Community-Level Vulnerabilities and Political Field Experiments

Most research ethics literature on vulnerability focuses on the vulnerability of individuals and populations defined by the potential vulnerability of their members (such as adults with intellectual disabilities or prisoners). However, research involving human participants does not always take the individual as the unit of analysis: political experiments may apply an intervention to a community as a whole. This paper argues that community-level vulnerability is not reducible to the sum of the vulnerabilities of community members, and that there is thus a need to consider vulnerability at the community level of analysis when analyzing the ethical implications of political field experiments. I first review ethical literature on community intervention research and the emerging scholarship on the ethics of political field experiments. I then highlight key accounts of the Abstract La plupart des documents relatifs à l’éthique de la recherche sur la vulnérabilité s’intéressent à la vulnérabilité des individus et des populations définies par la vulnérabilité potentielle de leurs membres (comme les adultes souffrant de déficiences intellectuelles ou les prisonniers). Cependant, la recherche impliquant des sujets humains ne prend pas toujours l’individu comme unité d’analyse : les expériences politiques de terrain peuvent appliquer une intervention à une communauté dans son ensemble. Cet article soutient que la vulnérabilité au niveau de la communauté n’est pas réductible à la somme des vulnérabilités des membres d’une communauté, et qu’il est donc nécessaire de considérer la vulnérabilité au niveau de l’analyse de la communauté lors de l’analyse des implications éthiques des expériences politiques sur le terrain. Je passe d’abord en revue la littérature éthique sur la recherche d’intervention communautaire et la recherche émergente sur l’éthique des expériences de terrain politiques. Je mets ensuite en évidence les principaux comptes rendus du concept de vulnérabilité au niveau individuel. En m’appuyant sur le concept de « torts collectifs » de Whitfield, je soutiens que les communautés peuvent être affectées négativement de manière distincte des torts causés aux membres individuels de la communauté, et que la variation de la susceptibilité à de tels torts au niveau de la communauté est largement conforme aux conceptualisations existantes de la vulnérabilité. Je suggère des questions que les chercheurs peuvent prendre en considération lors de la conception d’expériences politiques sur le terrain afin de s’assurer que les vulnérabilités au niveau communautaire sont prises en compte. Most research ethics literature on vulnerability focuses on the vulnerability of individuals and populations defined by the potential vulnerability of their members (such as adults with intellectual disabilities or prisoners). However, research involving human participants does not always take the individual as the unit of analysis: political experiments may apply an intervention to a community as a whole. This paper argues that community-level vulnerability is not reducible to the sum of the vulnerabilities of community members, and that there is thus a need to consider vulnerability at the community level of analysis when analyzing the ethical implications of political field experiments. I first review ethical literature on community intervention research and the emerging scholarship on the ethics of political field experiments. I then highlight key accounts of the concept of vulnerability at an individual level. Drawing on Whitfield’s concept of “collective wrongs,” I argue that communities can be negatively affected in ways that are distinct from harms to individual community members, and that variation in susceptibility to such wrongs at the community level is largely consistent with existing conceptualizations of vulnerability. I suggest questions that researchers should consider when designing political field experiments to ensure that community-level vulnerabilities are taken into account.


Article abstract
Most research ethics literature on vulnerability focuses on the vulnerability of individuals and populations defined by the potential vulnerability of their members (such as adults with intellectual disabilities or prisoners). However, research involving human participants does not always take the individual as the unit of analysis: political experiments may apply an intervention to a community as a whole. This paper argues that community-level vulnerability is not reducible to the sum of the vulnerabilities of community members, and that there is thus a need to consider vulnerability at the community level of analysis when analyzing the ethical implications of political field experiments. I first review ethical literature on community intervention research and the emerging scholarship on the ethics of political field experiments. I then highlight key accounts of the concept of vulnerability at an individual level. Drawing on Whitfield's concept of "collective wrongs," I argue that communities can be negatively affected in ways that are distinct from harms to individual community members, and that variation in susceptibility to such wrongs at the community level is largely consistent with existing conceptualizations of vulnerability. I suggest questions that researchers should consider when designing political field experiments to ensure that community-level vulnerabilities are taken into account. impliquant des sujets humains ne prend pas toujours l'individu comme unité d'analyse : les expériences politiques de terrain peuvent appliquer une intervention à une communauté dans son ensemble. Cet article soutient que la vulnérabilité au niveau de la communauté n'est pas réductible à la somme des vulnérabilités des membres d'une communauté, et qu'il est donc nécessaire de considérer la vulnérabilité au niveau de l'analyse de la communauté lors de l'analyse des implications éthiques des expériences politiques sur le terrain. Je passe d'abord en revue la littérature éthique sur la recherche d'intervention communautaire et la recherche émergente sur l'éthique des expériences de terrain politiques. Je mets ensuite en évidence les principaux comptes rendus du concept de vulnérabilité au niveau individuel. En m'appuyant sur le concept de « torts collectifs » de Whitfield, je soutiens que les communautés peuvent être affectées négativement de manière distincte des torts causés aux membres individuels de la communauté, et que la variation de la susceptibilité à de tels torts au niveau de la communauté est largement conforme aux conceptualisations existantes de la vulnérabilité. Je suggère des questions que les chercheurs peuvent prendre en considération lors de la conception d'expériences politiques sur le terrain afin de s'assurer que les vulnérabilités au niveau communautaire sont prises en compte.
