, the intrinsic badness of the pain caused by the poison contributes to the degree to which Joseph’s death is harmful. However, Pitcovski claims, this is an implausible result; “[t]his pain has nothing to do with the (presumably essentially deprivational) harm of Joseph’s death” (2023: 512).

Pitcovski formulates a simple version of the causation-based account as follows:

CAUSE: An event E is harmful for some subject S iff the total consequences of E are more intrinsically bad than intrinsically good for S. The degree to which an event E is harmful for some subject S is the degree to which the total consequences of E are more intrinsically bad than intrinsically good for S. (Pitcovski, 2023: 510)

CAUSE easily deals with the JOSEPH problem, as Joseph’s death does not cause the preceding pain. According to Pitcovski, however, CAUSE has another problem: it fails to accommodate the contribution to an event’s harmfulness made by the intrinsic value of what the event constitutes rather than causes. Pitcovski quotes, approvingly, the following passage from Duncan Purves (who assumes, in this context, that desire frustration is intrinsically bad for us):

Suppose that my mother desires that I not join a cult. If I join a cult, this frustrates her desire … But, it would misconstrue things to say that my joining a cult causes her desire to be frustrated … My joining a cult harms my mother by constituting her desire frustration. (Purves, 2019: 2638)

Pitcovski claims that we can strike a balance between COMPARE’s permissiveness and CAUSE’s restrictiveness by focusing on what the relevant event causally or non-causally explains. He proposes the following account:

EXPLAIN: An event E is a harm for S iff the totality of states that obtain (/fail to obtain) because E occurs are overall intrinsically bad (/good) for S. The degree to which an event E is harmful for S is the degree to which the states that obtain (/fail to obtain) because E occurs are overall intrinsically bad (/good) for S. (Pitcovski, 2023: 517–18)

This account, Pitcovski suggests, nicely avoids the problems for COMPARE and CAUSE. In JOSEPH, to begin with, it is not because Joseph’s death occurs that he experiences pain. Hence, unlike COMPARE, EXPLAIN rules out that the intrinsic badness of Joseph’s pain contributes to the harmfulness of his death. (By contrast, Pitcovski suggests, the intrinsic goodness of various states of affairs such as, for instance, Joseph’s being happy in the future does so contribute, as their non-obtaining is explained by his death.) Moreover, since EXPLAIN, unlike CAUSE, is not restricted to causal explanation, it also handles Purves’s case: “Even if my joining the cult constitutes my mother’s desire frustration, rather than causing it, my joining the cult explains my mother’s desire frustration” (Pitcovski, 2023: 518).

While Pitcovski does not mention this, it seems clear that this kind of problem arises not just for CAUSE but also for other theories of harm that focus on causation (such as those discussed in Carlson et al., 2022; see also Johansson & Risberg, 2022: 518–20). Thus, as Pitcovski’s positive proposal illustrates, such views could also be improved by being recast in terms of explanation. Accordingly, although we will focus on Pitcovski’s theory, our discussion will also be relevant to more general questions concerning the prospects and problems of such explanation-based views of harm.

In this paper, we have two main aims. The first is to show that Pitcovski’s theory, and his arguments in favor of it, can be substantially improved (Section 2). The second is to show that, even thus improved, his theory faces a dilemma. The dilemma concerns the question of what it takes for an event, E, to explain why a state, P, does not obtain. Does this require that P would have obtained if E had not occurred? Pitcovski’s theory faces problems, we shall argue, no matter how one answers that question (Sects. 3–4).

