Barry and Øverland on doing, allowing, and enabling harm

ABSTRACT In Responding to Global Poverty: Harm, Responsibility, and Agency, Christian Barry and Gerhard Øverland address the two types of argument that have dominated discussion of the responsibilities of the affluent to respond to global poverty. The second type of argument appeals to ‘contribution-based responsibilities’: the affluent have a duty to do something about the plight of the global poor because they have contributed to that plight. Barry and Øverland rightly recognize that to assess contribution-based responsibility for global poverty, we need to understand what it is for an agent to contribute to harm rather than merely failing to prevent it. Barry and Øverland argue that we should replace the traditional bipartite distinction doing and allowing with a bipartite distinction between doing, allowing and enabling. I argue that their discussion represents a significant contribution to this debate. However, more detail on their key ideas of ‘relevant action’ and ‘complete causal process’ is needed. Moreover, in cases involving the removal of barriers, the non-need based claims of those involved matter. Abbreviations: DAD: doing/allowing distinction; DAED: doing/allowing/enabling distinction


Introduction
In Responding to Global Poverty: Harm, Responsibility, and Agency, Christian Barry and Gerhard Øverland address the two types of argument that have dominated discussion of the responsibilities of the affluent to respond to global poverty. The first type of argument appeals to what Barry and Øverland call 'assistance-based responsibilities': the affluent are required to do something about global poverty simply because those in poverty need help to avoid serious suffering and the affluent are able to give such help. The second type appeals to 'contribution-based responsibilities': the affluent have a duty to do something about the plight of the global poor because they have contributed to that plight (2016,8). As Barry and Øverland note, assessing the second type of argument requires an understanding of the difference between contributing to an outcome and simply failing to prevent it. They, therefore, engage with (in their words) 'the literature on the so-called "doing/allowing" distinction' (9). In this paper, I'll outline Barry and Øverland's account of the ways an agent might contribute to a harmful outcome and then discuss some concerns. Barry and Øverland's account of the doing/allowing/enabling distinction Barry and Øverland do not think that talk of the 'doing/allowing distinction' (DAD) is helpful. They wish to replace the traditional bipartite DAD with a tripartite doing/allowing/ enabling distinction (DAED). The DAED recognizes enabling harm as a distinct category, both descriptively and normatively different from either doing or allowing harm.
On Barry and Øverland's view, clear-cut cases of doing harm possess two features: (1) relevant action: 'there is an answer to the question of how she was relevant to [the harm] that refers to something she did ' (2016, 116); (2) complete causal process: 'there is an intact sequence linking the relevant action [of the agent to the harm]' (118). Barry and Øverland note that in simple cases, a complete causal process is 'a physical process involving the transfer of energy and momentum from [agent to victim]' (118).
When both relevant action and a complete causal process are present, we have a clear-cut case of doing harm. When both are absent, we have a clear-cut case of merely allowing harm. Our judgements about such clear-cut cases exhibit two important features, convergence (pretty much everyone agrees about how the cases should be classified) and robustness (these judgements aren't swayed by factors relating to moral culpability: knowledge, intentions, avoidance costs, etc.). In contrast, Barry  Push is a clear case of doing harm. It involves both relevant action (Sue is relevant to Bill's injuries by pushing the cart) (Barry and Øverland 2016, 117, 118) and a complete causal process (energy and momentum is transferred from Sue to Bill) (118). Our judgements that it is a doing show convergence and robustness. As Barry and Øverland note, 'Sue does harm to Billwe know of no-one who would suggest otherwise' (115). And 'Sue does harm in Push whether or not she intends, foresees, or should have foreseen that harm will result from her conduct, and whether or not preventing harm would be costly to her' (115).
Stay Back is a clear case of merely allowing harm. There is no relevant action. Barry and Øverland claim: 'In Stay Back the answer to the how question … will … refer to something that Sue didn't do: her relevance to the outcome is that she did not interpose the rock' (117). Again, our judgements that it is an allowing show robustness and coherence.
