Compassion as an individual virtue

Collins et al. define a ‘virtue’ as an ‘acquired, stable disposition’ (2012: 252) and compassion as the disposition ‘to be in suffering with someone’ (2012: 255). An important feature of this definition is that it suggests a developmental process in which an individual comes to act with compassion not primarily because it is required of them in a given situation, but because compassionate understanding and action is an important aspect of their ‘character’. To be someone of whom we would say ‘they have the virtue of compassion’ this virtue must be a defining feature of who they are, rather than a role they play.

As we shall see, this constitutive relationship between virtue and identity has important implications for Collins et al. – as (I argue) their account does in fact collapse into an account of role-playing rather than virtue. Nevertheless, their characterisation of ‘compassion’ is an attractive one, and brings a useful conceptual framework to the application of social policy.

That framework comes from a long tradition of social thought that is based on a close relative of ‘compassion’. Both David Hume and Adam Smith – the latter the most famous of the ‘moral sentimentalists’ of the Scottish Enlightenment – insisted on the great moral and social power of ‘sympathy’. For Hume and Smith, it was an inescapable natural and psychological fact that close proximity to the suffering of others would arouse strong feelings of sympathy. Hume's account of sympathy is based on his celebrated empirical psychology, in which the observation of another person's experiences (good or bad) causes a kind of mental ‘contagion’, triggering images, associations and sympathetic feelings in the observer (see Fleischacker, Reference Fleischacker, Fricke and Føllesdal2012). The most basic and intuitive explication of this process is the feeling we have all had when watching someone trip and finding ourselves bracing for a fall. Significantly, for Hume, there is no process of judgement here. ‘Contagion’ is automatic.

This very cursory treatment of Humean sympathy inevitably comes with an element of caricature. But what is of crucial importance is that there would be no context in which feelings of sympathy could arise if there was not a sufficient degree of ‘proximity’. Equally important, the sympathy is naturally aroused; it is an inevitable feeling rather than a conscious cognitive process.

Smith offers a more sophisticated account of sympathy – one which comes close to the ‘mature empathy’ that Collins et al. adopt from some of the recent literature on social care (Hoffman, Reference Hoffman2000; Slote, Reference Slote2007). For Smith, there is a far greater role for judgement than Hume offers. Instead of the more immediate reaction that comes from contagion, Smith's ‘spectator’ goes through a more cognitive process. If we are to exercise the virtue of sympathy, we must be able to view the position of others from their own perspective (Smith, Reference Smith1969 [1759]: 47–8). A contemporary example here might be the case of someone who has committed minor benefit fraud or small theft (though the same process of judgement applies across the full range of moral approval or disapproval). In order to come to an impartial view, we would have to place ourselves in their shoes, to understand their motives, and see if we feel that their action is one that we ourselves could have taken. If we can understand their moral (or social) perspective, and thus understand their actions, we have then come to a sympathetic state of ‘fellow feeling’ (Smith, Reference Smith1969 [1759]: 49). This process, we should note, is not regarded as being infallible. We may mistake the motivations of those we judge, or we may misunderstand the context of their perspective. And we may come to a feeling of fellow sympathy and still judge an action to be wrong. But the crucial point is that judgement is based on a reflective form of empathy.

Something like this account, though not spelt out by Collins et al., could easily be articulated in their description of the virtue of compassion. The compassion thesis certainly gives a full role to judgement in what Collins et al. call ‘mature empathy’, and there is no reason why this could not be developed in a manner similar to Smith's (but not Hume’s) sympathy. A further feature of this ‘maturity’ is an ability to maintain an appropriate degree of detachment from the suffering. Thus, an empathetic response, for Collins et al., does not require a surrender of one's own self-interest or heroic self-sacrifice (2012: 62). For example, a response to domestic violence would be radically incomplete if it stopped at the provision of the most immediate practical solution, providing the victim with secure accommodation and other related necessities. An appropriate response would require understanding and long-term emotional support; a regard for emotional as well as physical wellbeing. What is not required, however, is neglect of one's own life and family, or constant and on-demand reaction to the emotional suffering of the victim.