Most research ethics literature on vulnerability focuses on the vulnerability of individuals and populations defined by the potential vulnerability of their members (such as adults with intellectual disabilities or prisoners). However, research involving human participants does not always take the individual as the unit of analysis: political experiments may apply an intervention to a community as a whole. This paper argues that community-level vulnerability is not reducible to the sum of the vulnerabilities of community members, and that there is thus a need to consider vulnerability at the community level of analysis when analyzing the ethical implications of political field experiments. I first review ethical literature on community intervention research and the emerging scholarship on the ethics of political field experiments. I then highlight key accounts of the concept of vulnerability at an individual level. Drawing on Whitfield's concept of "collective wrongs," I argue that communities can be negatively affected in ways that are distinct from harms to individual community members, and that variation in susceptibility to such wrongs at the community level is largely consistent with existing conceptualizations of vulnerability. I suggest questions that researchers should consider when designing political field experiments to ensure that communitylevel vulnerabilities are taken into account.

INTRODUCTION
Political field experiments test potential policy initiatives by applying experimental methods in a naturalistic and often largescale context (1,2). Field experiments are increasing in popularity as a tool in the ascendant "evidence-based policymaking" paradigm (3) and in political science scholarship (2,4). These experiments may involve interventions at a community level. For example, a political field experiment may involve modifying the structure of unemployment benefits in an effort to increase the employment rate or changing the built environment in a neighbourhood to promote the use of active transportation. Communities may be randomly or purposively assigned an intervention and compared to "control" communities that receive no intervention or "treatment as usual." Along with community-level interventions, these experiments may have communitylevel effects -benefits and harms, both intended and unintended -that can affect the community as an entity.
Adopting a community level of analysis in research raises distinct ethical questions. This paper focuses on the concept of vulnerability, asking how vulnerability may be understood at a community level in political field experiments. I argue that adding a community level of analysis to discussions of vulnerability in political field experimentation is essential because community vulnerabilities are not reducible to the vulnerabilities of a community's individual members. As such, failing to consider the vulnerability of communities may result in unanticipated adverse effects. While community-level intervention research is carried out in many disciplines (including community psychology, public health, and others) and methodological traditions (including community-based participatory research), I aim to contribute specifically to the emerging literature on the ethics of political field experimentation. As such, examples will be drawn primarily from the ethical and empirical literature in this area.
First, I highlight interdisciplinary scholarship on the ethics of community-level research. Second, I review ethical literature relating to political field experiments. Third, I provide a brief overview of definitional debates around individual-level vulnerability in research ethics. Fourth, I argue that communities can be negatively affected in ways that are not accounted for at an individual level of analysis. I will provide examples of domains in which community-level wrongs may occur and argue that susceptibility to incurring wrongs during political field experimentation can vary across communities and research situations. Fifth, I propose that this variation in susceptibility to wrongs is congruent with existing definitions of vulnerability, and therefore necessitates considering vulnerability at the community level when conducting political field experiments. Finally, I suggest considerations for researchers conducting political field experiments and call for further investigation and conceptualization of this issue.

ETHICS IN COMMUNITY-LEVEL RESEARCH
Political field experimentation is only one example of research conducted at a community level. Ethical questions in communitylevel research require attention to the concept of a community itself, and to the complexity and dynamism of community relations. I draw on interdisciplinary literature to highlight some of these issues below.