2 Pitcovski’s position improved

While we agree with Pitcovski that COMPARE struggles in cases involving backtracking, we are not convinced that these problems are best illustrated by the JOSEPH case. In particular, it seems far from clear that Pitcovski’s claim about what would have happened if Joseph’s death had not occurred—i.e., that Joseph would not have experienced the pain caused by the poison—is the most natural one to make. It seems to us at least as natural to claim instead, for instance, that Joseph would still have experienced the pain caused by the poison but managed to stay alive, albeit barely, for a while longer.¹ Of course, this latter suggestion and COMPARE jointly imply that Joseph’s death does not harm him, as it does not deprive him of any valuable future. But given the extraordinarily unfortunate circumstances in which Joseph finds himself, isn’t this implication perfectly acceptable? Pitcovski (2023: 513) suggests that we need to invoke the harmfulness of Joseph’s death in order to explain why it was such a misfortune for Joseph to be chosen by the lottery. Obviously, however, proponents of COMPARE have another explanation available—namely, that Joseph would have been overall better off if he had not been chosen by the lottery.

There are other illustrations available of the kind of problem for COMPARE that Pitcovski seeks to highlight. Here is one (taken primarily from Johansson and Risberg 2023: 515; cf. Carlson et al., 2022: 422):

OMNISCIENCE: O is an essentially omniscient being who essentially forms her beliefs about an event’s occurrence immediately after it happens. Robin feels intense pain at t1. Hence, O immediately thereafter, at t2, forms the belief that Robin feels intense pain at t1.

It seems that Robin would not have experienced the pain—and thus would have been better off—had O’s forming the relevant belief not occurred. COMPARE thus entails, counterintuitively, that O’s forming the belief harms Robin.

This objection to COMPARE seems to be more forceful than Pitcovski’s for several reasons. To begin with, OMNISCIENCE is in several respects less complicated than JOSEPH. Moreover, the relevant counterfactual about OMNISCIENCE—i.e., that if O’s forming the relevant belief had not occurred, then Robin would not have experienced the pain—is less problematic than the corresponding one about JOSEPH. In particular, given O’s essential properties, it is not plausible to suggest instead that if O’s forming the relevant belief at t2 had not occurred, Robin would still have experienced the pain at t1 but O would have formed the relevant belief only later than t2. (Indeed, the consequent of that counterfactual is metaphysically impossible.) Furthermore, while Pitcovski takes COMPARE to have implausible results only with regard to the degree to which Joseph’s death is harmful, it has implausible results not only with regard to the degree to which O’s forming the relevant belief is harmful, but also with regard to whether or not that event is harmful. Note also that EXPLAIN seems to handle OMNISCIENCE better than COMPARE does, since neither the obtaining nor the non-obtaining of any relevant states of affairs (such as Robin’s feeling intense pain) is plausibly explained by O’s forming the relevant belief. Thus, we suggest that the case for Pitcovski’s view can be strengthened by appealing to cases like OMNISCIENCE rather than JOSEPH.

This in turn highlights the issue of how an explanation-based account of harm of the kind Pitcovski endorses is best formulated. Recall his own formulation:

EXPLAIN: An event E is a harm for S iff the totality of states that obtain (/fail to obtain) because E occurs are overall intrinsically bad (/good) for S. The degree to which an event E is harmful for S is the degree to which the states that obtain (/fail to obtain) because E occurs are overall intrinsically bad (/good) for S.

Thus formulated, Pitcovski’s account implies that a state of affairs P that fails to obtain because E occurs is relevant to E’s harmfulness for S only if P is intrinsically good for S. This yields implausible results in many cases. Suppose for instance that hedonism is true, and that Shane feels pain to degree 10 at t1. Suppose that the pain is bound to continue undiminished until and including t2 unless Shane takes a pill at t1—an action that would result in Shane’s feeling pain to degree 5 at t2. Suppose also that Shane does take the pill at t1. Since the state of affairs of Shane’s feeling pain to degree 5 at t2 is intrinsically bad for him, and obtains because Shane takes the pill at t1, its intrinsic badness contributes to the harmfulness of his action on EXPLAIN. By contrast, the intrinsic badness of Shane’s feeling pain to degree 10 at t2, which fails to obtain because he takes the pill, does not. Since this state of affairs is not intrinsically good for Shane, EXPLAIN considers it irrelevant to the harmfulness of his action. Assuming that Shane’s action does not affect his hedonic level at any time other than t2, EXPLAIN (unlike COMPARE) thus implies that it harms him. This is clearly the wrong result.