Though we are not offered such an example, it is helpful to think about what this mature empathy might require in the kind of scenario that most of us are likely to face at some point in our private lives. A person who has been robbed in the street may need more than someone to phone the police for them (a minimal moral duty, and a legal one in countries with Samaritan laws; as we see, for example, in the German legal duty to administer first aid or assistance to a stranger in distress). They are likely to want you to wait while the police arrive, to reassure them and offer emotional support. A bereaved colleague needs more than an email from HR informing them of their statutory right to bereavement leave; they will require understanding and flexibility when they return. A friend who calls in distress after breaking up with a partner may (or may not) take priority over preparation for an important meeting.

As Collins et al. rightly recognise, these circumstances all require judgement (the extent of need, conflicting obligations) and a degree of wisdom, as well as relying on an empathetic reaction. There are also practical demands. This kind of ethical judgement also requires action: ‘virtue theory’ (on this reading at least) is practical, rather than just being reflective. Two of their own key examples of virtue in action are policy responses to domestic violence and bullying in schools (2012: 258), both of which would (in the scenarios we are presented with) require interactive understanding and care for the victims; perhaps alternative activities for bullied children aimed at rebuilding esteem, or shared support for domestic violence victims in sheltered housing. Crucially, in this account, it is likely that individual professionals will need training to enhance their compassionate responses and actions (2012: 267). Thus, it is not only the institutional and professional context that distinguishes these examples from those I have given, but also the likely need for a formalised developmental process.

Thus far there is little to disagree with in the compassion thesis. Yet the examples I have given (e.g. an individual cancelling an important meeting to comfort a friend) hint at at least one fundamental weakness in the argument I am critiquing: the compassion thesis assumes, without real argument, that the intuitive appeal of such examples – in which we are acting as individuals outside of any institutional context – remains once we apply the same account of virtue to professionals in an institutional context.

Thus, the compassion thesis goes on to make the following, central claim: ‘If virtue theory focussed only on how individuals should act it would be highly limited and likely to result in paternalistic judgements of individual behaviour’ (2012: 253). The argument that follows this claim is that, if we want to harness the virtue of compassion for the greater social good, we need to embed the virtue of compassion in institutions and certain types of social policy. Significantly, it is asserted that this cannot be achieved through the development of institutional values, which the authors describe as the dominant way of understanding the normative status of social policy, because an account of ‘values’ does not bring with it the action-orientated practical wisdom of ‘virtue theory’ (2012: 252).

As I am about to argue, however, on closer inspection it is not clear that ‘virtue’ is doing the work that Collins et al. think it is. Few would dispute the worth of their approach to domestic violence, but it is not at all clear that institutions rather than individuals are displaying the relevant sense of virtue, or that ‘values’ are as limited as suggested by the compassion thesis. Moreover, it is doubtful that even individual professionals seemingly acting with compassion are best described in terms of virtue theory. Duty based on professional obligation and rule-governed behaviour – rather than personal inclination – is likely to be more important.

The following section directly addresses this distinction between duty and virtue. Subsequent sections discuss the relationship between virtue and institutional action and, more specifically, the pivotal claim that a virtue account applied only to individuals would be both highly limited and liable to lead to paternalistic judgement.

The distinction between virtue and duty

Let us return to the examples of private compassion I have offered. One of the notable features of these examples is that in none of them (barring the Samaritan caveat) is the behaviour compulsory. Indeed, this is precisely why we say the person who habitually acts in this way displays the virtue of compassion. It is exemplary because it is an exercise in personal judgement or, in Smith's framework, personal feeling regulated by reflection. In direct contrast, the examples of compassionate behaviour we are given by Collins et al. are compulsory and rule-governed, at least if we follow the recommendation to embed such behaviour in concrete policy mechanisms.