In Remove, Barry and Øverland claim that there is a relevant action ('… the answer to how Sue is relevant to Bill's injuries refers to something she does: she removed the rock that would have prevented them' (121)). However, they claim there is no complete causal process:There may be such a process between Sue and something the rock hitsfor example, if Sue removes it by kicking it away and it slams into a nearby tree. However, there is no complete causal process running from Sue through the cart and down to Bill. (121)Our judgements about how to classify this case do not exhibit coherence or robustness. Some people classify Remove as a doing and others as a mere allowing. Barry and Øverland support claims about the judgements of ordinary people with empirical studies from previous work (Barry, Lindauer, and Øverland 2014), and cite examples to show similar disagreement amongst philosophers who have tried to analyse the distinction (2016,122). Judgements about Remove exhibit fragility: there is a much stronger tendency to judge that Sue does harm by removing the rock if she knows that Bill will be severely injured if she does so and the cost of refraining from doing so is small. Barry and Øverland conclude that these intermediate cases should fall into a distinct category: enabling harm. For Barry and Øverland, enabling covers any case in which there is relevant action but no complete causal process. On Barry and Øverland's view, enabling is neither equivalent to doing harm nor equivalent to allowing harm: Agents who enable harm seem, intuitively, to 'contribute' to harm, and in this sense are distinct from those who merely allow harm to occur… But enablers of harm contribute to harm in a quite different manner from doers of harm, since the latter are linked to these harms by a complete causal process. (127) Barry and Øverland hold that there is a moral difference between doing, enabling and allowing harm because of the moral significance of 'giving rise to cost'. 'A person gives rise to cost when his or her location, movements or agency has as a consequence that a person be harmed' (139). When a person gives rise to cost, they may be required to bear greater costs to prevent others from suffering harm. Except in cases of overdetermination, when an agent does harm or enables harm, he gives rise to cost; when an agent merely allows harm he does not give rise to cost. However, how much extra cost the agent may be required to bear depends on their ability to control for giving rise to cost. Barry and Øverland argue that those who do harm have a greater ability to control whether they give rise to cost than those who merely enable harm. They conclude that the DAED is morally significant in several ways. For example, an agent can be expected to bear much greater costs to avoid doing harm than to avoid enabling harm, but greater costs to avoid enabling harm than to avoid merely allowing harm. However, their core interest here is in the claim that enablers may be required to bear greater costs than mere allowers but lesser costs than doers to protect those who are currently under threat of harm. This claim is most relevant to contribution-based responsibilities for the affluent to respond to global poverty (141).
Barry and Øverland's account of the nature and moral significance of the DAED is an important contribution to the debate. However, I'll focus mainly on areas of disagreement or concern.
I found Barry and Øverland's analysis of clear doings and clear allowings too quick. I wanted further detail on both key ideas: relevant action and complete causal processes. Of course, their aim is only to provide enough of an understanding of the DAED to elucidate the discussion of contribution-based responsibilities to respond to global poverty. However, this lack of detail may have implications for their arguments.

Relevant action
Barry and Øverland take their idea of relevant action from J.J. Thomson's suggestion that we can understand the DAD using 'the How Question': how did the agent do or allow the harm? In cases of doing harm, Thomson notes, there is a clear answer to this question: 'a process or, as I'll call it, an enterprise by which we do'. In contrast, Thomson claims that when we allow harm, there is no clear answer to the How Question: 'No how about it!'. Thomson concludes by claiming that harm that we allow comes about 'because of something a person doesn't do ' (1996, 552). Barry and Øverland make important modifications to Thomson's account. They don't claim that when an agent allows harm there will be no clear answer to the harm question. Instead, they hold that when an agent does harm the answer to how the agent was relevant to the harm refers to some action or actions of the agent but when an agent allows harm it will refer to 'no relevant action' of the agent, but instead will refer to 'something [she] didn't do' (117). As I will show, this modification avoids some important problems that Thomson's account facesbut in doing so raises new questions that Barry and Øverland do not attempt to answer, leaving their account incomplete.