We think that an improvement of EXPLAIN is available. Let T be the set of states of affairs that obtain because E occurs and let T* be the set of states of affairs that fail to obtain because E occurs. A view in the neighborhood of EXPLAIN can then be formulated as follows:

EXPLAIN*: An event E is a harm for S iff (and to the degree that) the total intrinsic value for S of the states of affairs in T* is greater than the total intrinsic value for S of the states of affairs in T.²

EXPLAIN* avoids the problem just described. In particular, Shane’s feeling pain to degree 5 at t2 is intrinsically better for him than Shane’s feeling pain to degree 10 at t2. Thus, we will primarily focus on EXPLAIN* in what follows, though the dilemma we shall present also arises for EXPLAIN.

One set of problems for EXPLAIN that the move to EXPLAIN* does not help address has to do with the fact that both the states in T and the states in T* may stand in various logical relations to one another. For example, if Maria is in an accident which explains the obtaining both of Maria’s feeling pain in her hand and of Maria’s feeling pain in her leg, it seems plausible that the accident also explains the obtaining of the conjunctive state Maria’s feeling pain in her hand and in her leg. However, it is implausible to sum the intrinsic disvalues of all three states when determining the accident’s degree of harmfulness. For another type of example (taken with minor modifications from Pitcovski, 2023: 524), suppose that Maria dies just before entering an amusement park. This event may plausibly be taken to explain the non-obtaining both of Maria’s enjoying the hot air balloon at 2 p.m. and of Maria’s enjoying the rollercoaster at 2 p.m. However, it is implausible to sum the intrinsic values of both these states when determining the degree of harmfulness of Maria’s death, given that there is no possible world in which they both obtain—and even more implausible to also add in the intrinsic value of their conjunction.³ While these cases give rise to a host of further thorny issues, the dilemma we shall present is largely independent of them, so in what follows we shall set them aside.

3 A dilemma for Pitcovski’s view: first horn

Both EXPLAIN and EXPLAIN* raise several questions about how the relevant notion of explanation is best understood. Pitcovski’s account is meant to “remain neutral with respect to accounts of explanation” (2023: 520), and we shall follow him in not presupposing any particular account.⁴ However, there is one particular issue about explanations that we wish to highlight—an issue that is the focus of our dilemma for EXPLAIN*.

This issue concerns the relation between explanations of non-obtainings and certain counterfactuals. Consider these claims: (i) event E explains why state of affairs P does not obtain; (ii) P would have obtained if E had not occurred. It seems out of the question to regard (ii) as sufficient for (i). In OMNISCIENCE, for example, it seems clear that although the state of affairs of Robin’s not experiencing intense pain would have obtained if the event of O’s forming the relevant belief had not occurred, that event does not explain why Robin’s not experiencing intense pain does not obtain. Moreover, as Pitcovski (2023: 524) points out, he in particular must deny that (ii) is sufficient for (i), as his approach would otherwise lose its advantages over COMPARE in backtracking cases.

Even so, a possible view is that (ii) is necessary for (i). In other words:

EXPLANATIONS OF NON-OBTAININGS ENTAIL COUNTERFACTUALS (ENEC): If an event E explains why a state P does not obtain, then: if E had not occurred, P would have obtained.

This principle is central in our dilemma for EXPLAIN*. We shall argue that, no matter whether EXPLAIN* proponents accept or reject ENEC, the resulting package of views will be highly unattractive. This section considers problems that arise if ENEC is accepted; the next section considers problems that arise if ENEC is rejected.

