AN EXTRINSIC DISPOSITIONAL ACCOUNT OF VULNERABILITY

It is common to see vulnerability as either “ontological”or broadly “circumstantial.” Both views capture something morally important about vulnerability. However, there is a puzzle: how can the same concept refer to a necessary ontological fact and to a contingent circumstance? I address two solutions to this puzzle. First, I argue that Mackenzie et al.’s taxonomy of vulnerability is not a real solution (2013). Second, I addressMartin et al.’s dispositional account of vulnerability (2014). For them, vulnerability is both an intrinsic property and a disposition.This supposedly solves the puzzle: vulnerability can be intrinsic and yet be manifest in only some circumstances—such is the nature of dispositions. However, I argue that if vulnerability is indeed a disposition, it is better conceived as an extrinsic disposition (McKitrick 2003). Thus, vulnerability cannot be both intrinsic and dispositional; Martin et al. fail to resolve the puzzle. This, however, is no reason to fret. Indeed, an amended dispositional account of vulnerability, in which it is conceived as an extrinsic disposition, is metaphysically consistent and it satisfies our moral intuitions about human vulnerability, andmore.Given these advantages, I argue the solution to this dilemma is to abandon the ontological conception of vulnerability.

The concept of vulnerability plays an important role in ethical, political, and practical discourse. Ethical guidelines highlight vulnerability as something doctors and scientists should pay extra attention to (NCPHS 1979;CIOMS 2002). Many NGOs make some comment about prioritizing vulnerability in their mission statements ("What Is Vulnerability?" 2017; "WHO | Vulnerable Groups" 2017). Vulnerability is also referred to in law and legal theory to justify protective measures for certain people described as "vulnerable adults" (Dunn, Clare, and Holland 2008) or to assess human rights violations (Peroni and Timmer 2013). In all the preceding cases, the focus is on the vulnerability of human beings or human populations.
References to vulnerability as a morally salient feature are also made in broader contexts. Indeed, vulnerability seems applicable to a wide variety of objects. For example, we say that ecosystems are vulnerable to climate change, that marginalized social groups are vulnerable to systemic injustice or assimilation, that institutions are vulnerable to power struggles within a particular polity. In all these sentences, the term "vulnerability" is used to flag moral or political issues that supposedly warrant additional attention or special care. What this shows is that, at least in everyday language, we use the term "vulnerability" for human beings and for a wide variety of objects for analogous purposes. Even if most of the current theoretical literature about vulnerability pertains only to human vulnerability, the concept seems to have strong potential for a broad application in normative political philosophy. An account of vulnerability applicable to all these objects that would nonetheless track what makes situations of vulnerability, in general, morally salient is desirable. My goal is to provide the conceptual basis for such an account.
My thesis is that we should think of vulnerability as a dispositional property of beings with needs, interests, or integrity, broadly construed. An object is vulnerable, I argue, if it is disposed to be harmed or damaged in certain circumstances. This, in some circumstances, can constitute a wrong. 1 As a disposition, vulnerability is a property of objects that has particular circumstances of manifestation. This means that an object's vulnerability will not always be manifest-that the object will not necessarily be harmed or damaged. Vulnerable objects, however, are such that they are capable of being harmed or damaged because they are so disposed.
This article is mainly conceptual. The goal is to determine what kind or property vulnerability is. However, I think it reasonable to want our conceptualization of vulnerability to help us do the kind of normative work we already use the concept for. Importantly, our conceptualization of vulnerability should help us distinguish instantiations of vulnerability that we think are wrong from those that we think are morally neutral. We might think that we are always vulnerable to some extent, but that this constitutes a wrong only in some circumstances. What I am saying here is that our concept should help us make moral distinctions between just and unjust circumstances of manifestations of vulnerability. My claim is that the dispositional account presented in this article neatly explains how vulnerability can be wrong sometimes even if it is morally neutral or even morally desirable at other times.
Our concept of vulnerability should also be applicable to the kind of objects we already attribute it to. As I suggested, I think there are good reasons to have a concept of vulnerability that is compatible with statements about the vulnerability of objects like groups, institutions, and ecosystems. 2 Notably, I think we should be able to talk meaningfully and consistently about the vulnerability of minoritized social groups like women and racialized groups. In doing so, however, we should avoid the pitfalls of naturalizing or essentializing vulnerability; members of minoritized groups may be vulnerable, but they are certainly not so in virtue of their intrinsic properties. It is circumstances, relations-in brief, extrinsic properties-that make them vulnerable. Unfortunately, as I argue in section 5, the intrinsic ontological views of vulnerability lend themselves to the pitfalls of essentialism by treating vulnerability as an intrinsic and essential property. While I recognize objects like groups, institutions, and ecosystems cannot be described as vulnerable if vulnerability is conceptually linked to the kind of intrinsic ontological properties that make us capable of being harmed, I argue that they can be meaningfully described as vulnerable if we think of vulnerability as an object's disposition to be harmed or damaged in certain particular circumstances. Thus, if my account is conceptually sound, we will have the groundwork for a conception of vulnerability applicable to a wider array of moral and political considerations.
The dispositional account of vulnerability is sketched by Angela Martin, Samia Hurst, and Nicolas Tavaglione (Martin, Tavaglione, and Hurst 2014). Their goal, in using the dispositional account, is to resolve the conflict between two popular conceptions of vulnerability for human beings, conceptions I will call the "ontological" account and the "circumstantial" account of vulnerability. Martin et al. argue that a dispositional account allows us to treat vulnerability as an intrinsic ontological disposition that has relational circumstances of manifestation. In other words, Martin et al. argue that vulnerability is both a permanent feature of humanity and a phenomenon that arises or manifests in certain particular circumstances. This shows, they argue, that the famous debate between ontological and circumstantial accounts of vulnerability is a "pseudo-problem which emerged due to a lack of conceptual analysis" (Martin, Tavaglione, and Hurst 2014, p. 62). Martin et al. are broadly correct in their diagnosis. I also think the dispositional account of vulnerability is a necessary step to extend the scope of our normative discourse about vulnerability. Nevertheless, my goal in this paper is to show that their own version of the dispositional account of vulnerability is conceptually flawed. Vulnerability is not an intrinsic dispositional property. Thus, contrary to what they claim, vulnerability is not and cannot be both an intrinsic property of human beings and a dispositional property. This is because, if vulnerability is a disposition, it is an extrinsic disposition (McKitrick, 2003).
As it is well known, properties can be either intrinsic (they belong to the object, necessarily, in virtue of what it is) or extrinsic (they belong to the object, contingently, in virtue of its relationships with other objects). For example, my mass is intrinsic (I have it in virtue of my molecular make-up) and my weight is extrin- sic (it is a function of Earth's gravity and my mass). By the same token, McKitrick argues an object's vulnerability 3 is a function of its extrinsic properties-it is a function of the object's relationships with other objects or with circumstances. If vulnerability is a disposition, then an object is only disposed to suffer harm when it is in relation to external objects. 4 If this is correct, then vulnerability simply cannot be a disposition and an intrinsic property. This also means that Martin et al. fail to resolve the conflict between ontological and circumstantial vulnerability.
