Introduction
The global refugee crisis is undoubtedly one of the most pressing issues of our time. According to the United Nations High Commission for Refugees, “65.3 million individuals were forcibly displaced in 2015 as a result of persecution, conflict, generalized violence, or human rights violations,” which makes the number in 2015 higher than any other year since the end of World War II (UNHCR Global Trends Report 2015, 1-5). Despite the suffering experienced by these individuals and what seems to be a clear moral failure of some sort, affluent states are often quite reticent to let in more refugees.
Might there be some presumptive moral right that can justify permission for states to refuse entry to even the most desperate refugees? Christopher Heath Wellman has suggested that the right of a state to control its borders follows from the principle of freedom of association, and has reached the “stark conclusion” that all legitimate states have the presumptive right to refuse entry to “even refugees desperately seeking asylum from incompetent or corrupt political regimes that are either unable or unwilling to protect their citizens’ basic moral rights” (2008, 109). However, Sarah Fine has critiqued Wellman’s argument from freedom of association for its failure to take into account the potential harm that befalls refugees since “the very act of excluding people may thwart their interests” (2010, 345-347). It is certainly clear that refugees suffer harm by not being granted asylum in affluent countries; by any definition of well being, they would be better off if they were able to secure refuge in a safe country. Yet it is much less clear, I think, whether refugees ought to be considered to be actively harmed by such exclusion, or whether affluent states are merely failing to aid. Wellman seems to assume that states have merely positive duties to aid, whereas Fine, on the other hand, (at least implicitly) suggests that exclusion is actively harming the refugees, thereby violating a negative constraint against doing harm. Furthermore, it is highly contentious whether the distinction between doing and allowing harm ought to be morally relevant in a moral theory. Before being able to answer this question, we require a better grasp of the conceptual distinction between doing and allowing.
My aim in this paper, then, is threefold: to critically analyze the distinction between doing and allowing, to discuss the moral relevance of the distinction, and finally to discuss the bearing of these previous conclusions on the permissibility of a state’s moral right to close its borders to potential immigrants. I will argue that closing borders to refugees is a case of doing harm on the part of affluent states rather than a merely allowed harm, and that this is a morally relevant concern, which in turn places the threshold for justifying any exclusion of potential immigrants at a much higher level. In the first section I will discuss the conceptual distinction between doing and allowing. The second section will focus on a deontological defense of the doctrine of doing and allowing. More specifically, I will argue that Bennett’s positive/negative distinction is morally relevant for the way it normatively restricts the autonomy of agents. I will also argue that the DDA is a necessary component of a moral theory that reflects the elevated moral status of human beings as inviolable rational beings. These sections will draw heavily on the work of Jonathan Bennett, Philippa Foot, Jeff McMahan, Frances Kamm, and Fiona Woollard. In the third and final part of the paper, I will apply these arguments to refugee policy in the real world and discuss the moral permissibility of exclusion, and will conclude that affluent states actively harm refugees by excluding them and that there is therefore a stringent negative constraint on Wellman’s freedom of association argument.
1. CONCEPTUAL ANALYSIS OF THE DOING/ALLOWING DISTINCTION
1.1 Introduction to the Doing/Allowing Distinction
Before being able to determine whether there is any moral relevance to the distinction between doing and allowing harm, we must first have a better conceptual grasp of what it actually means to do something or make something happen, on the one hand, and what it means to allow something to happen or to let something happen, on the other. There does seem to be an intuitive sense of what it means to do harm, and what it means to merely allow harm. As Jonathan Bennett points out, descriptions with respect to a given state of affairs that results from an agent’s actions fall under either one of two types, doing/making or allowing/letting/permitting (1995, 62). To illustrate the intuitive distinction, consider the following pair of cases, in which Jane is out fishing at sea on her boat when she realizes that she has forgotten her vital medication at her beach house, which she needs quickly in order to survive:
Crush: On her way to shore, Jane sees Sam swimming in the water, blocking her way to the shore. She realizes that if she takes the time to go all the way around her, she will not make it to her medication in time and she will die. She navigates the rough waters to steer the boat straight toward the Sam. The boat hits Sam, knocking her unconscious and causing her to drown.
Lifesaver: On her way to shore, Jane sees that Sam’s boat has been sunk by a large wave and Sam will drown unless Jane stops to throw her a lifesaver and pull her onto the boat. Doing so will prevent Jane from reaching her medication in time and she will die. Jane goes straight to shore and Sam drowns.