This is in part an implication of the turn to institutions, but it is also a reflection of a desire on the part of Collins et al. to ‘argue there are specific qualities of compassion that differentiate it from charitable response and bring it closer to an intersection with rights-based policy’ (2012: 252). Given their concern with the inherent ‘paternalistic’ risks of virtue ethics, this ambition is understandable. But rights – especially ‘deontological’ rights – are ruled based; the right implies a concomitant duty for another person or institution to meet the demands of that right in a way that is ‘written in’ to the right itself, the vast majority of which will actually be formally codified. One of the key attractions of a virtue, however, is that it is a challenge that we can refuse (or neglect) to pick up. Yet this positive feature of virtue is at risk of being supplanted by rights-based rules in the Collins account, thereby obviating the real normative appeal of the account.

Another issue with this characterisation of the role of virtue is that it attributes virtue to individuals who may simply be acting under the guidance of rules. It is of course possible that many professionals will be compassionate individuals in their private lives. But many will not (we all know some). The most patient teacher (in tone and manner as well as in action) could easily mock his bullied children after work, or neglect a bereaved friend in favour of some trivial social engagement. It would be perverse to ascribe the virtue of compassion to this person – but he still does all that is required of him by Collins et al. A possible response here is to say that this teacher has a duty to display the virtue of compassion; but then the most natural thing to say would be that the individual virtue at work here is an inclination to honour commitments and obligations. Compassion is only secondary. Duty is doing the real work in this account.

Schiller captured this point with his caricature of the dutiful Kantian who is commended for doing the right thing even though every emotional impulse is demanding the opposite: the more he has to overcome his real ‘inclinations’, the more virtuous he is (Schiller Werke, Reference Schiller, Petersen and Beißner1943). But would we really think this is the virtue of compassion? Collins et al., if they are true to their premises, would have to say, somewhat implausibly, that we would. A similar description applies in another scenario. When I visit my doctor I hope that she is a compassionate human being; if she is not, then I expect her to act ‘compassionately’ – to pretend the virtue of understanding and an easy bedside manner – as part of her role as a doctor. In so doing she may be a superb actor, making all the right gestures and conveying a tone of concern. But the virtue of compassion is really doing no work here: the doctor is following rules and acting out a duty that is intrinsic to her professional position.

The ‘vice’ of paternalism

So, the relationship between virtue and the professional behaviours that Collins et al. want to embed in social policy is in fact highly problematic. But what about the other claim – that without application, at an institutional level, virtue theory would be ‘highly limited’? This is the claim that: ‘If virtue theory focused only on how individuals should act, it would be highly limited and likely to result in paternalistic judgments of individual behaviour’ (Collins et al., Reference Collins, Cooney and Garlington2012: 253). There are in fact two claims in this statement, neither of which are justified by argument and, in practice, both serve the purpose of unquestioned premises for the compassion thesis as a whole. I deal with the second claim – the charge of paternalism – first.

The first, immediate, objection to this claim is that it is not actually clear what is wrong with ‘paternalism’. Indeed, in the given example of bullying it seems slightly odd to just assume that paternalistic judgement or intervention is in some way inappropriate. The bullied child may, for example, have internalised a derogatory nick-name and made it her own. Here there is a powerful reason to override, through concerted persuasion and support, the apparent preference and choice of the child. This very simple illustration unsurprisingly becomes harder interventions with adults, where there should indeed be a presumption of individual autonomy. But this is still not to say that paternalistic judgement is never appropriate. A common example is adult addiction and substance abuse. We would of course want to intervene in a way that maximises the recipient's control and dignity – but we also do in fact sometimes know what is better for him. Great care is needed in such cases. Ongoing judgement, reflection and dialogue will be crucial. But complete disavowal of ‘paternalistic’ responsibility for others – strangers as well as intimates – looks more like a moral abnegation rather than respectful distance.