To begin with, note that if ENEC is true, EXPLAIN* will inherit virtually all of the well-known problems that COMPARE faces (other than those related to backtracking). For example, consider the following case, in which one intuitively harmful action prevents an intuitively even more harmful action (for similar problem cases for COMPARE, see, e.g., Boonin, 2014: 62–63; Bradley, 2012: 407; Feit, 2015, 2023: ch. 3; Hanna, 2016; Immerman, 2022; Johansson & Risberg, 2019; Norcross, 2005: 165–66):

THREE DRUGS: Serena has three drugs at her disposal, each of which she can give to Rupert. Drug 1 would bring Rupert 200 units of pleasure, drug 2 would bring him 100 units of pain, and drug 3 would bring him 200 units of pain. Serena gives Rupert drug 2. If she had not done so, she would have given him drug 3.

Since Rupert is even worse off in the nearest world in which Serena does not give him drug 2, COMPARE counterintuitively implies that Serena’s action does not harm him.⁵ Assuming ENEC, Serena’s action does not explain why Rupert does not receive 200 units of pleasure. Hence, assuming ENEC, EXPLAIN*, just like COMPARE, considers the fact that drug 1 would bring Rupert 200 units of pleasure irrelevant to the harmfulness of Serena’s action. If EXPLAIN* is to avoid the result that Serena’s action does not harm Rupert, it must be assumed that Serena’s giving Rupert drug 2 also fails to explain why he does not feel 200 units of pain. Pitcovski might suggest that this assumption is warranted, since the description of the case does not mention any relevant causal (or constitutive) relation between the putative explanans and explanandum. But we can simply add such a causal relation to the case. Suppose, for example, that drug 2 reliably has two effects. It gives you a severe headache, involving 100 units of pain, but it also anaesthetizes the rest of your body, ensuring that you cannot feel more than 100 units of pain. Under these assumptions, it seems hard to deny that Serena’s giving Rupert drug 2 explains why he does not feel 200 units of pain.⁶ And surely, the presence of this causal relation does not make the judgment that Serena’s action does not harm Rupert any less counterintuitive.⁷

Pitcovski briefly discusses the possibility of combining his view of harm with an account of explanation which, like ENEC, takes counterfactual dependence to be necessary for explanation (2023: 525).⁸ He also notes, as we have done, that this move would render his view susceptible to some of the problems that COMPARE faces, and acknowledges that this is “definitely a price to pay” (2023: 525). Of course, it is controversial how strong these objections are; in particular, various COMPARE proponents take themselves to have good responses to them (see, e.g., Feit, 2015, 2023: chs. 4–5; Jedenheim Edling, 2022). However, a problem that arises on this horn of the dilemma which is independent of those controversies is that ENEC itself is also questionable. In particular, as we shall now argue, cases which involve non-causal metaphysical explanations (of the kind that figures in Pitcovski’s objection to CAUSE; cf. Section 1) pose significant problems for ENEC.

The standard view about grounding explanations is that disjunctive facts are fully grounded in, and thus explained by, their obtaining disjuncts (see, e.g., Fine, 2012). On this view, for instance, the fact [Emmanuel Macron wins the 2022 French presidential election] fully grounds (and thus explains) the fact [Either Emmanuel Macron wins the 2022 French presidential election or Emmanuel Macron is a European]. That is so even though Macron is a European, and even though he would still have been a European if he had not won the 2022 election. Given all these claims, which we take to be orthodox, it would surely be implausible to deny that the event of Emmanuel Macron’s winning the 2022 French presidential election explains why the state Emmanuel Macron’s either winning the 2022 French presidential election or being a European obtains. Similarly, the fact [the ball is red] plausibly fully grounds the fact [the ball is colored] even if the ball would have had a different color if it had not been red. If that is granted, it seems clear that the state the ball’s being red explains why the ball’s being colored obtains. And if explanations of why states obtain do not presuppose the relevant type of counterfactual dependence between explanandum and explanans, as these examples illustrate, it is hard to see why we should think that explanations of why states do not obtain do.

This objection can be sharpened. The following principle seems highly attractive:

EXPLANATIONS OF OBTAININGS ENTAIL EXPLANATIONS OF NON-OBTAININGS (EOEEN): If event E explains why state P obtains, then E explains why the negation of P (i.e., the state not-P) does not obtain.

