1. Pure risks and harm

Beth drinks too much, gets into her car and attempts to drive home. With impaired reflexes and some swerves here and there, she continues her drive. On Beth’s route is a pedestrian on the sidewalk, Adam, whom she narrowly misses. Adam is completely unaware of Beth’s presence. One question we can ask is whether Beth’s behaviour is objectionable. On a number of plausible accounts, Beth wrongs Adam¹For example, Feinberg (1987: 34) states that ‘one person wrongs another when his indefensible (unjustifiable and inexcusable) conduct violates the other’s right’. We could plausibly think that Adam has a right against Beth treating him this way. Further, a contractualist account states that X wrongs Y when X does not treat Y as someone with the standing as a person to be owed justification (Kumar 2015: 30). It is not relevant for my purposes to determine which account of wronging is superior, only that Adam’s behaviour is wrongful.. A further question is whether the risk of harm that Beth imposes on Adam while drunk driving can itself constitute a harm to Adam. A number of authors (Finkelstein 2003: 972, Oberdiek 2017: 71, Placani 2017: 86) have recently argued in the affirmative, and have endorsed the following thesis:

(Risk Harm Thesis) An imposition of a risk of harm can itself constitute a harm.

I will call this ‘RHT’ throughout. This paper argues that three recent attempts to ground RHT fail. In §2 I argue that RHT is not true on a subjective understanding of risk, and in §3 I argue that it is not true on an objective understanding of risk. In §4 I conclude the discussion.

Whether RHT is true is of great practical import for at least three reasons. First, if a risk of harm were itself a harm, we would have reason to avoid imposing risks on people (over and above any other reasons we may otherwise have) since we have reasons to avoid harming them. Second, many people believe that victims of wrongdoing are owed compensation only if they were also harmed²For a contrasting view, McCarthy (1996) questions the necessity of harm for liability and argues that we can owe people compensation for imposing risks on them even if we do not harm them by doing so. Further, if RHT is false, we may have to give up on the requirement of (foreseeable) loss or harm as a necessary condition for compensation, as this may have been thought necessary for compensation in part because of the belief that risks of harm are themselves harms.. For example, in tort law a claimant is owed compensation only if there is a foreseeable loss or harm. So, if RHT is true, many more actions would demand compensation than if it were false. Third, whether RHT is true will be of broad interest for person-affecting views in ethics, which state that an act can only be bad if it is bad for someone. If RHT is false, then an imposition of a risk cannot be bad for someone in virtue of its purported harmfulness. This would thereby limit the scope of available arguments in the ethics of risk.

For the purposes of this paper, a risk of some event is a probability of harm occurring³I adopt this definition to keep in line with the target views of this paper (Perry 1995: 321, Finkelstein 2003: 967; Oberdiek 2017: 17, Placani 2017: 78).. This paper focuses on ‘pure risks’ of harm. These risks do not result in the harm that they threaten, nor are the potential victims of the risk aware of its imposition or have any beliefs about it (Oberdiek 2017: 68). The three authors targeted in this paper’s discussion endorse RHT for cases of pure risk (Finkelstein 2003: 971, Oberdiek 2009: 374, Placani 2017: 91). This can be contrasted with ‘impure risks’, which lead to the harm that they threaten (Thomson 1986: 173). Throughout, the focus is on cases where risks are culpably imposed.

Traditionally a distinction is drawn between subjective and objective accounts of probability⁴i will not attempt to provide a full account of the distinction here, but each of the three authors employs this distinction (Finkelstein 2003: 973, Oberdiek 2017: 21–34, Placani 2017: 79).. A subjective account of probability makes essential reference to the beliefs of an agent. A subjective probability is an agent’s degrees of belief or credence in some event’s occurrence. For example, Jane believes that the probability that it will rain tomorrow is 0.6. The probability of the event is a reflection of her credence that the proposition ‘it will rain tomorrow’ is true. Agents can arrive at different beliefs about the probability of an event given their evidence and prior beliefs. As such, there is no ‘true’ subjective probability of an event.

An objective account of probability focuses on a mind-independent feature of the world (Gillies 2000: 1–2). The relative frequency account of probability is adopted in some form by each of the three target authors (Finkelstein 2003: 973, Oberdiek 2017: 35–65, Placani 2017: 79). On this view, the probability of some event is its relative frequency. For example, as the number of tosses of a fair coin approaches infinity, the relative frequency of ‘coin landing tails’ approaches 0.5. As such, we can say that the probability of a ‘coin landing tails’ with any one coin toss is 0.5.