We can make more sense of the objection to paternalism if we consider the following explanatory claim:

Here, in the last clause of the sentence, we see a (partial) description of disrespectful control, though this is not necessarily best described as ‘paternalism’. But while the described behaviours are undoubtedly objectionable, it is hard to see why Collins et al. think that shifting the focus of virtue from the individual to institutions will help deter such behaviour. It is equally likely that ‘institutional virtue’ will develop into an uncritical culture that forsakes judgement for process, and thereby loses the critical wisdom required of virtue. This is precisely why we talk of issues such as ‘institutionalised’ racism or sexism. Collective and cultural sexism, supported by all the trappings and power that comes with formal and administrative structures, is far more powerful and pernicious than the isolated judgement of individuals. We have seen this repeatedly across developed economies, where collective culture rides to the rescue of individual prejudice. High profile cases of ruined careers on Wall Street and in the City of London are simply too numerous to list.

Before turning to the role of institutions in more detail, however, there is one final observation to be made on the role of judgement and enforcement. As we have seen, judgement is needed to exercise a particular virtue. But it also requires us to judge the success of an individual in that exercise. Collins et al., as we have seen, express the legitimate anxiety that this judgement could be oppressive. But to complain of individual judgement and ‘enforcement’ is to miss the attraction of virtue theory; there is no point to it otherwise. This kind of judgement cannot be shed without losing the very essence of ‘virtue’.

The enforcement, however, need not be unduly oppressive or prescriptive. A comparison with the ‘perfectionism’ – the more common term for the ‘paternalism’ of Collins et al. – helps to draw this out. Thus, Mill is certainly perfectionist in his account of human flourishing. That is to say, there are standards of individual virtue that are appraised and judged by other people, and by society as a whole. But Mill's perfectionism is open and inclusive (Mill, Reference Mill2006). The enforcement in question is restricted in scope in Mill's perfectionism precisely because it is left to individual judgement; social rather than institutional censure. Doubtless this can be unpleasant. A case in point would be the social oppression of sexual norms. But it is still limited in an important way that is lost when we attempt to embed virtues at an institutional level, or when we attempt to professionalise them. Moreover, Mill's perfectionism suggests a way in which government and other institutions can create the space for virtue and human flourishing without defining the content of specific virtues (Raz, Reference Raz1988), thereby avoiding excessive prescription or the dominance of one (prescribed) vision of the good life. In this light-touch perfectionism, central institutions place a high value on reflective individual autonomy (itself a virtue), and then provide the framework in which we can all pursue this perfection (if we wish) – perhaps through Mill's famous ‘experiments in living’.

The brief detour into perfectionism illustrates two crucial points. Firstly, one can't take virtue ethics without a dose of perfectionist judgement but, secondly, this judgement need not be oppressive. A third crucial conclusion of this section is that the formalisation of virtue at an institutional level in fact, contrary to the compassion thesis, carries greater risk of oppression than a virtue theory confined only to individuals.

As I am about to suggest, many of these difficulties could be avoided if we substituted the value of compassion for the virtue of compassion. Yet Collins et al. insist that the concept of ‘value’ is not up to the task of bringing greater compassion to social policy. I turn to this claim in the following section.

From private to public: collective ‘virtue’ versus institutional ‘value’

One of the central claims of the compassion thesis is that value-based accounts – in contrast to virtue theory – are inherently limited because they do not have the ‘active-orientated meaning’: they do not guide compassionate behaviour (2012: 252). Indeed, this alleged failing of value is presumably the basis of the turn to virtue in the first place. But is ‘value’ really so limited? Or, perhaps more importantly, could ‘virtue’ really do a better job than value for those who want to see a more compassionate welfare state?

In answer to the second question, we must, at least provisionally, say that the virtue theory does not offer an improvement, for the simple reason (as this paper has been arguing) that the account we are offered ends up not really being an account of ‘virtue’ at all. This leaves the question of values. Here there is another distinction to be considered. On the one hand, virtue theory is praised because it ‘links the individual with the collective and frames the discussion for political discourse in shared understandings’ (2012: 253), thereby extending the scope of virtue beyond the apparent limitations of individual judgement and action. Yet, on the other hand, value is dropped from the discussion before it even gets a look in as an alternative means of thinking about our ‘shared understandings’ in general, and about compassion as a value in particular.