Catriona Mackenzie recently criticized Martin et al. for this failure (Mackenzie 2016). The problem, she says, is that Martin et al. effectively reduce vulnerability to a circumstantial property. Mackenzie is right, but, unlike her, I see no reason to regret this result. Indeed, despite their mistake, Martin et al. give us very important insights for our conceptual and moral reflections about vulnerability. Conceptually, their article provides us with a neat and elegant explanation for the fact that vulnerability seems to be a universal feature of the human condition while still being manifest in only some circumstances. Morally, it shows that an account of normative vulnerability-an account designed to track the kind of manifestations of vulnerability that are morally problematic-is always incomplete if we do not pay close attention to its circumstances of manifestation. Vulnerability, in other words, matters morally because it is causally related to the relationships that hold between the world and vulnerable objects. Both these features are preserved in an extrinsic account of vulnerability. Moreover, as I will argue in section 5, the extrinsic account of vulnerability is compatible with a wider set of objects describable as vulnerable and it does not lead to the political and moral problems other accounts of vulnerability can lead to (Brown 2011;Luna 2009). Mackenzie and others might worry that my view effectively reduces vulnerability to an extrinsic and relational property. Reducing vulnerability to such properties arguably downplays the moral significance of some of our intrinsic properties (e.g., our finitude, our frail bodies, etc.) that allegedly make us inherently vulnerable. My conclusion is that this is no reason to fret. The gains we make with my amended version of the dispositional account should trump such concerns.

IS VULNERABILITY ONTOLOGICAL OR CIRCUMSTANTIAL?
After years of debate, it seems two fundamental intuitions have crystalized into conflicting, or at least competing, definitions of "vulnerability." 5 Discussing this basic distinction between what I call the ontological accounts of vulnerability and circumstantial accounts has become a sort of passage obligé in the literature. Hopefully, I can provide some basis to foresee some closure to this debate. 6 For many philosophers, vulnerability is an essential and universal property of the human condition; it is an ontological feature of humanity. For these authors, human beings are fundamentally and essentially vulnerable (Butler 2012;Callahan 2000;Fineman 2008;Gilson 2011Gilson , 2014Kottow 2003;Lévinas 1968;Rendtorff 2002). 7 As such, most of our specifically human traits like our sociability, our need for care, and our duties to care for others, etc. are a function of our inherent vulnerability. The normative upshot of this conception is that everyone is entitled to proper ethical concern in virtue of one's inherent vulnerability. As such, vulnerability cannot and indeed should not be eliminated. This approach admits of variability. For example, this view accepts that infants are more vulnerable than adults and that people living with disabilities are more vulnerable than people who do not. Nonetheless, everyone is vulnerable to some degree. Since, on this account, everyone is always vulnerable to some degree and since vulnerability is described as an intrinsic ontological property, I'll refer to this type of account as the "ontological account of vulnerability." Others see vulnerability mainly as circumstantial and relational (Chambers 2006;Goodin 1986;Macklin 2003;Shivas 2004;Vrousalis 2013). Goodin, for example, argues "'vulnerability' is essentially a matter of being under threat of harm," which means that there is necessarily "some agent (actual or metaphorical) capable" of causing that harm (Goodin 1986, p. 110, p. 112). Someone's vulnerability arises from this relationship. Others use the term "vulnerable" to describe only those who are particularly at risk in some situation. According to this basic intuition, a person is vulnerable if conditions are such that he or she runs a higher than normal risk of being harmed or wronged (Macklin 2003). Understood this way, vulnerability is not some kind of fundamental frailty. It is a state that necessarily includes a relationship with circumstances. These circumstances include particular socio-relational contexts, but they also include the fact that certain individuals have certain intrinsic particularities (e.g., a medical condition) that make them more likely to be harmed or wronged in general. The important point here is that vulnerability is a feature that arises with circumstances and relations. Because of this, I will call this account "circumstantial vulnerability".
Both these accounts of vulnerability focus on different aspects of the human condition. Ontological vulnerability puts the emphasis on universally shared features of beings like us-all of us have needs and interests that are at risk of being frustrated, and all of us are constituted such that we are susceptible to being physically or psychologically harmed. The advantage of this view is that it enables us to formulate universal duties and responsibilities in terms of protection, or at least recognition, of the vulnerable. Circumstantial vulnerability focuses on particular circumstances or relations that seem to affect certain particular people. The advantage of this view is that it allows us to identify our special duties and responsibilities-those that arise only with particular circumstances, and which make the situation of certain people relatively more urgent.
Both these accounts come with their own problems. For example, some people prefer the circumstantial account of vulnerability because they worry that our definition of vulnerability loses its normative sway if it entails that everyone is vulnerable (Levine et al. 2004, p. 45;Luna 2009, p. 128). Indeed, given that vulnerability is appealed to in order to call for special attention for some people, if everyone is entitled to special attention, it seems that this extra attention is no longer special (Levine et al. 2004, p. 46). In the ontologist camp, some people worry that a focus on circumstantial vulnerability will hide "morally salient features of our humanity [e.g., finitude, fragility, dependence on others] that … are … central to much of our everyday moral discourse and practice" (Mackenzie 2016, p. 84). Others worry that the circumstantial approach reduces vulnerability to something negative, something we need to eliminate. For these authors, at least some manifestations of vulnerability are morally important, if not morally good. For example, the kind of vulnerability that is inherent to intimate and loving relationships or to relationships of care cannot and should not be eliminated (Gilson 2014;Straehle 2016).
As we see, both conceptions come with their advantages and with their disadvantages. It could be the case that we are not faced with an either-or situation. As we will see in the next section, we could follow Mackenzie, Rogers, and Dodds and be content with a polyvalent concept of vulnerability that refers to distinct but overlapping phenomena and that comes with a set of distinct but overlapping moral commitments. The fact that the same concept applies to a wide variety of phenomena and comes with a wide variety of moral commitments is not necessarily a problem. But there is something genuinely puzzling about a concept that refers (1) to a seemingly permanent and necessary fact and (2) to a seemingly contingent and occurrent relation. Ideally, we would have a conceptual solution to this puzzle. As I argue in the following, Mackenzie, Rogers, and Dodds's taxonomy unfortunately does not solve it.

DISTINCT BUT OVERLAPPING KINDS OF VULNERABILITIES
A prominent response to the debate sketched in section 1 was proposed by philosophers Wendy Rogers, Catriona Mackenzie, and Susan Dodds. Rather than identifying necessary and sufficient conditions that would be compatible with all manifestations of ontological and circumstantial vulnerability, Mackenzie et al. propose a taxonomy of vulnerability designed to capture "distinct but overlapping kinds of vulnerability" (Rogers, Mackenzie, and Dodds 2012;Mackenzie, Rogers, and Dodds 2013, p. 7). This taxonomy is meant to provide a vocabulary that describes the different phenomena associated with the concept of vulnerability, but also to help us understand the different moral and political duties associated with different kinds of vulnerabilities (Rogers, Mackenzie, and Dodds 2013, p. 8).