Defining communities
The question of what constitutes a community is critical in community-level research. Writing in the field of community psychology, Krause and Montenegro propose seven characteristics of community: social relationships, shared culture, shared identities, a sense of community, collective action and motivation, and spatial and temporal dimensions (including virtual or other spaces) (5). In an article on community engagement in biomedical research, Weijer and Emanuel suggest ten characteristics of communities, while acknowledging that not all communities will have all characteristics. Excluding healthspecific domains, their list of characteristics suggests that key features of a community are shared traditions, identities, and resources, along with mechanisms for collective decision-making and representation (6). Literature on community-based participatory research (7-9) and public health (10) has defined communities in terms of social ties and shared identities.
Communities, however, are not homogenous, but rather have internal divisions, hierarchies, and conflict (7)(8)(9). Moreover, it is important to note that defining the boundaries of a community can be fraught with challenges. Demarcating the limits of a particular community necessarily requires exclusion as well as inclusion -it involves stating who is in and who is out (5). In intervention research, these boundaries may have important implications both for the distribution of resources, and for the application of safeguarding processes. As such, community-level ethical issues may arise as soon as researchers attempt to define the community with which they are engaging.
In this paper, I consider a community to be a group that includes social relationships, shares elements of identity, and engages in purposeful collaboration. In light of the above discussion, this description is not intended to imply homogeneity, nor is it suggested that this definition alone is sufficient to neatly bound any given community.

A community level of ethical analysis in research
As noted above, many fields have longstanding traditions of community-level intervention research. These fields each have distinct values such as humility in community-based participatory research (11), and frameworks for ethical intervention such as Nuffield's ladder of interventions in public health (12). Researchers conducting political field experiments can nonetheless turn to this ethical scholarship to understand some of the issues at stake in research with communities. These issues include the obligations of researchers towards communities, differential effects within the community, and difficulties identifying who can validly represent the community.
The obligations of researchers working with communities are less well-established than those pertaining to individuals, but this is not to say the former are unexplored. Four decades ago, Rappaport identified a dialectic of needs and rights that arises in relation to community intervention, noting that a needs-based view implied requirements for protection, even paternalism; a rights-based view suggested a focus on freedom and self-determination (13). Rappaport called for a synthesis of these ideas, which he termed "empowerment" and described as "enhance[ing] the possibility for people to control their own lives" (p.15). Empowerment continues to hold sway as an important concept in the ethics of community-level research (14). Meanwhile in the field of public health, ethicists have argued for an obligation to support communities' "flourishing," a condition that encompasses resilience and the enabling of individual and collective capacities (15). Like autonomy in individual-level research, empowerment and flourishing are principles that can help to orient ethical research design.
However, the internal heterogeneity of communities creates challenges for actualizing these principles. Interventions targeting a community as a whole may have differential effects within that community. That is, an intervention may benefit some subgroups but harm others; moreover, while some interventions may only target a subgroup within a community, interrelationships within a community may lead to unexpected effects on "non-participants" or the community as a whole (16). Complicating the picture further, different value systems within a community mean that there may be disagreement about what constitutes a benefit or a harm, and for whom (16). This complexity can introduce dilemmas and tensions into the conduct of research.
Beyond benefit and harm, the concept of consent is also a thorny issue when interventions are applied at a community level. Given the complexity of communities as described above, who can speak for the community as a whole? Engaging with community governance or other forms of representation has been argued to be necessary but insufficient to safeguard communities as a whole (11). This insufficiency derives from the diversity within communities, which may not be represented in formal governance structures. Similarly, writing about cluster randomized trials, Gallo and colleagues suggest: "A gatekeeper may give permission for the cluster to participate in the study if they have legitimate authority with respect to the individuals involved and if their authority extends to the decision at hand" (17, p.9); however, they go on to describe the challenges of determining a gatekeeper's legitimacy and accounting for within-group diversity. Questions of "speaking for" a community can also arise with respect to representing or portraying that community, as may occur in the dissemination of research findings (18). Communities cannot be presumed to speak with a single voice.