But, in fact, value seems to have the following advantage over virtue in this context. It is precisely because values are abstract that they are suitable for the collective framing of meaning. We collectively frame and dispute – through innumerable social and communicative processes – the value and meaning of compassion in our society. We might find, for example, that compassion extends to animals in a way that it does not in other societies, or different societies may have very different views of compassion in end of life care. But the point of compassion as a virtue is that it adds to that meaning by locating it in a specific context for a particular individual. If values set the abstract meaning, then the exercise of a virtue displays a particular interpretation.

We do of course sometimes collectively judge these interpretations. The merits and follies of public figures – or those made public by exceptional circumstances – help us think collectively about virtue, and may feed back into collective reflections on the underlying value. Judgement in these cases may force us to collectively reflect on the guiding value. High profile cases of assisted suicide are a case in point: the action of compassion problematises the value of life. But, far more typically, virtue judgements are made about the local and the particular; they are not directly about meaning, and nor do they extend beyond relatively small circles of judges.

These considerations – without adding more complexity – suggest that, prima facie, ‘value’ is in fact the more appropriate way of thinking about the collective framing of compassion in social policy. In the remainder of this article, however, I suggest a different account of the relationship between compassion and social institutions, one that circumvents the need for an in-depth account of framing and the social creation of shared understanding. The account I offer is intended to apply specifically to the virtue of compassion and not across all types of virtue and is based on two crucial assumptions: firstly, the Humean assumption that we have a natural tendency to sympathy if we are close enough to the suffering of others; and, secondly, that welfare institutions have a legitimate role to play in creating those conditions of closeness.

The institutional structure of empathy

I have set aside the issue of value formation in part because of its complexity and in part because we do not need to address it directly in this context. But we do need to make an important value-based assumption in order to ground the remaining argument of this paper. The assumption, simply, is that it is indeed desirable to see more compassion in modern welfare states. Few would perhaps say that they don't want such an outcome, but the credibility of such denials is severely strained by so much contemporary welfare rhetoric. The argument here is thus addressed to those who share the underlying values that underpin the desire for a more compassionate treatment of welfare recipients. The values of equality and dignity play a major role in this context. It is these that drive the desire to develop institutions that, while not actually exercising the virtue of compassion themselves, create social circumstances in which individuals may develop more compassionate beliefs about the virtues of those that need to rely on ongoing welfare provision.

Constructing such an account would be a full research agenda – and far beyond the scope of this article – but the key principles are relatively clear. The general principle is that institutional and policy design can create the conditions in which individuals are inclined to experience feelings of compassion, or similar feelings of sympathy and empathy. Put simply, if we never come across someone suffering from material poverty (or from other concerns of social policy), it is unlikely that we will be moved to compassion or sympathy; and that means that there is no trigger for, or opportunity to exercise, any of the related virtues. This is the central insight of Hume's and Smith's moral philosophy: we need to witness and feel the suffering of others to be moved to action. Hume and Smith are therefore actually more optimistic about the power of individual compassion than Collins et al. The presumption is that closely witnessed suffering will arouse individual sympathy.

Of course, if we want this phenomenon to lead to social policy outcomes, rather than individual acts of compassion and charity, we will need to draw an institutional link. There are a range of ways in which this might be achieved. Broadly speaking, however, the range is set by a contrast between direct and indirect drivers of empathetic feeling. The direct drivers I will go on to discuss are the result of social and public policies that bring people into direct contact either in institutional settings or in everyday life. In the next section, however, I first discuss the ways in which different forms of welfare distribution can either encourage or discourage empathetic feelings.