According to Mackenzie et al., vulnerability has three kinds of nonmutually exclusive sources. Some are inherent, others are situational, and, finally, some situational vulnerabilities are pathogenic. A source of vulnerability is inherent when it is related to an object's intrinsic properties-our bodily needs and our dependence on others are inherent sources of our vulnerability. An object is situationally vulnerable when its vulnerability is causally related to certain situations (Mackenzie, Rogers, and Dodds 2013, p. 7). This kind of vulnerability is context specific and relational. Finally, a source of a vulnerability is pathogenic when it stems from "morally dysfunctional or abusive interpersonal and social relationships and sociopolitical oppression or injustice" (p. 9).
This taxonomy incorporates the idea that vulnerability can be conceived as circumstantial (as in pathogenic and situational vulnerability), as well as univer- sal, depending on its source. It also acknowledges that some vulnerabilities are manifest while others seem to be merely latent. Indeed, different sources of vulnerability can lead to occurrent or merely dispositional vulnerabilities. For example, I am dispositionally and inherently vulnerable to hunger-that is to say, some of my intrinsic properties dispose me to feel hunger. However, my situation is such that I am not occurrently vulnerable to hunger-indeed, I am not in a situation in which I am likely to suffer from hunger.
Mackenzie et al.'s taxonomy is interesting because it enables us to acknowledge the complex interplay among different manifestations of vulnerability. It also helps us to see the complexity of responding to vulnerability, as it forces us to reflect on the fact that responding to one type of vulnerability may further another. Thus, this taxonomy is very useful to map the debate and to highlight why different situations of vulnerability raise different moral concerns. In other words, it offers us a concept that captures "different but overlapping" phenomena and it helps us identify which of these phenomena are morally urgent. In other words, Mackenzie et al. provide us with an account that helps us do the normative work we want to do with the concept of vulnerability.
The problem, however, is that this taxonomy does not give us a neat solution to the conceptual debate we started with. When they use the expression "distinct but overlapping kinds of vulnerabilities" (Mackenzie, Rogers, and Dodds 2013, p. 7, my emphasis), it is not clear whether these different "kinds" of vulnerabilities can be captured by the same concept. Indeed, how can a situation of "increased risk of suffering harm" (Mackenzie 2016, p. 88) be captured by the same concept as "the fragility of our embodiment and agency" (p. 84)? On the one hand, we have a relational property that describes a state of affairs by drawing a contrast with some kind of base state of "risk": vulnerability is about increased risk, not simply about plain risk. On the other hand, we are told that vulnerability is an intrinsic ontological property, our "fragility," which, while it may vary among individuals, refers to the metaphysical possibility of suffering harm. Even if the word "vulnerability" is used in both instances, it is not clear that the same concept is at work. The mere possibility of X cannot be expressed by the same concept as the increased likelihood of X to actually manifest. Thus, it seems that someone's underlying propensity to be harmed, in general, is conceptually different from someone's actual and quantifiable risk of being harmed in a particular situation.
Again, we seem to be stuck in the conceptual and metaphysical debate: is vulnerability ontological and intrinsic or circumstantial and extrinsic? Unfortunately, Mackenzie, Rogers, and Dodds do not provide a clear solution to this conceptual dilemma, even if they make some progress on the normative front. Ideally, we would have a single concept that could resolve the conceptual dilemma and that would do all the normative work we want vulnerability to do. In the following, we will see an elegant, but ultimately flawed, solution to the conceptual dilemma, and one that meets some our normative desiderata.

THE SAME CONCEPT WITH DIFFERENT LIKELIHOODS OF MANI-FESTATION
Another and perhaps more satisfying attempt to reconcile the universal and circumstantial conceptions of vulnerability is made by Angela Martin, Nicola Tavaglione, and Samia Hurst. They go further than Rogers et al., who think of the circumstantial and ontological vulnerabilities as "distinct but overlapping kinds of vulnerabilities" (Rogers, Mackenzie, and Dodds 2013, p. 7) and claim that all ascriptions of vulnerability actually refer to the same concept. The apparent tension between the ontological and the circumstantial conceptions, they say, is a "pseudo-problem which emerged due to a lack of conceptual analysis" (Martin, Tavaglione, and Hurst 2014, p. 62).
To make their claim, Martin et al. hold that vulnerability is a permanent intrinsic property of beings like us and a disposition to be harmed and/or wronged in certain circumstances. This explains why our shared vulnerability is not manifest for all of us, all of the time. In other words, the fact that vulnerability is a dispositional property explains why it can be a universally shared property with different likelihoods of manifestation for different people.
Martin et al.'s account of vulnerability is very promising. Appealing to the notion of disposition allows us to think of vulnerability across the circumstantial/ontological dichotomy. Indeed, we can begin the analysis by recognizing a universally shared ontological property-human embodiment-that causes our vulnerability (i.e., some likelihood of being harmed). In the second instance, we can show that circumstances increase vulnerability by making it more likely to manifest. We could say that, given human embodiment, humans beings are disposed to suffer bodily injuries-this is an intrinsic ontological property. However, a person's vulnerability to such injuries may increase in some contexts. For example, if the person is a woman, she is in many social contexts more likely to be the target of sexual violence. This contextual vulnerability to sexual harm becomes exponential if the person is in a relationship with a violent man. What this means is that vulnerability, even if it is an intrinsic ontological property, can be understood as a matter of degree and as context relative. A disposition may be more likely to manifest in some specific context, even if it has always been there. Thinking of vulnerability as an intrinsic disposition would therefore explain how vulnerability can be both ontological and circumstantial; this is Martin et al.'s contention, anyway.
In the following, we will see that Martin et al. make a crucial conceptual mistake in their analysis. If vulnerability is a disposition-and I agree that it is-then it cannot be an intrinsic property, contra the ontologists. Indeed, as Jennifer McKitrick shows, vulnerability is an extrinsic disposition (McKitrick 2003). Vulnerability, in other words, is a relational, we could say circumstantial, property-it is a function of an object's relationship with other objects. Evidently, if vulnerability is an extrinsic property it cannot be an intrinsic property, as Martin et al. suggest. This means they have failed to resolve the conflict between the ontological and the circumstantial conceptions. However, as we will see in section 5, this is not a problem.

Martin et al.'s most important assumption is the following:
Any definition of particularly vulnerable individuals in need of special protection needs to be embedded into a larger understanding of vulnerability: that is, vulnerability as a permanent intrinsic property of all beings with certain types of interest, but with different likelihoods of manifestation. (Martin, Tavaglione, and Hurst 2014, p. 62, my emphasis) According to this framework, circumstantial vulnerability is quite simply a case of ontological vulnerability that is, because of external circumstances, more likely to become manifest. While Martin et al. say ontological vulnerability is an essential and fundamental property of human beings, they argue both conceptions of vulnerability "depend on each other, since they refer to the very same concept with different likelihoods of manifestation" (p. 53, my emphasis). This reframing seems to make the ontological and circumstantial accounts compatible: vulnerability is a universally shared intrinsic property that is made manifest by external circumstances. If this is correct, Martin et al. successfully bridge the gap between the ontological approach and the circumstantial approach.