Intuitively it seems that in the first case, Jane clearly does harm to Sam. In the second case, by contrast, Jane merely allows the harm to befall Sam. More precisely, in Crush, Silvia kills Georgia, whereas in Lifesaver, Jane merely lets die. Yet despite what seems like a clear intuitive sense of which cases ought to be considered doing and which ought to be considered allowing, upon further reflection it becomes much more difficult to pin down the distinction in a principled manner. One of the most obvious ways of understanding the doing/allowing distinction is in terms of action and inaction, respectively. We might be tempted to think that one allows harm to occur when one chooses inaction despite knowing that such inaction will result in the harm, and one does harm when one chooses action despite knowing the harm will occur.
However, other examples seem to separate the action/inaction distinction from the doing/allowing distinction. As Philippa Foot notes, although allowing perhaps tends to be associated with inaction and doing with action, “there is no other general correlation between omission and allowing, commission and bringing about” (1967, 5). For instance, suppose that, for whatever reason, a famous actress chooses to lie in bed (i.e., remain inactive) and thus misses her show. It does seem, intuitively, that the actress has done more than merely allow the show to be ruined. Although she remained inactive, we might say that she has done something to the show rather than allowed something to happen to happen to the show. It might very well be that a fact about an agent’s behavior should be classified as a doing but not as an action, or that another fact should be classified as an allowing despite being an action. Therefore, it seems that we should keep the action/inaction distinction separate from the doing/allowing distinction. While the latter, thus far, seems to track our intuitions about moral blameworthiness, the former does not necessarily do so.
1.2 Sequences
In this section, I will follow Foot and Woollard in suggesting that the doing/allowing distinction is best understand in terms of sequences of events that lead to the relevant harmful outcome. Woollard, drawing on the original suggestion of Foot and refining it, has argued that we should distinguish between doing and allowing in terms of the way that the agent is relevant to the sequence of facts that lead to the upshot in question (2015, 25-26). In Crush, the sequence leading to the harmful upshot might look something like this:
Jane steers the boat toward Sam. → The boat crushes Sam. → Sam is rendered unconscious. → Sam drowns.
Similarly, in Lifesaver, the harmful upshot might be explained sequentially as:
The storm creates large waves → The large waves sink the boat → Sam drowns.
While these seem to be the intuitively correct events to describe what happened when we ask why Sam drowned in each case, it remains unclear what it is about these facts that make them suitable to belong in the sequence. Why, for instance, is the fact that Jane refrains from throwing the lifesaver to Sam not included in the sequence? Woollard argues that facts should be included in the sequence if they are either positive facts about the agent’s behavior, or if they are negative facts about the absence of barriers to harm and a) the victim has some right to ownership over the barrier or b) the barrier to the harm requires no further sacrifice from the agent (2015, 35). Any other facts that do not meet these conditions are not suitable to be in the sequence.
It is important here to note that a fact about an agent’s behavior must be directly connected to the harmful outcome through a chain of other facts that belong in the sequence. If, say, the fact about her behavior is positive but it is relevant to the upshot through a negative fact, then the agent does not count as doing harm; the “chain” connecting the relevant action of the agent has been “broken” because it is only relevant to the upshot through a “mere condition” (Woollard 2015, 31-33). In the following sections, I will come back to the concept of sequences to demonstrate how they can account for shortcomings in other analyses of the doing/allowing distinction.
1.2 Bennett’s Positive/Negative Analysis
The main way that facts about an agent’s behavior are to belong in the sequence leading to some harmful outcome is by being positive facts. To understand this, we need an account of what it means for a fact to be positive. Bennett suggests that an agent should be considered to be doing when a positive proposition describes the way that the action is relevant to the upshot, and that the agent is allowing when the behavior is negatively relevant to the upshot (1995, 88-89). He defines negative and positive facts as follows:
A proposition about how Agent moves at T is negative if it is fit to be represented by a region that covers nearly the whole of the space of possibilities for him at that time. That is, a proposition and its complement are positive if they divide the Agent’s behaviour space extremely unevenly; the highly informative one is positive, and its almost empty complement is negative (1995, 91-92).
Implicit in Bennett’s description of the positive/negative distinction with respect to propositions is the metric that he uses to measure the “space of possibilities” for the agent, namely, the possible ways in which the agent may physically move. To illustrate these ideas, let us return to the original cases Crush and Lifesaver. In Crush, Jane is relevant to the harmful upshot at time T2 (Sam drowns) through the proposition about her action (steering the boat), at time T1. To determine whether the proposition Jane steers the boat is negatively or positively relevant to the harmful upshot, we must compare that proposition to the possible ways that Jane could have moved, and ask how those movements would have affected the upshot (i.e., how the sequence would have progressed from T1 to T2). In other words, if we consider how Jane might have acted differently, under how many of those possibilities would Sam still have drowned in the same manner and at the same time? It seems obvious to me that Jane’s steering the boat in that precise way at T1 is one of the few actions that she could have chosen that would have resulted in Sam’s death at T2. Had she steered the boat differently, stopped the engine, or done just about anything else, Sam’s drowning would not have occurred at T2 (by supposition).