Welfare distribution and the mediation of empathy

Over the last twenty-five years, there has been a burgeoning debate on the role of welfare ‘regimes’ in shaping attitudes to the welfare state (Esping-Andersen, Reference Esping-Andersen1990). Thus, countries that tend towards wider welfare or even ‘universal’ coverage – in other words countries in which all people are entitled to certain benefits or services as a right of citizenship – are likely to have stronger welfare states than regimes that target support selectively, focussing only on those in greatest need. The typical methodology is to take a large data-set (often the Luxembourg Income Study) and to compare ‘universal’ welfare states with more targeted regimes, leading to the conclusion that targeted regimes spend less on welfare, thus indicating a lower level of public support for the welfare state. And a typical contrast is between Nordic countries, with broad or universal coverage and strong (generous and popular) welfare states, and ‘Anglo Saxon’ countries – recognisably the USA and the UK – that target welfare support only on the neediest (and with increasing degrees of conditionality).

A more sophisticated approach also takes in the role of stigma, with some research suggesting that highly targeted regimes also tend to be characterised by a high degree of stigma (Rothstein, Reference Rothstein1998; Larsen, Reference Larsen2008), with recipients of welfare and targeted services held in low esteem by their fellow citizens. There is also a wealth of literature on the social–psychological dynamics of this type of process. Much of it is based on the creation of ‘in’ and ‘out’ groups and the process of ‘othering’ (Taifel and Turner, Reference Taifel, Turner, Worchel and Austin1986; Opotow, Reference Opotow1990; Montada and Schneider, Reference Montada and Schneider1989), which, in turn, can affect perceptions of ‘deservingness’ (van Oorschot, Reference van Oorschot2000) and willingness to redistribute wealth and welfare resources (Brewer and Kramer, Reference Brewer and Kramer1986). Taken together, these two bodies of literature suggest a picture of the institutional structure of the virtue of empathy, or the lack of it. What we are seeing here, if the conclusions of these two bodies of literature are correct, is a failure of empathy, vividly depicted in the language of othering – the very opposite of empathetic feeling.

This is not a new phenomenon. The printed press in Britain has long been a factor in attitudes to poverty and the welfare state, a relationship that has been described in detail by Golding and Middleton in their classic study of the creation of ‘scrounger-phobia’ in late 1970s Britain (Golding and Middleton, Reference Golding and Middleton1982). More recent evidence from the US suggests the media has an important role in ‘framing’ the way that social policy is perceived (Tighe, Reference Tighe2010) and, in turn, shaping attitudes towards welfare recipients.

Empathy and the market: Titmuss and the gift relationship

There are also other institutional drivers of compassion and related concepts and virtues. One whole area that is ripe for re-evaluation is the distribution of welfare goods and services through the market, or through the so-called ‘quasi-market’. In both Britain and the US, there has been a growing faith in the power of the welfare consumer to drive efficiencies and higher standards through the power of consumer choice. There is a clear empirical question here of how far market mechanisms do achieve the stated goals of efficiency and satisfaction. The vastly expensive American system of private insurance – in fact amongst the most expensive in the world in terms of the amount spent per person (Mahon and Fox, Reference Mahon and Fox2014), with notoriously unequal coverage – is a case in point.

But there is also the more normative question of how the turn to the market might have an impact on feelings of empathy and support for the welfare state. On this question it would to be useful to return to the work of Richard Titmuss, over four decades old but increasingly relevant in a marketised welfare state. For Titmuss, one of the key features of social policy is that it creates a sense of mutual interdependence (Glennerster, Reference Glennerster2014). His view was that social policy unites people and creates a common sense of identity, in contrast to the individualising effects of the market. Titmuss applied this analysis to the provision of replacement blood for medical purposes. In The Gift Relationship, he found that the fully voluntary British system, in which there was no financial reward or other incentive for blood donation, fared better than a system based on payments (typical at the time in the US), both in terms of the quantity and the quality of the blood made available for medical use. Famously, the result of the study was that people were inclined to offer less blood when paid than when it was given freely on a voluntary basis, with the only reward being the ‘warm glow of giving’ (Costa-Font et al., Reference Costa-Font, Jofre-Bonet and Yen2013). This motivation does not survive the cash nexus, at least in the case of blood donation.