To make their claim, Martin et al. begin by distinguishing two turns of phrase. On the one hand, we can say that "x is vulnerable to y"; in this kind of sentence the word "vulnerable" can be easily replaced by the word "susceptible." On the other hand, we can say "x is vulnerable," tout court, a sentence which, per Martin et al., can be translated as "'x exemplifies the intrinsic property of vulnerability'" (Martin, Tavaglione, and Hurst 2014, p. 53). The foregoing expression is the focus of their analysis because their discussion aims at identifying the conditions that allow us to ascribe the (intrinsic) property of vulnerability to certain beings.
To count as vulnerable-as having the intrinsic property of being vulnerablean object must have welfare or agency interests, which "may be frustrated by the individuals themselves, external circumstances or other living beings" (p. 55). The reason we are vulnerable is that we have agency or welfare interests. 8 As such, all of us are vulnerable because we all have interests that are potentially frustrated and we have those interests intrinsically. However, even if we assume that we are intrinsically vulnerable, this view is compatible with the fact that some of us are more likely to have their interests frustrated. Indeed, vulnerability also has specific circumstances of manifestations that are more or less likely to occur.
According to Martin et al. this is because vulnerability tout court is a dispositional property. Hence, "while vulnerability is an intrinsic property, its conditions of manifestation are relational: a manifestation of vulnerability occurs due to some interaction of the vulnerable individual with the world" (p. 58). That is to say, this basic intrinsic vulnerability can manifest itself in various circumstances, in many ways, and with different likelihoods depending on the context.
To understand how this is possible, we have to see what it means to say that vulnerability is a dispositional property. The classic example is that of fragility. When we say that a glass is fragile, that it "exhibits the intrinsic property of fragility," we mean that it is disposed or prone to shatter if certain conditions obtain (e.g., when it is thrown against a hard surface). This does not mean that all fragile objects actually shatter; only in the right conditions will fragile objects break. Thus, dispositions are properties that are not always manifest even if the object remains disposed to x.
Dispositions admit of variations in the probability of manifestation. The glass's fragility is more likely to become manifest if it is laid on the side of a table, but the glass remains equally fragile when it is safely tucked in a padded box. We could actually say that taking the glass out of that box actually makes the glass vulnerable; we effectively create circumstances that makes it disposed to be damaged. We will see in section 4 why that matters, but let us leave this point aside for now.
If we think of vulnerability as a disposition, we can say that human beings with welfare or agency interests will be harmed (i.e., their vulnerability will become manifest) if certain conditions obtain (e.g., when they are "physically or mentally adversely affected" [Martin, Tavaglione, and Hurst 2014, p. 56]). Some people may be more likely to be harmed, but everyone is prone to be harmed under the right circumstances. According to Martin et al., the dispositionality of our intrinsic property of being vulnerable explains why it becomes manifest only under some conditions. If this argument works as is, the consequences are quite important. Indeed, once we show that vulnerability is a disposition, we have a neat explanation as to why it can be both permanent and variable. This would effectively resolve the conflict between the two main competing conceptions of vulnerability.
This would also help us explain why we have different moral intuitions about different manifestations of vulnerability. For example, saying that everyone is owed adequate consideration to their interests in virtue of their intrinsic vulnerability is compatible with saying that some people require "special consideration." Indeed, once we think of vulnerability as a disposition, it is possible to argue that, for those people who are identifiably more likely than others to see their interests unjustifiably frustrated-those who are especially "vulnerable"adequate consideration should involve "special protective measures." Therefore, the vulnerability of some-those identifiably more likely to be wrongfully harmed-can be more morally significant. This is so, not because they have more vulnerability, but because their basic vulnerability is more likely to become manifest. We can also make normatively significant distinctions on the basis of the circumstances of manifestation. For example, if vulnerability becomes manifest as the result of an accident, this may not give rise to moral obligations.
If we follow Martin et al., the fact that there are morally distinct manifestations of vulnerability does not mean that there are, in fact, different concepts at play. Vulnerability remains a permanent intrinsic property (universally shared) even if this property is manifest only when certain conditions obtain and even if these conditions are more likely to obtain for some people. The tension between the circumstantial and the ontological accounts of vulnerability is a false problem because the vulnerability faced by particularly vulnerable people is actually "embedded into a larger understanding of vulnerability" that is part of our intrinsic properties (p. 62). Unfortunately, vulnerability is not an intrinsic property.

THE PROBLEM
We just saw why Martin et al. conclude that the conflict between alternative views on vulnerability is solved if we think of vulnerability as an intrinsic dispositional property (p. 55). This claim, however, works only if we assume what metaphysicians call the "Intrinsic Dispositions Thesis" (IDT). According to the IDT, dispositions are "intrinsic properties of their bearers" (Armstrong 1973;Lewis 1997;Molnar 1999, p. 3). If we accept this thesis, there is absolutely no problem in saying that vulnerability is both an intrinsic property and a disposition.
The problem is that Jennifer McKitrick-whom Martin et al. actually cite to make their claim-explicitly refutes the IDT. In fact, she claims quite explicitly that vulnerability is an extrinsic disposition. If Martin et al. follow McKitrick to argue that vulnerability is a disposition, it would seem that they make the rather strange claim that vulnerability is both an extrinsic and an intrinsic property. This sounds absurd. As McKitrick reminds us, "every property is either intrinsic or extrinsic" (McKitrick 2003, p. 158). How can we make sense of Martin et al.'s claim to have resolved the conflict in these circumstances?
First, let us take a closer look at the notion of disposition. Something has a disposition, simpliciter, when it is "prone to act in certain ways in certain circumstances" (p. 156). As I said, most philosophers adopt the IDT and argue that dispositions are "intrinsic properties of their bearers" (McKitrick 2003, p. 155;citing Molnar 1999, p. 3). To say that a disposition is intrinsic means that, the laws of nature remaining fixed, perfect duplicates would necessarily have the exactly same dispositions. Indeed, intrinsic properties of an object belong to this object necessarily and do not depend on external, extrinsic, relational, or circumstantial properties to exist. A glass is intrinsically fragile for it has fragility as one of its intrinsic properties. By the same token, Martin et al. argue human beings are intrinsically disposed to be harmed for they have vulnerability as an intrinsic disposition.
The IDT is a strong thesis. It holds that, no matter what, dispositions are intrinsic properties of their bearers, "regardless of what is going on outside of [them]" (McKitrick, 2003, p. 158). McKitrick argues that the IDT is wrong. At least some dispositions are extrinsic and, as I said, Martin et al. are in trouble because vulnerability is one of them-we will see why in the following.