Thus it seems that Jane’s behavior space, made up of the possible ways that she could have moved her body, is unevenly divided between the ways that would have led to Sam’s death at T2 and ways that would not have led to Sam’s death at T2. In other words, there are many ways that she could have acted that would have resulted in Sam remaining alive, while there are very few ways that she could have moved that would have resulted in Sam’s death. Because the proposition Jane steers the boat is within the very small subsection of her behavior space, we can conclude that Jane is positively relevant to Sam’s death; Jane kills Sam.
However, it seems Bennett’s use of the positive/negative distinction to analyze the doing/allowing distinction might be objected to as follows. Recall the story of the actress who remained in bed rather than going to perform at the theatre. If we analyze this case in terms of negative and positive facts, it might seem that we arrive at the conclusion that the relevant fact about the action of the actress, refraining from going to her performance, is negatively relevant to the harmful outcome, the ruination of the show. The same outcome would have obtained had the actress moved in nearly any other way other than the few ways that would have resulted in her moving from her home to the theatre to perform. Therefore, we must conclude that, according to our analysis thus far, the negative proposition about the action falls on the allowing side of the doing/allowing distinction and the actress has merely allowed the show to be ruined. Bennett’s analysis here yields a counterintuitive conclusion; as I have suggested earlier, it seems clear that the actress has done more than merely allowed the show to be ruined.
Bennett’s account also seems to get it wrong in the opposite direction, so to speak, when the use of the positive/negative distinction classifies an intuitively allowed harm as an actively done harm. Consider the following case:
Hide: Jane is being pursued through the forest by an enraged bear. She sees a small cave in which she can hide if she remains perfectly still, but knows that if she does so the bear will continue straight ahead to where Sam is having a picnic. Jane hides in the cave and Sam is mauled to death by the bear.
According to Bennett’s account, Jane is positively relevant to the harmful upshot in this case. The relevant fact about her behavior, i.e. that she remained still, is a positive fact because it represents the small subsection of her possible movements that would have resulted in Sam’s death. Had she moved in any other way other than hiding and remaining still, the bear would have killed her instead. Thus it seems that Bennett would be obliged to conclude that Jane has made the harm happen to Sam, rather than merely allowed it to happen. But again this is absurdly counterintuitive; it seems, rather, that Jane has merely allowed Sam to be killed by the bear.
We can again draw on Woollard’s account of sequences in order to avoid this objection. In Hide, although the fact about Jane’s action is positive (she remained still), the fact that connects her action to the upshot (there was nothing to distract the bear from heading toward Sam) is negative. It does not give us anything more than a negative condition, one of the many that would have been necessary for the upshot to occur. For instance, it also had to be the case that there was not a sudden forest fire between the bear and Sam, or that there was not a bear trap. But when asking why the upshot occurred, it would be absurd to include such negative facts in the sequence. Therefore, the chain of facts that connects the relevant action of the agent is broken by a negative fact about the background conditions, and the agent merely allows harm; Jane merely allows the bear to kill Sam.
1.3 Withdrawing Aid and Denying Resources
On which side of the doing/allowing distinction do cases of withdrawing aid or withdrawing barriers to harm fall? It seems that some cases of withdrawing aid intuitively should be considered cases of doing harm while others should be cases of merely allowing harm. Consider the following cases:
The Snake Charmer: A snake charmer is hired to deal with a poisonous snake infestation in the village. He places the snakes in a large cage outside of the town and locks the gate. He later returns and removes the lock, allowing the snakes to escape and return to the village. Many people are bitten and die as a result.
The Mountaineer: A mountaineer is climbing near the summit of a mountain when he notices a massive boulder about to roll down the slope and crush many people in the small town below. He bravely holds the boulder in place, but eventually he becomes cold and hungry. He removes his body from the path of the boulder and many people die.
Prima facie The Snake Charmer and The Mountaineer appear to be quite similar; both involve the removal of barriers to harm and both result in the death of many individuals. Intuitively it seems that the snake charmer kills the villagers, whereas the mountaineer merely allows people in the town to die.
Speaking in terms of positive and negative facts, it can be seen that neither the snake charmer nor the mountaineer are directly positively relevant to the harmful upshot; the relevant fact about the mountaineer’s movement is negative, while the
Which features of these cases, then, might separate our intuitions with respect to the doing/allowing distinction? Jeff McMahan has attempted to explain the difference in terms of the status of the aid and the relationship of the agent to the resource. He argues that when an agent removes a barrier to harm that she has not provided or when she removes a barrier to harm that she has provided but was “complete and self-sustaining,” the agent actively does harm to the victim; on the other hand, if the barrier to harm requires “further contributions” (i.e. it is not complete and self-sustaining) in order to prevent the harm, the agent merely allows harm (1993, 257). McMahan’s analysis does seem to accurately track the differences in these two cases. The snake charmer’s action should be counted as an instance of doing harm because, although he has himself provided the protection from the lethal threat, it was complete and self-sustaining. In other words, the lock on the gate required no further contribution from the snake charmer to continue to prevent the snakes from escaping and killing the villagers. By contrast, we can see that McMahan’s distinction characterizes the mountaineer’s action as an allowed harm, since the mountaineer would be required to constantly be contributing to make the barrier to the harm (i.e., his body) remain effective.