In the face of these ethical challenges, scholars in fields and methodologies that focus on community-level work have suggested extensions or adaptations to transdisciplinary ethical frameworks such as the Belmont principles. For instance, Campbell and Morris suggest that community psychologists interpret and extend the Belmont principles of respect for persons and beneficence to consider the implications of research for groups, communities, and cultures (19). While adopting a community level of analysis does not resolve the tensions of differential values and effects, it can surface them. Once identified, community psychology scholar Sánchez Vidal argues for a systematic approach to weighing options in community-based research (16): But if there are clear incompatibilities among values or if the beneficence distribution is unbalanced, it is necessary to rank-order values (or actors) so that an option is selected that is most beneficial (or less harming) to the primary values or actors chosen. Although in many instances there will not be a perfect option, it must always be an option that is globally better than the rest in terms of values, consequences, and methods involved (p.78).
This approach enables a systematic management of differential effects and may be of interest to researchers in other fields coping with similar research challenges.

THE ETHICS OF POLITICAL FIELD EXPERIMENTS
Literature on the ethics of political field experiments has not directly confronted the issue of a community level of analysis, although related issues have been raised. The ethical relevance of a naturalistic context and the potential for widespread effects will thus be discussed below, while a more detailed discussion of the ethics of political field experimentation can be found in Phillips' recent review (4).
Political field experiments are carried out in the naturalistic context of the community, rather than in a laboratory or other contrived setting. Naturalistic contexts have implications for informed consent, as some political field experiments may be designed to be "unobtrusive" to further maximize the validity of findings (20). In these cases, participants may have partial or no awareness of the experiment taking place. Minimal risk is one requirement of justifying unobtrusive designs (20). Further, political field experiments not only intervene in natural settings but do so within the framework of an experiment, which requires comparison. Kukla (21) offers an alternate vision of equipoise for this type of research. Her argument centres the contextual appropriateness -rather than the abstract "equality" -of trial arms. Teele (22) also describes the complexity that natural contexts introduce into experimentation. Drawing on examples from international development, she points to jealousy between sites receiving and not receiving a desired intervention, as well as tensions between community and researcher values, as risks that are visible only when the context of the field is thoroughly understood. Teele argues that fulfilling the principle of beneficence in field experiments requires a deep, contextualized understanding of risks.
The benefits and harms of political field experiments may also accrue to individuals and entities beyond those directly enrolled in a study. For instance, in a political field experiment distributing election information, friends and neighbours of those receiving the information may also be affected (4,23). McDermott and Hatemi argue that the widespread impact of political field experiments -and accompanying widespread risk of harm -requires a new ethical principle to guide field experimentation: respect for societies (1). This principle would require field experiment researchers to consider the impact of their experiments not only on participants, but also on non-participants and on the broader population and society in which those participants exist. Whitfield (24), meanwhile, argues that field experimentation in policy research "is distinctively concerned with people in or as committees, associations, states, polities, nations, corporations, and other types of groups engaged in some manner of collective decision making" and "the agency and attendant rights of the relevant collection as a group" (24, p.530; italics in original). Whitfield goes on to argue that this focus on groups as groups requires attention to particular types of wrongs, namely those impinging on groups' legitimate decision-making. I will return to Whitfield's argument below to inform a provisional concept of vulnerabilities occurring at the level of a community.

DEFINING VULNERABILITY
Vulnerability is an important concept in research ethics, where it is used to flag a need for special attention or protection (25). The American Political Science Association's 2020 Principles and Guidance for Human Subjects Research states "Researchers should be especially careful to respect participants' autonomy when conducting research with low-power or vulnerable participants and communities" (26). But what does it mean for a community to be vulnerable? While the concept of vulnerability is subject to ongoing debate in the research ethics literature, this debate has centred almost entirely on the concept of vulnerability as applied to individual research participants. Even concepts of "vulnerable groups" in fact consider vulnerability at an individual level of analysis. To provide grounding for a concept of vulnerability at a community level, I review existing definitional debates around individual-level vulnerabilities below.
Guidance relating to vulnerability in research -including in major research ethics frameworks such as the Belmont Report, the CIOMS International Ethical Guidelines for Biomedical Research Involving Human Subjects, and the Declaration of Helsinkifocuses on individuals who are vulnerable because of personal characteristics, or interactions between personal characteristics and context. The 1979 Belmont Report called attention to issues of capacity for consent when addressing vulnerability, and listed groups whose individual members may be considered vulnerable (such as people who are poor or very ill) (27). More recently, the 2013 Declaration of Helsinki defines vulnerability as "an increased likelihood of being wronged or of incurring additional harm" (28). This definition is referenced in the 2016 CIOMS guidelines, which like earlier reports, points to specific potentially vulnerable groups -including those with reduced capacity for consent, pregnant women, and institutionalized individuals -but also calls for attention to context when determining vulnerability (29).