Extrinsic Dispositions
McKitrick's strategy to challenge the IDT is to argue that if "perfect duplicates can differ with respect to having certain dispositions keeping the laws of nature fixed," then at least some dispositions are extrinsic (p. 155). In other words, if it is possible to show that two ontologically identical objects have different dispositions, then at least some dispositions are extrinsic. To relate this point to the concept of vulnerability, McKitrick's thesis entails that changes to my extrinsic properties would be sufficient to make me vulnerable, and this without any changes to my intrinsic properties. Changes to my extrinsic properties could make me acquire the disposition to be harmed or wronged in certain circumstances.
Let us begin with a neutral example of an extrinsic disposition (p. 163). Recognizability is a disposition. People who are recognizable are prone to be recognized. Bill Clinton is recognizable. However, his recognizability is not a function of his intrinsic properties (e.g., his physical appearance). It is a function of his living in a society where Bill Clinton is famous-which is evidently an extrinsic and relational property of Bill Clinton. Bill Clinton would not be recognizable, as such, in a world where he was not famous. In other words, if circumstances were different, Bill Clinton might lose his disposition to be recognized (p.173).
Vulnerability is extrinsic for the same reasons. McKitrick holds that something is "vulnerable if it is disposed to suffer as a result of an attack" (McKitrick 2003, p. 161). This is arguably narrower than being disposed to be harmed or wronged in certain circumstances, but the latter certainly includes the former. McKitrick asks us to imagine a city that is made invulnerable to attacks by a "Star Warslike defence system" operated and maintained outside the city limits (p. 161). The system is thus part of the city's extrinsic properties. If this system were to be turned off, the city would become vulnerable without any changes to its intrinsic properties. Therefore, vulnerability is an extrinsic disposition. Similarly, walking alone or with bodyguards changes one's vulnerability without modifying one's intrinsic properties. I am not vulnerable to suffer from attacks if I have bodyguards; I am if am alone. I do not have the space to fully go over the debate over dispositions and the IDT, but it seems McKitrick's example is a clear counterexample to an extreme view of the IDT. Hence, I will assume that there are extrinsic dispositions (i.e., that IDT is false). If vulnerability is an extrinsic disposition of human beings, then it cannot be a permanent intrinsic property of human beings, as Martin et al. argue. Given the importance of dispositionality for their thesis, it would seem that Martin et al. did not successfully resolve the conflict between the two basic views of vulnerability. So, to successfully resolve the conflict, they can do one of at least three things.
The first way out of this is to argue that McKitrick's example of vulnerability is not one of vulnerability tout court-one in which vulnerability can be expressed  (Martin, Tavaglione, and Hurst 2014, p. 55). But without an explanation as to why McKitrick's example is not one of vulnerability tout court, this argument would be begging the question. Indeed, it is not clear that one is warranted to maintain a strong distinction between the locutions "x is vulnerable" tout court and "x is vulnerable to y" on the basis that the first sentence denotes an intrinsic property and the other one does not. If vulnerability is indeed an extrinsic disposition, saying "x is vulnerable" would simply not translate into "x exemplifies the intrinsic property of vulnerability." Another way out would be for Martin et al. to stress that they sought to talk about only the particular kind of vulnerability that is relevant to health care. This vulnerability, they could claim, is indeed an intrinsic disposition-after all, even though they cite McKitrick, who clearly states that vulnerability is an extrinsic disposition, Martin et al. merely say that vulnerability is a dispositional property and postpone a full discussion of the metaphysical difficulties related to dispositions (p. 55, p. 68 n. 5). In other words, Martin et al. could say that McKitrick's examples-the city (un)protected by a defense system and the person walking with(out) bodyguards-are not of the right kind of vulnerability. This would not be incompatible with McKitrick's thesis since she is merely saying that at least some dispositions are extrinsic. Maybe the kind of vulnerability relevant to health care is distinct from the kind of vulnerability discussed by McKitrick.
Let us assume that vulnerability in health care is an intrinsic disposition. This means that, the laws of nature remaining fixed, perfect duplicates would necessarily retain their vulnerability. This strategy sounds promising if we hold that the reason we are vulnerable is that we have the permanent intrinsic property of having welfare and agency interests. Indeed, it seems that in most cases this property would be shared by perfect duplicates; if this property is the reason we can say human beings are vulnerable, it should be the case that all objects who have these interests are vulnerable in this basic sense, even if this vulnerability does not or even cannot become manifest.
Let us further assume that having interests which can be frustrated is an intrinsic property that does not come and go with changes to one's extrinsic properties. Even in a scenario where a machine analogous to the Star Wars defense system of McKitrick's example would fulfill all my interests, I would remain vulnerable because vulnerability is an intrinsic property related to my having those interests.
The problem with this argument is that there are clearly cases in which perfect duplicates would differ in terms of having vulnerability without any changes to their interests qua intrinsic properties. Indeed, let us assume that human beings, especially in the context of health care, have welfare and agency interests to be respectfully treated with regards to their sociolinguistic background. For instance, they have a particular interest in communicating in their native language. Having this interest is the reason many people are vulnerable in many health-care contexts; linguistic minorities are more likely to see this interest frustrated and should be granted special attention. What happens in health-care contexts that involve only one linguistic community? It seems that in such contexts, people would not be vulnerable with regards to that particular interest, even if we assume they would have it qua human beings. While people would still have an interest in being able to communicate in their native language, it would be a stretch to say that members of a monolingual community are vulnerable to being spoken to in some other language. If, as Martin et al. suggest, vulnerability refers to some likelihood of being mistreated, I doubt a likelihood close to zero would count as vulnerability in the relevant sense. This, I think, clearly shows how two perfect duplicates could possess the same welfare and agency interests and still differ with regards to their capability of seeing them frustrated. While changes to my community may indeed cause me to become vulnerable to being spoken to in another language, this disposition arises only out of changes to my extrinsic properties. If vulnerability is a disposition, it is extrinsic.
Martin et al. could make a final stand and argue, following David Armstrong, that having vulnerability, qua disposition, "depends wholly on the intrinsic properties of [a person 9 ], and does not depend on any extrinsic properties of the [person]" (McKitrick 2003, p. 172, citing Armstrong 1973. 10 According to this claim, my propensity to be harmed depends wholly on my intrinsic properties, even if the conditions of manifestation of vulnerability are causally related to some of my extrinsic properties. That is to say, even in the absence of circumstances and causal chains that would make my vulnerability manifest, I would retain vulnerability as a disposition. This is because the only extrinsic properties that are relevant to my being harmed-to my vulnerability becoming manifestare those "properties that are instantiated in the circumstances of manifestation" (i.e., the initiating cause and the circumstances) (McKitrick 2003, p. 172, my emphasis). These properties clearly do not belong to me intrinsically. Thus, they do not affect my having the property of being vulnerable. Therefore, my vulnerability is intrinsic.