There is considerable disagreement about what exactly constitutes a harm, but all three of the authors I am targeting adopt Joel Feinberg’s account, which holds that a harm is a setback to the legitimate interests of an agent (Feinberg 1987: 36). Feinberg draws a distinction between ‘welfare’ and ‘ulterior’ interests. Welfare interests need to be secured in order for individuals to pursue their goals and to lead flourishing lives. Examples of these interests include physical health and freedom from coercion (Feinberg 1987: 37). Ulterior interests relate to those goals and projects that individuals wish to pursue, such as professional achievement or political power. For those who defend some version of it, RHT applies to welfare interests.

2. Why subjective risks do not harm

John Oberdiek provides an account of risk that builds on a ‘reasonable person’ evidence-relative perspective (2017: 35–66). This characterization of risk is subjective in that essential reference is made to the evidence a reasonable person would use to ground their estimation of a risk⁵In determining the magnitude of a risk, Oberdiek argues that we should characterize the risk in a way that is justifiable to the imposer’s and victim’s perspectives (2017: 61–63). This avoids risk imposers ‘playing down’ a risk. Although this view is ‘thicker’ than a minimalist account of subjective probability, such as the one canvassed in §1, essential reference is nonetheless made to the beliefs held in a reasonable person evidence-relative perspective.. This does not preclude reference to objective probabilities, as Oberdiek states that, where available, estimations of risk may be grounded in objective probabilities (2017: 49). According to Oberdiek, a risk of harm can itself be a harm in virtue of setting back an agent’s interest in having a sufficient number of safe options to choose between. This account is grounded in Raz’s (1986: 370–399) conception of autonomy, which requires that individuals have enough options available from which to plot their own life. Oberdiek argues that some impositions of a risk of harm can make options unsafe. For example, Beth laying a trap on a path makes Adam’s option of using that path unsafe (Oberdiek 2017: 86). Beth makes safe passage unavailable to Adam, and this in turn sets back his interest in having a range of acceptable alternatives from which to choose. It is for this reason that Oberdiek argues that the risk itself can constitute a non-material harm (2017: 71).

This view is vulnerable to the following objection:

(Interference Objection) A risk of harm, in itself, cannot interfere with a target individual’s interests in order to set back those interests.

To motivate this objection, consider the following case:

(Russian Roulette Aggressor) Completely unbeknown to Adam, who is enjoying the weather on a park bench, Beth is aiming a six-shooter at him with the intention to kill him. Beth loads a single bullet into the cylinder. However, Beth in fact loads a blank, but does not realize this. Reasonably believing that there is one bullet and five empty chambers in the cylinder, Beth pulls the trigger.

Beth’s action is clearly subjectively wrongful. But in what sense can we say that Beth imposes a risk of harm on Adam? On a subjective understanding of risk, Beth is imposing a risk of harm on Adam, since Beth’s credence that her pulling the trigger will kill Adam is 0.167 (a 1-in-6 chance of a ‘bullet’ firing). Yet, if the gun had fired, Adam would be unscathed. The gunsmoke would help Beth confirm her belief that she had shot Adam. But there is no objective chance of Adam being shot; there is merely a subjective risk at the level of Beth’s credence.

A subjectively imposed risk is unable to set back an agent’s interests, because a mere subjective risk of harm cannot itself interfere with the interests of the victim in order to set these interests back⁶Parr and Slavny (2019: 82) press the same point when they say that ‘many pure risks do not affect their victims, and instead exist only in the mind of the risk-taker’. The aim of these authors, however, is to establish when pure risks of harm are wrongful, and not whether a risk of harm can itself constitute a harm.. This is because the subjective risk exists only in the mind of Beth. A credence itself cannot plausibly set back anything. Therefore a subjective risk of harm is itself not a harm.

The autonomy account falls foul of the Interference Objection owing to its reliance on a predominantly subjective account of risk⁷One may wonder whether the autonomy account would fare better against the Interference Objection if an objective account of risk was adopted instead. The objective risk would not exist in the mind of the risk imposer, but the account would be vulnerable to the objections to objective risk accounts in §3.. From the reasonable evidence-relative standpoint of Beth, she believes that the risk that she is imposing is grounded in an objective 1-in-6 chance that the bullet fires. Beth is of course mistaken about the objective probability, but her mistake is not unreasonable given the evidence that is available to her.