These shifting guidelines reflect debate and advancement in the ethics literature on vulnerable individuals. Hurst (30) and Luna (31) offer two of the most prominent scholarly conceptualizations of vulnerability. Both critique consent-based definitions, suggesting that these fail to address the full spectrum of possible risks in research, and do not justify the perceived vulnerability of some groups. For instance, pregnant women are commonly labelled as a vulnerable group, but face no particular barriers to providing consent (30,31). Meanwhile, both Hurst and Luna also argue that labelling entire groups as vulnerable runs the risk of stereotyping (30,31). While their critiques of existing work on vulnerability share similarities, Hurst and Luna put forward distinct alternatives. Hurst argues that vulnerability can be defined as an increased likelihood of incurring wrongs. She further states that applying this definition involves asking first if a potential research participant faces a greater likelihood of wrongs, and second whether the researcher or review board has a share in the obligation to minimize that likelihood. Luna argues against defining vulnerability, suggesting the concept should remain fluid. Her work shifts the focus from what vulnerability is, to how it is produced; she suggests that contextual features, individual traits, and study design can all act as dynamic "layers" of vulnerability that wrap around an individual, and as such a person who is vulnerable in one setting may not be vulnerable in another.
More than a decade after Hurst and Luna's initial articulations of each of their positions, the precise nature of vulnerability continues to be subject to lively debate in the literature. However, this debate continues to focus on the vulnerability of individuals. An individual focus on vulnerability is explicit in Hurst's writings (30), narrowing it to an individual's risk of incurring wrongs. While Luna's (31) work brings attention to context, she is interested in context insofar as it "wraps around" the individual. Even early accounts identifying vulnerable groups are, in fact, focused on individuals who belong to those groups, and not the group as a collective unit. For instance, labelling adults with profound intellectual disabilities as a vulnerable group suggests that any person from this group could face increased risks when engaging in research. However, this move from singular to group vulnerability is neither inevitable nor always appropriate. Labelling a group as vulnerable is a categorical decision based on shared traits; it does not assume ethically-relevant social relationships or interdependencies among vulnerable group members which, as discussed above, are key features of communities.
In some instances, vulnerability may be generated by hierarchical or coercive institutional contexts such as prisons, educational institutions, or the military (32). The relationship of interest in these circumstances is not among participants, but rather between individual participants and the institutional order in which they are embedded. For instance, ethical concerns relating to research with people who are incarcerated revolve around the limits to individual prisoner's exercise of autonomy (32), and not the "community" of people in prison as a whole. Prisoners are in this case treated as a category of individuals who share specific, institutionally-imposed threats to their ability to consent.
Thus, although the phrase "vulnerable communities" appears frequently across research ethics literature, the question of what it means for a community to be vulnerable is left unanswered. As such, it is unclear whether current definitions as described above are sufficient to guide research undertaken at a community level.

COMMUNITY-LEVEL VULNERABILITIES
In seeking to address community-level vulnerabilities, I follow Hurst's conceptualization of vulnerability as variable susceptibility to wrongs. I first draw on Whitfield to extend the concept of wrongs to a community level, then argue that communities have varying susceptibility to these wrongs; I suggest that the concept of vulnerability can be applied at a community level of analysis. I also follow Luna (31) in treating vulnerability as an emergent condition rather than an intrinsic property. Luna argues that individuals are rendered vulnerable by circumstance and that vulnerability is layered rather than dichotomous; in keeping with this approach, I will suggest ways in which community-level vulnerabilities can emerge through the process of political field experimentation, rather than seeking to label certain communities or types of communities as inherently vulnerable.

Communities and collective wrongs
If individual vulnerabilities can be understood in terms of (individual) wrongs, community vulnerabilities can be understood in terms of collective wrongs. Whitfield, in his call for ethical principles specific to empirical political science research, provides just such a concept. He distinguishes between three types of wrongs, the first and most familiar of which are individual wrongs, i.e., those perpetrated against distinct individual persons. Second are diffuse wrongs, which pertain to the shared attributes of individuals, for instance the perpetuation of stereotypes. Third, and most relevant here, are collective wrongs, which "are done to the moral status and due consideration of groups" (24, p.532). While Whitfield takes a narrow view of groups as "constituted collectives that function as intentional agents in their own right" (p.532), such as governments or other decision-making bodies, I extend the concept of collective wrongs to a broader array of communities that may be subject to political field experimentation, using the provisional definition of "community" described above.