However, as McKitrick argues, even if this particular argument works for dispositions like fragility, it is clearly not the case for all dispositions. For example, Bill Clinton's recognizability is a disposition that clearly depends on more than the properties instantiated in the circumstances of its manifestation (e.g., being outside, not wearing a disguise, etc.). Living in a world where Bill Clinton is famous is an extrinsic property that is clearly relevant to his being recognized and to his being recognizable (McKitrick, 2003, p. 173). Extrinsic properties are thus necessary both for the existence of the disposition and for its actual manifestation. Similarly, living in a racist and sexist society is an extrinsic property that allows us to infer that the interests of women and racial minorities are likely to be frustrated in this society. Living in a racist and sexist society does not depend on the properties instantiated when someone is suffering racist or sexist behaviour. At least some of their interests could be frustrated in a nonracist and nonsexist society, but the fact would remain that some of the reason they are currently vulnerable (having the welfare and agency interests of not being forced into gendered forms of life; having the welfare and agency interests of not being treated differently on the basis of race) would simply not exist if their extrinsic properties were different. Thus, even if some dispositions are indeed intrinsic, vulnerability is not one of them. Vulnerability, qua disposition, clearly varies with context, and extrinsic properties are relevant to its existence.
Furthermore, even if we accept Martin et al.'s claim that the reason we are vulnerable is that we have certain permanent intrinsic properties (i.e., we have welfare and agency interests) and even if we accept that vulnerability is "wholly dependent" on these intrinsic properties, that does not show that vulnerability itself is an intrinsic property and that does not show that it is not an extrinsic disposition as defined by McKitrick. Of course, we do have intrinsic properties, like being embodied, carbon-based, or aerobic, or having agency, that may underlie or ground our propensity to be harmed. We might even say that our vulnerability supervenes on these intrinsic properties. But that would not mean these properties are the same as our disposition to be vulnerable, just like being made of glass is an intrinsic property of a glass that is distinct from its fragility or from its vulnerability-when it is standing at the edge of a table.
Emphasis is required here to make the point as clear as possible. A glass is fragile in virtue of its molecular make-up (its intrinsic properties). It will be fragile in all circumstances, in all possible worlds, no matter what. However, insofar as we can say that a fragile glass G on the edge of a table is more likely to shatter than it would be if it were tucked inside a padded box, it seems possible to say that G is vulnerable when it is laid on the edge of a table-when it has the extrinsic property of being in such position. Of course, if G were made of plastic, we would not say that it is vulnerable, in either situation. This shows that there is indeed an intimate relationship between an object's vulnerability and its intrinsic properties-human beings would not be vulnerable if they had a different ontological nature. However, this relationship is not one of identity-my fragility, frailty, or mortality is not the same as my vulnerability. Some might want to say that vulnerability supervenes on our intrinsic properties or that these properties ground our vulnerability (Epstein 2015). I leave a full discussion on the proper terminology to express this relationship to another time. What is important to note at this time is that, while our vulnerability might only be possible because we are constituted as we are, as a disposition, our vulnerability also depends on our extrinsic and circumstantial properties.
A note is in order concerning this last sentence. In a previous version of this article, I wrote "vulnerability wholly depends on our extrinsic and circumstantial properties." 11 This, as many have urged me to mention, suggests that the possibility of sudden death syndrome, brain aneurism, or congenital health conditions-all of which seem to depend wholly on intrinsic properties-does not amount to vulnerability. This goes against the way we use the term "vulnerability." I see the force of this argument. There is definitely some sense in which the term "vulnerability" denotes something that is wholly dependent on our intrinsic properties. However, as I argue in the next section, if we want a concept of vulnerability that captures something normatively significant, we must focus on a concept of vulnerability that refers to our extrinsic disposition to be harmed or damaged. Essentially, my claim is that the kind of vulnerability that matters for normative political and moral theorizing is wholly extrinsic even if, as I have just argued, this property supervenes on some of our intrinsic properties.
To reiterate, I agree with Martin et al.'s definition of vulnerability as a property of beings with interests and, while I salute their attention to the circumstances in which vulnerability becomes manifest, I do not think they have successfully shown that "the controversy concerning the question of whether vulnerability is a property of all living beings or just of some" is a pseudo-problem (Martin, Tavaglione, and Hurst 2014, p. 62), because they did not show how vulnerability is both intrinsic (and ontological) and extrinsic (and circumstantial).

WHAT IF IT WERE ALL CIRCUMSTANTIAL?
So far, the focus of my analysis has been to circumscribe the concept of vulnerability. The conflict between the two fundamental conceptions, "ontological" and "circumstantial" vulnerability, would have us believe that vulnerability is either always and universally present or only sometimes present. I think the best way to explain this phenomenon is to think of vulnerability as a dispositional property of beings who can be harmed or damaged. I have argued, against Martin et al., that this disposition is extrinsic. In this section I want to briefly discuss the normative issues related to this conceptual approach to vulnerability.
As I stated at the end of section 4, I must acknowledge that some instances of vulnerability seem to depend wholly on a person's intrinsic properties (e.g., that person's vulnerability to brain aneurism or to sudden death syndrome). This is true even if this vulnerability is described as a dispositional property. Indeed, human bodies seem to be intrinsically disposed to have brain aneurisms when some specific circumstances obtain. However, these instances of vulnerabilityinstances that are impossible to predict or prevent-are not particularly interesting from a normative political or moral standpoint. I am capable of suffering from a brain aneurism just as I am capable of suffering from a lightning strike. In both instances, it seems that the word "vulnerable" does not convey particularly important moral or political information. It just says that my body will be damaged if something we cannot predict or prevent happens. It is not absurd, but it is not very interesting either. In the introduction, I argued that our concept of vulnerability should be compatible with the way we use it in normative theorizing. Vulnerability, I said, is used to flag important issues. Because the mere possibility of certain harms or damages is not particularly interesting for normative theorizing, I leave aside this aspect of our everyday use of the word "vulnerability." That does not mean, however, that all intrinsic properties related to our vulnerability are uninteresting from the standpoint of normative theorizing. For many people, our shared finitude, embodied fragility, and mortality are morally significant. Thus, a concept of vulnerability that would be circumstantial all the way down would also miss something important. Along this line of thought, Catriona Mackenzie argues that Martin et al.'s definition of vulnerability does not resolve the conceptual conflict between ontological and circumstantial conceptions of vulnerability because their account is merely another version of the latter (Mackenzie 2016, p. 86). She notes that Martin et al.'s view collapses into the circumstantial conception because it "does not take seriously the implications of the view that vulnerability is an ontological condition of our finite, embodied humanity" (p. 86). As Mackenzie points out, in at least some sense, our vulnerability is "an ontological condition of our embodied humanity" (p. 87). Our finite and fragile bodies are all susceptible to suffering, and these properties are what, according to the "ontological view, and in our everyday use of the concept" makes us vulnerable (p. 88).