Suppose that E explains why P obtains, and that P would have obtained even if E had not occurred. By EOEEN, E explains why not-P does not obtain even though not-P would not have obtained if E had not occurred. This result directly contradicts ENEC.

4 A dilemma for Pitcovski’s view: second horn

Turn now to the second horn of the dilemma. That is, suppose that proponents of EXPLAIN* reject ENEC, and thus claim that E can explain P’s non-obtaining even if the latter is not counterfactually dependent on the former. They will thereby avoid the problems discussed in the previous section. To begin with, EXPLAIN* will no longer inherit the problems for COMPARE that THREE DRUGS illustrates. Since proponents of EXPLAIN* no longer have to deny that Serena’s action explains why Rupert does not receive 200 units of pleasure, they no longer have to deny that it harms him. Obviously, moreover, the problem that ENEC is itself questionable is no embarrassment for proponents of EXPLAIN* who reject ENEC.

However, the denial of ENEC renders EXPLAIN* vulnerable to another serious problem, which does not arise for COMPARE. Consider the following variation on THREE DRUGS:

TWO DRUGS: Serena does not have access to drug 1. Hence, her options are to give Rupert either drug 2, drug 3, or no drug. Drug 2 would give Rupert a severe headache, involving 100 units of pain, and anaesthetize the rest of his body, ensuring that he would not feel more than 100 units of pain. Drug 3 would bring Rupert 200 units of pain. If Serena were to give Rupert no drug, he would not feel any pain. Serena gives Rupert drug 2. Had she not done so, she would have given him no drug.

The obvious verdict in TWO DRUGS is that Serena’s action harms Rupert. This is also what COMPARE implies. But if Serena’s action explains not only why Rupert is not free from pain, but also why he does not feel 200 units of pain, EXPLAIN* implies that it does not harm him. In order to avoid this very implausible result, a defender of EXPLAIN* must claim that Serena’s giving Rupert drug 2 cannot explain why he does not feel 200 units of pain. But if ENEC is false, it is hard to see how this claim could be established. Since the causal effects of drug 2 ensure that Rupert will not feel more than 100 units of pain, it is very plausible to conclude that Serena’s action explains why he does not feel 200 units of pain.⁹

It might be thought that the problem posed by TWO DRUGS can be solved by modifying how an event’s harmfulness is taken to depend on the intrinsic values of the states of affairs in the sets T and T*. Instead of subtracting the sum of the intrinsic values of the states in T* from that of the states in T, as EXPLAIN* dictates, we might consider some other function of the intrinsic values of the relevant states. However, we doubt whether any such function would yield substantial improvement on EXPLAIN*, in part because an argument that we have presented elsewhere (Carlson et al., 2021: 171–73) against corresponding modifications of COMPARE seems to us to be applicable here as well. While this is not the place for any detailed discussion of this issue, it is worth considering the arguably most natural proposal along the present lines—namely, that E is a harm for S if and only if (and to the degree that) the average intrinsic value for S of the states of affairs in T* is greater than the average intrinsic value for S of the states of affairs in T. It is easy to see that this variant of EXPLAIN* is no improvement. In TWO DRUGS, we may suppose that Rupert’s experiencing 100 units of pain is the only state that both has intrinsic value for Rupert and obtains because Serena gives him drug 2. Similarly, we may suppose that Rupert’s experiencing 0 units of pain and Rupert’s experiencing 200 units of pain are the only states that both have intrinsic value for Rupert (presumably neutral in the former case) and fail to obtain because of Serena’s action. Assuming that units of pain correspond to units of intrinsic (dis)value, this implies that the average intrinsic values for Rupert of the states in T and in T* are both − 100, and hence that their difference is 0. Just as in the case of EXPLAIN*, the upshot is that Serena’s action does not harm Rupert.¹⁰