Whitfield's taxonomy may imply that collective wrongs exist independently from individual and diffuse wrongs. In fact, this is unlikely to be the case. Individuals are embedded in communities, and communities are comprised of individuals; feedback loops among levels of analysis can shape intervention effects, harms, and vulnerabilities (33,34). However, while collective wrongs are not fully separable from individual-level effects of an intervention, nor are they reducible to a simple sum of individual effects. Rather, I posit that wrongs at a community level impinge or affect the elements that define a community, such as social relationships, shared identities, and collective functions. These effects cannot be understood without adopting a community level of analysis.
Income inequality provides a good example. If one of the functions of a city is to support the livelihood and flourishing of its inhabitants, we can say a city is "doing well" when the quality of life there is high, life expectancy is long, and people are happy to live there. Conversely, a city is "doing poorly" when the quality of life is poor, and its residents live short and unhappy lives.
Research in public health has demonstrated that income inequality is associated with poorer and more unequal population health, independently of other major factors including overall wealth (35,36). Various pathways have been hypothesized to explain this effect, including increased chronic stress (36) and weaker public infrastructure and services (37). An increase in inequality may arise from various differential effects within the community, including benefits to some groups and smaller benefits, neutral effects, or losses to others. However, the problem of inequality is relative, by definition, and so may not be visible at an individual level of analysis.
Threats to social cohesion are another instance of a community-level effect. The term "social cohesion" is used to describe the extent of trust and intra-community networks within a group. Like social equality, social cohesion is independently associated with wellbeing (38)(39)(40). If benefits accrue -or are perceived to accrue -to one group and not to another, cohesion may be undermined (40). Tension may also arise when social discourses of "deservingness" are breached (41). Workers, for instance, may object to the provision of more generous benefits to the unemployed; or law-abiding citizens may object to increased supports for incarcerated people. In these hypothetical situations, benefits to a subgroup within the community may have ramifications for the cohesiveness of the community as a whole.
Veering closer to Whitfield's original argument, we can also see threats to governance as a collective wrong that can be incurred by a community. Governance reflects the ability of a community to make decisions that affect the collective. Governance requires both authority (the power to make decisions) and legitimacy (social acceptance of authority) (42). Where governance is undermined, a community loses the capacity to self-organize, support its members, and manage its resources. Political field experiments relating to voting may undermine the legitimacy of elections, a clear threat to governance (23). Meanwhile, field experiments have gained particular prominence in research on international development (43), and scholars have noted that well-intentioned international aid may have the effect of undermining governance -when essential services are provided by an entity other than the government, the government may lose legitimacy as a body capable of meeting population needs (44). This can spur a cyclical effect, as a government that has lost legitimacy will be hobbled in efforts to gather and distribute resources, leading to increased need for outside support and further reduced legitimacy. As in the above examples, this effect can occur without harming or wronging any particular member of the community. In fact, external aid may provide tangible benefits like health or social services. But despite these benefits to individuals, the community's decreased ability to self-govern represents a collective wrong.
The above descriptions are only examples of collective wrongs affecting communities and are in no way intended to represent a comprehensive or exhaustive list. In fact, such a list would be difficult to compose, as it would require a fully-articulated inventory of the ethically-relevant interests of groups. Instead, I hope to illustrate the possibility of analyzing community-level wrongs as a distinct concern, one that cannot be deduced by tallying individual-level wrongs. This is important because analyzing wrongs solely at an individual level may miss risks of the sort described above, thereby creating the potential for unintended, adverse effects.

Community vulnerabilities
If we accept that collective wrongs are those affecting a community's ability to function as a community, and that collective wrongs are not reducible to a collection of individual wrongs, we can then ask whether all communities are at equal risk of being wronged in this way. This potential for a range of levels of risk is central to many conceptualizations of vulnerability. In Hurst's work (30), and in the 2013 Declaration of Helsinki (28), this issue of risk is explicit -both describe vulnerability as an excess risk of wrongs (and, in the latter case, harms). In Luna's argument, layers of vulnerability may accrue or be peeled back depending on context (31). The metaphor of layers therefore implies a spectrum or range of vulnerability, although risk per se is not addressed. In all these accounts, vulnerability tells us that something is more than usually susceptible to damage, or in need of greater care.