I think Mackenzie is right. Properly conceived, Martin et al.'s dispositional account of vulnerability does collapse into a circumstantial/relational account. They fail to resolve the conflict. But frankly, I do not see Martin et al.'s failure in that regard as a problem. Indeed, as I argue in the following, with my amended dispositional account, we may not need the ontological account after all. My claim is that we can capture everything normatively significant about the ontological view with an extrinsic and dispositional account of vulnerability. The concept of vulnerability relevant for normative political and moral theorizing may be circumstantial all the way down, and, as I argue, this is probably an advantage, especially in normative political and moral theorizing.
First, let us deal with the moral importance of some of our intrinsic properties. As I argued in section 4, even if human vulnerability is intimately related to properties like fragility, embodiment, etc.-properties that are both intrinsic and universally shared-this relationship does not have to be one of identity. We can say that our vulnerability supervenes or is grounded on these intrinsic properties to explain why we would not be vulnerable without these properties. This does not mean, however, that vulnerability itself is intrinsic. Vulnerability simply arises from a particular arrangement of the world in which an object's integrity, needs, or interests can be frustrated because some of its intrinsic properties are likely to be negatively affected.
Authors like Martha Fineman and Catriona Mackenzie hang onto the ontological view because they think it morally important to focus on vulnerability as a "shared" property of humanity (Fineman 2012(Fineman , 2013Mackenzie 2016). Indeed, it seems quite intuitive to say that all human beings are, to some extent, vulnerable. It even seems natural to say that this is a universal feature of the human essence. We are all vulnerable to elements, agents, etc., because of some of our intrinsic ontological properties. This, as Mackenzie argues, "help[s] to flesh out the intuition that vulnerability is a morally salient feature of our humanity and hence can help to explicate why the judgment that someone is vulnerable carries normative force and seems to demand from us some kind of response" (Mackenzie 2016, p. 92). However, we can certainly say vulnerability is morally salient even when it is wholly related to circumstantial and relational properties that have nothing to do with our intrinsic properties. A women's vulnerability to sexism has nothing to do with her "womanhood," whatever that is-of course, some of her intrinsic properties may be negatively affected, but again this does not mean that her vulnerability is identical to the properties that are affected. Moreover, our extrinsic properties affect our vulnerability in more than the circumstances of its manifestation. We all are vulnerable because we all are in relationships with external objects. What this means is that the intrinsic/extrinsic dichotomy cuts across the universal/particular dichotomy.
I am perfectly comfortable saying that all human beings are stuck in circumstances that make them vulnerable. Circumstantial vulnerability is indeed universal. But, importantly, this does not mean this property is intrinsic, and this is precisely what is so powerful about the dispositional account. What the dispositional account of vulnerability highlights is that something can have vulnerability as a property even if this is not always manifest-even if at a particular moment that thing is not likely to be harmed or damaged. Moreover, as we saw with the example of the vulnerability to being unable to communicate in one's native language, the extrinsic dispositional account explains why particular circumstances can bring about vulnerability. Thus, vulnerability does not need to be an ontological or an essential property to be universal in the way proponents of the ontological view think of this concept. Even if vulnerability is indeed an extrinsic disposition, it could certainly be argued that all humans are in some kind of relation with the external world (or, in other words, all humans have extrinsic properties) that make them vulnerable. Therefore, we can think of vulnerability as a universal property without changing the metaphysical status of this disposition; this property comes and goes with changes to an object's extrinsic properties even if, as it were, all relevant objects have the kinds of extrinsic properties that make them vulnerable. Now let us move on to the advantages of an extrinsic account of vulnerability. I suggested in the introduction that objects like groups, institutions, and ecosystems can be vulnerable, at least metaphorically, if we treat vulnerability as the likelihood of seeing one's interests, needs, or integrity frustrated. These objects may not have the kind interests or needs crucial to human vulnerability, but they have certain properties that can, in some circumstances, be under threat. I think we have good normative reasons to apply the concept of vulnerability to these objects. Indeed, in applying the concept of vulnerability to an object, we can, it seems, flag important moral and political concerns. We should monitor pollution because ecosystems are vulnerable to pollutants; we should strike a balance of power among branches of government because institutions are susceptible to hostile take-overs; national minorities are vulnerable to cultural assimilation. In situations where these entities are deemed morally valuable, saying they are vulnerable provides important moral information. However, if we are going to allow for this kind of normative statement with regards to objects like institutions, ecosystems, or groups, we simply cannot treat vulnerability as conceptually related to the kinds of intrinsic properties that make us vulnerable. Indeed, those objects do not and cannot have vulnerability in virtue of their embodiment or their need for care. They are not embodied in the same way, they do not hold close relationships with care-givers and dependents, etc. Yet, it is clear that those objects appear to be normatively vulnerable. Indeed, it is possible for them to face particular circumstances that make them more or less likely to be damaged. Again, it may be true that this vulnerability is somehow related to these objects' intrinsic properties, but it does not make a difference on a metaphysical level. Nor should it make a difference from a moral perspective. Thus, to account for their vulnerability, we cannot reduce vulnerability to the expression of our ontological properties in certain circumstances.
I could also run the argument in another way. Since groups do appear to be normatively vulnerable in at least some circumstances, and since an intrinsic account of vulnerability is incompatible with the notion of group vulnerability-because groups do not have the kind of universal ontological properties that are supposed to ground conceptions of intrinsic vulnerability like ontological vulnerability and vulnerability qua dependency-an intrinsic account of vulnerability cannot be right.
When Black Lives Matter activists denounce the fact that Black Americans are vulnerable to police violence, they, of course, implicitly refer to their embodied fragility. Indeed, their claim makes sense only if we assume that Black Americans, as human beings, are ontologically constituted such that they can suffer physical harm. However, the force of their moral claim comes from the injustice of the circumstances Black Americans face. Indeed, members of a socially identifiable group should not be made or kept vulnerable in a society in which everyone is supposed to be treated with equal respect. This situation comes with a vivid sentiment of vulnerability. We could say that all people can see themselves as vulnerable, intrinsically, if they really think about it. But I think it is clear that this sentiment is not the same as knowing oneself to be vulnerable because of unjust circumstances. I would also wager that the former sentiment of vulnerability is morally worse. When members of a minoritized group realize that they are vulnerable because of their group membership, they are likely to see this as an injustice. It seems that it would be even worse if these people were told that, in fact, when you really think about it everybody is vulnerable.
Note that we did not have to treat vulnerability as an ontological property to highlight the moral urgency of the situation. An extrinsic account of vulnerability was entirely sufficient. More importantly, this approach has the advantage of being protected against the criticisms according to which attributing vulnerability to a group or a person applies a stigmatizing label to groups and individuals (Brown 2011;Brown 2012;Luna 2009). This worry is especially strong when we think of the vulnerability of groups. As Luna shows, if you think of vulnerability as a permanent intrinsic property, then saying that a particular group or a particular subpopulation is vulnerable seems to suggest, first, that the group is homogeneous and that all its members are equally vulnerable and, second, that the group has some intrinsic properties that make its members equally vulnerable (Luna, 2009, pp. 123-124). Both assumptions are clearly false in many situations. It is certainly correct to say that Black Americans are vulner- able, but this does not mean that all Black Americans are equally vulnerable in all contexts and this certainly does not mean that their vulnerability is attributable to some of their intrinsic properties, qua group.