A reply is that the sort of interest that is set back, namely the interest in having a sufficient range of acceptable options to choose between, is non-material. As such, there need not be any ‘contact’ between a risk and the victim’s interests. Non-material interests are a part of a person’s ‘normative life’: the part of one’s life that goes beyond their experiential and biological existence⁸For example, Thomas Nagel writes that ‘A man’s life includes much that does not take place within the boundaries of his body and his mind, and what happens to him can include much that does not take place within the boundaries of his life’ (1979: 5). Examples of actions that may cross the boundary are deception and betrayal.. Having an acceptable range of options from which to choose appears to be such an interest, as an acceptable menu of options lies outside of the mind and the body. But it is implausible that a subjective risk can interfere with a person’s option in the manner that a trap does. The trap example suggests that the risk itself makes an option defective. But there is no way for this to happen given that a subjective probability exists in the mind of the risk imposer.

A response could be that it is the recognition of the risk by the target individual that makes their options unsafe and thereby harms them. If Adam were to realize that the gun was pointing at him, the option of sitting on the bench would be thereby rendered unsafe. However, the subjective risk will nonetheless still exist in the mind of the risk imposer. Instead, the awareness of the imposition of risk can lead to harms distinct from the risk itself, such as fear or anxiety (Nozick 1974: 66), which will transform the case into one of an impure rather than a pure risk (Thomson 1986: 173).

3. Why objective risks do not harm

Finkelstein (2003: 973) argues that a risk of harm must be objective for it to set back a welfare interest. She states: ‘agents have a legitimate interest in avoiding unwanted risks. A person who inflicts a risk of harm on another damages that interest, thus lowering the victim’s baseline welfare’ (2003: 966). Finkelstein also argues that, other things being equal, we would prefer to receive the chance of a benefit and disprefer the chance of a harm. A satisfied preference benefits us and a thwarted preference harms us. Therefore, as a chance of a harm thwarts our preference to avoid harm, it sets back a welfare interest.

An immediate issue with objective risks can be drawn out with the following case:

(Deadly Bingo) Beth has a sophisticated lottery ball machine that triggers a loaded gun the moment ball number 13 is drawn. The gun is pointing at Adam, who is sitting on a park bench blissfully unaware of Beth’s contraption. Once the machine is turned on, twenty lottery balls spin in the container.

Here there is a live possibility that Adam could be killed by Beth’s contraption⁹Although it is not possible to recast and engage with the debate here, I will assume that objective chances can be compatible with determinism (Glynn 2010, Bradley 2017). This is also an assumption made by the three authors I engage with here.. We can also suppose, following Finkelstein, that Adam would prefer that an objective risk of harm not be imposed on him. Since Adam’s preference appears to be thwarted, Beth’s imposition of risk appears to harm Adam. But then we can ask: if an imposition of risk is itself a harm, when does it begin to be so?

One immediate answer is that the risk of harm harms at the establishment of the risky contraption. But this is not plausible, since there is no risk before the device has been activated.

A second answer is that the risk harms Adam the moment Beth presses the ‘start’ button on the lottery ball machine, causing the lottery balls to tumble around. The use of the contraption is clearly wrongful, but the objective risk represented by the tumbling balls does not itself interfere with Adam’s interests. The following example explains why. Suppose that a perfect predictor, knowing all antecedent causes, could predict whether ball number 13 would be drawn when Beth turns on the machine. Since objective risks are mind-independent, the predictor’s knowledge does not interfere with any causal processes. Suppose that the predictor perfectly predicts that ball number 13 will not be drawn. Does the objective risk nevertheless harm Adam? Since it is perfectly predicted that the risk is ‘impotent’, the risk’s impact on Adam’s interests is the same as pointing a certainly empty gun at Adam and pulling the trigger. A rational guardian of Adam’s interests who is aware of the perfect prediction (while Adam himself is not) would not be willing to pay anything to remove the risk itself, because the risk will not result in Adam being shot. But if an objective risk of harm is itself a harm, then, other things being equal, it would be rational to want to remove it. A response is that, from the imperfect perspective of agents in the world, it is not possible to perfectly predict the outcome, and it is our imperfect knowledge about objective risks that explains why the risk itself is harmful. If this is true, then the view will fall back to a subjective understanding of risk, and it has been established in §2 that a subjective risk cannot itself be a harm.