It is a mistake to say that when an agent merely allows harm the harm occurs because of something she doesn't do. 3 Sue's failure to interpose the rock is not something she did not do. She did fail to interpose the rock. The harm to Bill did occur because of something Sue did not do: it occurred because someone or something else pushed the cart down the hillbut that is not the answer to how Sue was relevant to Bill's injuries. When we are thinking about how agents allow, we should not say the harm occurs because of something the agent does not do. We should say that the harm occurs because the agent does not do something.
This matters because there are various ways of failing to do something. This is why Thomson original account fails. It's just not true that there is 'No how about it!' when it comes to allowings. 4 When I fail to pick up our children on time, my husband invariably asks the How Question: Was I struggling to start a broken down car? Was I playing with a puppy? There is often a sensible question as to how an agent allowed harm. Indeed, even asking whether Thomson's version of the How Question (how did the agent do or allow harm?) is answered by reference to an act doesn't distinguish doing and allowing: the potential answers to my husband's question both refer to actions.
Barry and Øverland's version of the How Question is much better in this respect: it asks how the agent is relevant to the harm. The answer to this question picks out the relevant fact about the agent's behaviourand whether the relevant fact about an agent's behaviour tells us that he did something or that he did not do something is clearly important to the DAD. However, this has already moved way beyond Thomson's initial suggestion that when an agent does harm there is way that he does harm. The revised account raises new questions: How do we pick out the relevant fact about the agent's behaviour? What makes a fact tell us that the agent did something rather than that he did not do something? Different answers to these questions may result in different classifications of some cases.

Complete causal process
In explaining the claim that there is a complete causal process linking Sue's action and Bill's injuries in Push, Barry and Øverland say … there is an intact sequence linking the relevant action of Sue in Push with Bill's injury. In Push, this complete causal process is a physical process involving the transfer of energy and momentum, from the cause to the effect. (118) However, they do not see transfer of energy as necessary for a complete causal process. They discuss another case: Dislodge: A cart stands at the top of a hill, its wheel held in place by a rock. Sue removes the rock. The cart rolls down the hill and injures Bill, who is sitting there. (119) Barry and Øverland note, 'Sue does not transfer energy to the cart. She is, however, linked to Bill's injuries through a complete causal process. In Dislodge, this process involved the release of stored potential energy….' (119).
So on Barry and Øverland's account, a complete causal process can proceed either by energy transfer or by energy release. I'd have liked to know what these two mechanisms have in common that makes them both able to form parts of a complete causal process. I'd have also liked to know if there are other mechanisms which can make up a complete causal process. I worry about the appeal to notions such as energy and momentum. The notion of an intact sequence which we use when classifying cases as doings or allowing seems to extend to cases where these ideas do not apply straightforwardly. A registrar pronouncing a couple husband and wife, a donation of money through Internet banking, and the enactment of a new law can all be parts of an intact sequence. Barry and Øverland note that they do not want to attempt to extend their analysis to cover such cases (114). Nonetheless, given that the idea of an intact sequence applies so naturally elsewhere, it seems preferable to try to look for some more general analysis. In addition, as I shall show, some of these non-physical ways of being part of an intact sequence may apply in what Barry and Øverland see as the simple cases, leading them to misdiagnose such cases.