Empirical and conceptual literature provides examples of a range in risks of wrongs at a community level. Inequality may pose a greater threat in communities that do not have effective mechanisms for redistribution of resources (35,37). For instance, income inequality may be mitigated by taxation policies. At a smaller scale, neighbourhoods may redistribute resources through informal or formal networks like places of worship. However, in communities without redistributive capacities, the full force of increasing inequality will be felt. Threats to social cohesion, meanwhile, are a more prominent concern in communities with pre-existing divisions and tensions. Teele, for instance, comments on a field experiment carried out by Paluck and Green in Rwanda, where a radio program was developed in which actors voiced dissent with the aim of assessing the effect on listeners' tendencies to defer to authority. Teele argues that promoting the democratic norm of contestation could in fact pose a threat to the fragile balance of a post-genocidal society (22). Threats to governance may be most pronounced where governance structures are weak or are already undermined. This issue is reflected by Cragoe, who notes that traditional governance and other forms of leadership in some Indigenous communities may exist in tension with governance structures instituted by the state. He argues that researchers working with Indigenous communities should consult with the most appropriate community leaders to ensure that the interests of participants are legitimately represented, and to avoid replicating and reinforcing damaging, paternalistic colonial relations (45). These examples suggest that social features and processes -whether internal to a community or imposed by outside forces -can place some communities at additional risk of incurring collective wrongs during the process of political field experimentation.
However, it is critical to note here that the objective of naming community vulnerabilities is not to label specific communities as vulnerable. As described above, research with communities has adopted a focus on empowerment that dialectically acknowledges the risks faced by communities and seeks to actualize community rights and capacities (14). In keeping with this synthesis, I suggest that community-level vulnerabilities in political field experiments are not fixed properties of the community. Rather, community-level vulnerabilities emerge through the interaction between a study or intervention design, and features of the community receiving that intervention. Inequalities may be deepened when an experiment distributes goods to the already-advantaged; social cohesion may be weakened when a study produces benefits that accrue to a specific group, or that violate social norms of fairness; governance may be weakened by an intervention that circumvents or overrides existing, local decision-making structures. Naming these vulnerabilities is not grounds for excluding communities from research, but rather is intended to prompt responsive study designs that minimize harm while respecting community-self-determination.
Creating such designs requires researchers to ask of their political field experiments: In what ways does this study design interact with this specific community's context and makeup to affect the likelihood of wrongs -including wrongs affecting social relationships, shared identity, collective activity, and other elements that make this community a community? What wrongful effects might this study have on subgroups within the community? How might the susceptibility to these community-and subgroup-level wrongs be reduced or offset in the context of the experiment? How might differential effects be fairly and appropriately balanced? Moreover, the dynamic nature of communities mean that any changes introduced to a field experiment to offset vulnerabilities may introduce new, and newly differential, effects. Thus, managing vulnerability will be an iterative process. Sánchez Vidal's framework quoted above can provide a helpful structure for managing that complexity while aligning decisions with relevant values and minimizing overall harm (16).
These questions and processes are critical because a failure to consider community-level vulnerabilities in political field experiments risks incurring detrimental effects. This risk is present even when individual vulnerabilities have been carefully taken into account. For instance, ensuring consent processes are adapted to the needs of individual participants is critical, but will prevent none of the potential community-level wrongs described above such as increased inequalities, damaged cohesion, or undermined governance. On the other hand, tailored strategies rooted in a deep understanding of local contexts including conflicts, resources, and norms, along with collaboration with community members and community leaders, may help prevent unintended wrongs to communities, as communities.

A CALL FOR FURTHER EXPLORATION
This paper adds to emerging conversations on the ethics of political field experiments. While existing ethical frameworks are oriented towards protecting individuals, political field experiments often concern communities. Applying the concept of vulnerability at the community level alerts us to the possibility of increased susceptibility to collective wrongs -wrongs that impede community relationships, identity, and functioning. While not independent of individual-level effects, these wrongs may be overlooked when political field experiments are subjected to individual-level ethical analysis alone. Future scholarship can create a firm definition of community-level vulnerabilities; suggest obligations implied by these vulnerabilities; further consider vulnerabilities at a subgroup level and the interactions of vulnerabilities across levels of analysis; and identify strategies for identifying community-level vulnerabilities and mitigating the risk of associated wrongs. While I have drawn here primarily on literature relating to field experiments in political science and policy, future work can also address the relevance of communitylevel vulnerabilities in other disciplinary traditions.