A third answer is that the harm of the risk occurs at the moment that the ball is drawn, leading to the simultaneous triggering of the gun. But at the moment the ball is drawn, there is no risk. The drawing of the ball is the result of a chancy process, but the event of the drawn-ball simultaneously triggering the gun is not itself risky. The lottery has been settled at the moment the ball is drawn. It is no different from a modification to Russian Roulette Aggressor, where there is one live bullet in the cylinder, and the cylinder is spun rapidly before pulling the trigger. The mind-independent causal process of the lottery ball machine is just as separate from Adam’s welfare interests as Beth’s credence is from Adam’s welfare interests.

One might argue that it is too quick to disavow an objective risk based account. An alternative account may provide a better explanation of the interests that are set back by an objective risk.

A further attempted defence of RHT comes from Placani (2017: 86–89), who argues that an imposition of a risk of harm can itself be a harm in virtue of being wrongful¹⁰Parr and Slavny (2019: 78) consider a reversal of this claim when they outline The Harm Account, which states that ‘pure risking is wrong, when it is, because risks are harms’.. Placani employs a hybrid account of risk that draws upon objective frequentist risks (2017: 79). In particular, an imposition of a risk of harm wrongs the victim by aiming at unconsented harm, and thereby harms the target by setting back a dignitary interest. This dignitary interest is a ‘core human interest [that] demands that we treat others in such a way as not to violate their inherent moral worth’ (2017: 87).

Although an imposition of a risk of harm may be clearly wrongful, as in Russian Roulette Aggressor, a wrongful attempt to impose a risk of harm need not itself constitute a harm. Risk can be separated from wrongful risking behaviour. This can be done by identifying a particular risky action type, such as ‘aiming a Russian roulette gun at someone and pulling the trigger’, ‘or driving home whilst inebriated’. Whether the action was wrongful does not rest on whether there was in fact an objective chance that the victim would be harmed. Beth wrongs Adam by pointing (what she believes to be) a loaded revolver at Adam. This is true whether or not an objective risk is imposed. This action may be an affront to Adam’s dignity – for Beth acts as if she can arbitrarily take Adam’s life into her hands – but the wrongness of this action does not depend on there being an objective chance of Adam being shot. As such, the dignity account fails to provide grounds for thinking that a risk itself can harm, when the work can be done exclusively with reference to wrongful (and risk-free) acts.

Another route is to claim that wrongful acts, in virtue of being wrongful, set back the victim’s interests and are thereby harms. Beth in Russian Roulette Aggressor, for example, wrongfully carries out an act that is widely understood to be risky. Wronging a person can harm, as wronging invades their interests in some respect (Feinberg 1987: 35–36). But this misses the mark, as the harm that arises from a wrongful act itself makes no reference to an objective risk of harm. All this would establish is that wrongful behaviour can harm, not that the risk present in wrongful behaviour can harm.

4. Conclusion

The preceding three accounts struggle to defend RHT against the Interference Objection. Subjective accounts of risk cannot interfere with a person’s interests, as the subjective risk exists in the mind of the risk imposer. This means that the autonomy account fails to provide grounds for RHT. The preference account claimed that a risk of harm was itself a harm in light of an objective risk of harm being something that we disprefer. But on this view there is no plausible answer to the question of when a risk of harm harms the victim. The perfect predictor example demonstrated that an objective risk in itself does not interfere with the interests of a victim, because the deadly bingo will lead to Adam either being shot or escaping unscathed. It is not the probability of harm itself that interferes. The dignity account was unable to establish RHT because a wrongful action that one believes to be risky can be separated from an objective risk of harm.

The above conclusions do not rule out the fact that intentional impositions of risk can be wrongful. But the wrongness of impositions of risks of harm are diminished by the fact that they are not themselves harms. Since a subjective risk of harm cannot itself be a harm, one route forward is to try to identify an interest that an objective risk can set back that is immune from the above objections. But, even if such a view is established, it will struggle to be action-guiding, since whether, and to what extent, one imposes an objective risk of harm is typically not accessible to agents in the world¹¹I am grateful to Emma Curran, Todd Karhu, Chris Marshall, Asbjørn Schmidt, Martin Smith, Tony Zhou and two anonymous referees for their very helpful comments..