Paying attention to the particular circumstances of manifestation also allows us to prevent what Luna calls the "'naturalization'" of vulnerability, which leads to "the thinking that it is normal or natural to be vulnerable, and that all research subjects are in one way or the other vulnerable" (p. 128). If vulnerability is natural, ontological, and universally shared, it might be more difficult to justify special attention to some people in virtue of their vulnerability. Essentially, if everyone is vulnerable, nobody is vulnerable in a meaningful sense. This does not mean that attempts to prevent or to respond to extrinsic vulnerability will not have these adverse effects, but these consequences will not be attributable to the concept itself.

CONCLUSION
I argued in this article that an amended version of Martin et al.'s dispositional account of vulnerability can neatly solve the conceptual puzzle with regards to ontological and circumstantial vulnerability without implying the problematic view that vulnerability is both an extrinsic and an intrinsic property. I have also shown that my extrinsic dispositional account can capture most of what we think matters with regard to vulnerability. Contra Mackenzie, Rogers, and Dodds, I think that a typology does not clearly solve the conceptual puzzle we started with. I also hold that vulnerability, at least the kind of vulnerability relevant to normative moral and political theorizing, is circumstantial all the way down. Our vulnerability to aneurisms is very real, but it refers to the mere possibility of suffering harms or damages, a metaphysical possibility that is, I argue, different from the likelihood of suffering harms or damages to which we usually refer with the word "vulnerability." We also saw that the fact that an extrinsic disposition is underpinned by intrinsic properties does not mean the disposition is intrinsic. An object's vulnerability is distinct from its fragility or its mortality. Moreover, the fact that all human beings are vulnerable does not mean our vulnerability is an intrinsic ontological property. I argued this means we do not need an ontological account of vulnerability to make the kind normative judgments and distinctions Mackenzie and others use the ontological view for. In particular, we do not need an intrinsic ontological account of vulnerability to say that all human beings are vulnerable (i.e., that vulnerability is universally shared) and that vulnerability is an embodied property (i.e., that vulnerability is intimately related to some of our intrinsic properties).
I have also shown that the ontological account of vulnerability prevents us from consistently applying vulnerability to other kinds of objects. Not only do these objects (groups, institutions, and ecosystems) lack the properties that make us vulnerable, importing the intrinsic ontological view in our statements about their vulnerability comes with problematic baggage. In particular, an ontological approach to vulnerability seems to suggest that women or racial minorities, both of which can reasonably be described with the word "vulnerable", are essentially and intrinsically different from other human beings. This, of course, is a problem because there is nothing intrinsic about a person's vulnerability to racism or sexism. My extrinsic account of vulnerability avoids this problem entirely.
Thinking of vulnerability as an extrinsic disposition and, therefore, as a relational and circumstantial property, albeit grounded on some of our intrinsic properties, is more metaphysically sound; it is compatible with our differing moral judgements with regards to different manifestations of vulnerability and it applies to more than just human beings. Given these advantages and given that vulnerability as an extrinsic disposition can capture important features of the ontological view, I conclude we have no reason to cling to the ontological conception of vulnerability. With their dispositional account, Martin et al. began to pave the way for discussions about vulnerability that are metaphysically robust and normatively compelling. we call a "wrong." As will be clear in section 5, I think there are reasons to adopt my account that pertain to the normative dimensions of vulnerability, but this article is mainly conceptual. 2 In an earlier version of this article, I defined vulnerability as "the disposition to be harmed (sometimes wrongly) in certain circumstances." I also argued that vulnerability could be either needs-or interests-based. This, as the reviewers aptly noted, made my application of the concept of vulnerability to "groups," "institutions," and "ecosystems" rather dubious. Understandably, they weren't convinced by my rather vague suggestion that "institutions" and "ecosystems" may have "interests" and "needs," in some sense of the terms. I still think it important to note we use the word "vulnerability" when we talk about groups, institutions, and ecosystems. This use of language provides important moral and political information. Even then, as a reviewer pointed out, this may sound as though I am using "misleading anthropomorphizing metaphors." By adding "damaged" in my definition and leaving open the possibility of an interests-based, needs-based, and integrity-based understanding of vulnerability, I think I respond at least to their conceptual worries. On the moral front, however, I will dig in my heels and maintain that groups, institutions, and ecosystems can be vulnerable in a morally and political significant sense. Because I do not make vulnerability a function of uniquely human properties, conceptually, I think I am warranted in making this normative move. I wish to thank both reviewers for pushing me to clarify this point. 3 McKitrick defines vulnerability as a disposition to "suffer harm as a result of an attack" (McKitrick 2003, p. 161). Arguably, my own definition is broader than that, but this does not change much for the nature of that property. 4 We will see that this vulnerability can be grounded in an object's intrinsic properties, but that this does not mean vulnerability itself is the same as the intrinsic properties it is grounded on. For a more complete explanation of the "grounding relationship," see Epstein 2015, chap. 6. 5 Some authors offer further subdivisions. Samia Hurst talks of "consent-based," "harm-based," and "comprehensive" definitions of vulnerability; Rogers, Mackenzie, and Dodds talk about "inherent" and "situational" sources of vulnerability and add that both these basic aspects of vulnerability can be "dispositional" or "occurrent" and that in some cases situational vulnerabilities are "pathogenic" (Hurst 2008 p. 192;Mackenzie, Rogers, and Dodds 2013, pp. 24-25). These qualifications highlight important features of the concept of vulnerability, but for my purposes I will concentrate on the following basic distinction between "universal" and "circumstantial." 6 It should be noted that this debate pertains to human vulnerability. As far as I know, nobody has systematically discussed the application of the concept of vulnerability to other kinds of objects. I will say why this might be a problem in section 5. 7 Butler talks about "precarity," but she makes a similar distinction between "universal" precarity and "special" precarity (Butler 2009, p. xx). See also Gilson 2014, chap. 2. 8 I will not address the debate between needs-based accounts of vulnerability and interestsbased accounts. See Mackenzie 2016.

V O L U M E 1 2 N U M É R O 2 -3
A U T O M N E / F A L L 2 0 1 7 9 I'm paraphrasing; the original says "glass" because the disposition under study is fragility. 10 McKitrick notes that Armstrong is probably not making a distinction between "depending wholly on intrinsic properties" and "being an intrinsic property." However, she also says that Armstrong clearly supports the view that dispositions are intrinsic, so for my purposes I am not going to make the distinction either. 11 I thank reviewers for catching this issue. I also want to express my gratitude to Daniel Engster, who provided me with the brain aneurism example, and to Éliot Litalien and Natalie Stoljar for very helpful discussions on the normative dimensions of this problem.