Intermediate cases and enabling harm as a third category
Barry and Øverland argue against both the view that all cases involving removing (or preventing the creation of) a barrier to harm are doings (128-132) and the view that all cases involving removing (or preventing the creation of) a barrier to harm are allowings (159)(160)(161)(162)(163)(164)(165). Both these sections are extremely impressive and persuasive. Because I broadly agree with them, I will not discuss them here. They do not discuss the view that some cases involving removing (or preventing the creation of) a barrier to harm count as doings and others count as allowings. I refer to 'cases involving removing (or preventing the creation of) a barrier to harm' instead of 'enablings' deliberately. This is because part of the issue here is that the removal/creation of a barrier to harm may not count as an enabling. On my own view, all enablings count as allowings, but only some cases involving the removal of a barrier to harm count as enablings.
Like Barry and Øverland, I think the notion of an intact sequence is key to analysis of the DAD. Thus, I think their notion of an enabling is useful, but I would tweak this analysis to remove the appeal to 'complete causal sequences'which I argued above is both unclear and too narrow. Where Barry and Øverland hold that An agent enables harm if and only if an action of the agent is relevant to the harm but there is no complete causal process linking the relevant action to the harm.
I suggest An agent enables harm if and only if an action of the agent is relevant to the harm but there is no intact sequence linking the relevant action to the harm.
As Barry and Øverland elsewhere explain 'complete causal sequences' in terms of intact sequences (118), this is still within the spirit of their proposal. Of course, the notion of an intact sequence requires further analysis. I attempt to provide this elsewhere (Woollard 2015).
On this account, it is an open question whether all removals of barriers count as enablings. I do not think they do. Consider Drive-Away (Victor's Car): A boulder is rolling towards Victor who is trapped on the slope below. Victor's car is in the path of the boulder and, if it remains where it is, will bring the boulder to a halt. Bob drives Victor's car away. The boulder hits Victor and crushes him to death. (Woollard 2015, 35) On my view, there is an intact sequence linking Bob and Victor's injuries, even though Bob is only relevant to Victor's death through the removal of a barrier to harm. The absence of Victor's car from the path of the boulder is part of the sequence leading to harm. Thus, Bob's removal of that car is part of the sequence leading to harm. Bob counts as doing harm.
This claim is not intended to be merely a linguistic claim nor a purely moral one. I am not claiming simply that we would intuitively describe Bob as harming Victor (although I think we would) nor simply that Bob would be required to bear greater costs to avoid this behaviour than to avoid driving his own car out of the boulder's path (although I think he would). The claim is that we think that Bob must bear these higher costs because otherwise he would be harming Victor. 5 If I am right, this provides support for my claim that we should switch away from talking about complete causal processeswhere this is understood to point us towards ideas such as energy transference. Drive-Away (VC) is surely one of the 'relatively simple cases', which Barry and Øverland would want to cover in their discussion (Barry and Øverland 2016, 114). But even in such a case, removal of a barrier can count as part of a sequence not because it involves transfer or release of energy, but because it results in the removal of a person's rightful property that was being used to protect him. I argue elsewhere that what matters is whether the relevant facts are 'substantial' enough to count as part of a sequence. Negative facts (such as the fact that a barrier is not present) are usually not substantial enough to count as part of a sequence and thus break the sequence between agent and harm. However, because the default is that our property that we are using remains untouched by others, facts about the absence of such barriers are substantial enough to be part of the sequence (Woollard 2015).
There are three different categories in this part of Barry and Øverland's discussion: (1) intermediate cases (cases about which our judgements don't exhibit coherence or robustness); (2) removals/preventions of barriers (cases where the agent removes, or prevents the creation of, a barrier to harm; (3) enablings (cases where there is relevant action but no intact sequence). Barry and Øverland appear to assume in their arguments that any case that falls into one of these categories falls into the other two: they present the lack of coherence and robustness of cases involving removal or prevention of barriers as evidence that enabling is a distinct category. If I am right that Drive-Away (VC) is not an enabling, then categories (2) and (3) can come apart. At the very least if Barry and Øverland want to argue that any removal of a barrier is an enabling they need to provide further argument. I suspect that this will require further analysis of the notion of a 'complete causal process' and further defence of that notion and analysis. I'd be interested to know whether Drive-Away (VC) is an intermediate case. If it is not, then categories (1) and (2) can come apart; if it isand I am right that it is not an enablingthen categories (1) and (3) can come apart. If the three categories do come apart, then this doesn't undermine Barry and Øverland's overall argument but it does mean that much care must be taken in applying it. First, we need to know more about which types of case exhibit a lack of coherence and fragility to assess the import of these findings. Second, if not all removals of barriers are enablings, we need to be very careful to check whether a given removal of barrier is an enabling before drawing normative conclusions.