Yet in other cases it seems that McMahan’s account gets the answer intuitively wrong. Namely, it misidentifies cases that intuitively seem like allowed deaths as killings:
The Impoverished Village: Jane finds that ten percent of her income is being used to save the lives of people in a remote impoverished village. She can phone her lawyer and tell him to cancel the monthly donation. Jane phones her lawyer and the people in the village die of starvation (McMahan 1993, 258).
In this case, it seems as if one merely lets the impoverished villagers die. But because the fund will already be used to save the lives of the people (it is complete and self-sustaining), McMahan’s account yields the conclusion that one kills the villagers. To attempt to deal with this problem, McMahan draws a distinction between what he calls “operative” aid and “as-yet-inoperative” aid, where cases of operative aid can be already used by the potential victim of the harm, and cases of as-yet-inoperative aid require some further aid or protection of aid from the benefactor (1993, 260-261). McMahan concedes that thus far his account of withdrawing aid might appear to yield the conclusion that one is killing the people in the impoverished village by pressing one’s claim to prevent the money from being sent, despite the fact that this seems intuitively incorrect. However, he draws on his notion of operative aid in order to resist this conclusion by suggesting that the aid has not yet become operative; the impoverished people still require some further aid from the benefactor.
McMahan claims that his notion of operativeness allows him to avoid misclassifications of cases. However, despite the plausibility of McMahan’s account in tracking the doing/allowing distinction in the cases discussed thus far, it struggles to explain other cases in which it becomes even less clear whether the barrier to harm should be considered “complete and self-sustaining” or requiring “further contributions… to be effective.” Kai Draper points out this problem with McMahan’s account of withdrawing aid with the following case, where
Intercept II: Sam is slowly starving to death, so Jane sends her a check for enough money to buy food. After Sam has received the check and cashed it, Jane changes her mind and takes the money back. Sam dies of starvation (2005, 265).
It is not really clear how McMahan’s account might cope with this case. It seems that comparing Intercept II with The Impoverished Village highlights the fact that McMahan’s account struggles to draw the line between operative aid and as-yet-inoperative aid. As Draper asks rhetorically, “does the aid become operative when Jones receives the check, cashes it, purchases food, or begins to eat” (2005, 265)? I think McMahan would be hard-pressed to satisfactorily answer this question by drawing on the notions of operative and as-yet-inoperative. Why, for instance, does the potential victim require further aid when the trust fund is set to donate money, but not when he has received and cashed the check? It seems to me that if the trust fund is set to donate money, it does not require any further action from the agent.
1.5 Rights and Sequences
In the previous two sections, I have drawn attention to ways in which the
Draper suggests that we might be better equipped to understand how cases of withdrawing aid should by located in the doing/allowing distinction by abandoning discussions of the concept of “operative” and instead analyzing such cases in terms of rights. The relevant conceptual difference between cases in which the check has been sent but not yet received and cases in which the check has been received and cashed, then, might rather be that in the former instance the money remains the property of the donor while in the latter instance the ownership of the money has been transferred (Draper 2005, 269-270).
2. THE DOCTRINE OF DOING AND ALLOWING
2.1 Intuitions and the DDA
With the preceding analysis of the conceptual distinction between doing and allowing in mind, I now turn to the question of its moral relevance. In the following sections, I will argue that the doing/allowing distinction does have moral relevance. According to the Doctrine of Doing and Allowing (DDA), it is generally more difficult to justify doing certain harm than allowing the equivalent harm. In other words, the DDA maintains that, ceteris paribus, the moral restriction against doing harm is more resistant to other the intervening factors such as self-interest or maximizing well-being (Kai Draper 2005, 253). It is important here to note that the DDA does not maintain that doings are absolutely impermissible and allowings are absolutely permissible. It is clearly not the case, that all when everything else is equal, a harm that is done will be impermissible while a harm that is allowed will be permissible. I am clearly not morally permitted to kill someone to avoid stubbing my toe on the door. Yet neither am I morally permitted to refrain from saving someone from dying on the grounds that in doing so I will also stub my toe on the door. Similarly, restrictions against doing harm to others are not absolute. If in the process of saving my own life, or