I suggested that Barry and Øverland should have discussed the view that some cases involving removing (or preventing the creation of) a barrier to harm count as doings and some count as allowings. On my view, Drive-Away (VC), in which Bob drives Victor's car out of the path of the boulder, counts as doing harm. In contrast, if it is Bob's car that is in the path of the boulder (Drive-Away (BC)) the absence of the car (and thus Bob's removal of the car) does not count as part of the sequence leading to harm (Woollard 2015, 9). Because this is simply a negative fact about Bob's own property, there is nothing to make it substantive enough to count as part of this sequence. Because there is no intact sequence connecting Bob and the harm, Bob counts as enabling harmand on my view merely allowing harm. However, on my view, both the DAD and the action/inaction distinction are morally relevant. This means that allowing harm through inaction is morally distinct from allowing harm through action. While Drive-Away (BC) counts as merely allowing harm, it is not morally equivalent to allowing harm through inaction. Nonetheless, countenancing harm in cases like Drive-Away (BC) is far easier to justify than countenancing harm in cases like Drive-Away (VC). Suppose that Bob has been bitten by a snake and can only save his life if he immediately jumps in the car and drives to the hospital, leaving Victor exposed to the deadly boulders. I think it is permissible for Bob to do so if it is Bob's own car that Victor is using as a shield, but impermissible for Bob to take Victor's car to get to the hospital. 6 Barry and Øverland and I agree that relevant action and intact sequence/complete causal processes matter. We both draw a morally significant line between clear allowings and enablings and another between enablings and clear doings.
However, I think that, while enabling and clear allowing are morally distinct, they have enough in common to both be treated as cases of allowing for the purpose of drawing a binary distinction between doing and allowing. 7 Barry and Øverland disagree. This may be, in part, be simply a disagreement about how to use the word 'allow'. 8 Nonetheless, substantive disagreements appear if Barry and Øverland also hold that all removals of barriers to harm count as enabling harm. As the Drive-Away cases show, not all removals of barriers are morally equivalent. The non-need-based claims (for example, claims based on ownership) which the relevant individuals have over the barrier matter enormously. This has important implications for Barry and Øverland's later discussion applying the DAED to contribution-based responsibility for global poverty. In discussing whether affluent nations count as contributing to global poverty by negotiating unfavourable trade agreements, we need to consider not just (a) are they relevant through action? and (b) is their behaviour correctly characterized as removing a barrier or preventing a barrier from forming? but also (c) what are the claims to this barrier of the various relevant parties? If we hold that the victims of global poverty have a non-need-based claim to the barriers, the behaviour of affluent nations is likely to seem relevantly similar to doing harm. If we hold that the affluent nations have nonneed-based claims to the barriers, then their behaviour is likely to seem more similar to merely allowing harm.

Conclusion
Barry and Øverland rightly recognize that to assess contribution-based responsibility for global poverty, we need to understand what it is for an agent to contribute to harm rather than merely fail to prevent it. Their discussion represents a significant contribution to this debate. However, more detail on their key ideas of 'relevant action' and 'complete causal process' is needed. Even for the type of simple cases on which Barry and Øverland focus, looking only at physical mechanisms such as energy transfer and energy release is not enough. We require a broader understanding of what can make up an intact sequence. In cases involving the removal of barriers the non-needbased claims of those involved matter.

Disclosure statement
No potential conflict of interest was reported by the author.