Yet the insights of The Gift Relationship have been curiously neglected in contemporary discussions of the moral psychology of attitudes towards welfare distribution. Any future debate about the role of virtue and compassion in social policy will need to engage with these issues, especially the role of the market in shaping (or eroding) the virtue of compassion. This is likely to involve, among other things, conceptual analysis of the relationship between altruism and compassion, and the way in which these both interact with more duty-based notions of reciprocity. And this, in turn, should lead to a more empirical research into which institutional structures – and which specific policies – best meet the normative objectives that arise from such an analysis. A key test case could, for example, be the use of ‘personalised budgets’ – which give individuals the cash value of their care to spend as they choose – in social care.

Thus far I have outlined two institutional drivers of attitudes to welfare and distribution. The first is the balance between targeted and universal welfare provision and the way in which this can have an impact, positive or negative, on perceptions of welfare recipients. The second, as we have just seen, is the under-explored relationship between market-based welfare distribution and broader attitudes to the welfare state; a relationship that is likely to have important implications for the compassion thesis. In the final sections of this article, I am going to explore a third institutional driver: the extent to which social policy and public services bring net welfare ‘losers’ (in crude terms, those who pay in more than they take out) into everyday contact with welfare recipients.

The proximity thesis

Titmuss’ study of the distribution of blood and the more recent literature on welfare regimes share at least one crucial characteristic: the ‘relationships’ fostered by these institutions and polices are indirect and impersonal. This is perhaps clearest in the case of the blood gift, in which donors had no idea of who would receive their blood. As we shall shortly see, the situation is more complex in the case of welfare regimes, but the institutional structures studied in the literature are predominantly indirect too: the universal-targeted axis is overwhelmingly conceptualised and measured within the framework of different social security systems, meaning that the relationship between individuals is indirect and mediated by administrative structures.

In contrast, what I am calling the proximity thesis seeks to develop and understand policy mechanisms that create a direct connection between different types of people. Two prime candidates are education and housing policy, both of which have a very direct bearing on the degree of social proximity to be found in different welfare states. My argument here, of course, is that this will also have a direct bearing on the extent to which the Humean natural propensity to empathy will be given the opportunity to emerge as a social virtue. It is to be stressed that the account offered here is at the level of intuition and hypothesis. But it is in principle a testable hypothesis. This section outlines what we might expect in the way of proximity and compassion in the case of housing policy in Britain and the US.

The intuitive case is simple. On recent commutes to work through a relatively deprived district of London, I have observed, variously: a Somali mother who could not speak English insist that her children did, regardless of her exclusion from the conversation; a single mother being shouted off the bus for delaying the journey by trying to use someone else's pass, when it transpired that she was desperately trying to get her child to school; and, more directly and personally, the loss of a friendly neighbour who, despite holding down a full-time job, could no longer afford to live in the area following cuts in housing support. None of these experiences would have occurred had I lived in a segregated middle-class area; and there would have been no immediate context and experience to counter-balance the negative narratives of ‘the poor’ that are so dominant in the media. Distant judgement rather than more immediate compassion may well have been the stronger force.

The particular social mix of this area is actually, in part at least, a result of an interventionist social policy. The early years of the Blair government in Britain gave rise to a dedicated attempt to address the issues of concentrated poverty and social exclusion in urban areas. As early as 1997, the new government formed the Social Exclusion Unit, which later came to see neighbourhood as a key battle ground, stating the ambition that ‘within 10 to 20 years, no one should be disadvantaged by where they live’ (SEU, 2001). The SEU stated in 2000 that ‘communities function best when they contain a broad social mix’ (SEU, 2000). This belief in the value of social mix (in practice mixed housing tenure) was then embedded in central planning guidance issued to local housing authorities (DCLG, 2006) and has since been accepted as an orthodoxy in British social housing.

In the US, there is a longer tradition of intervention but, equally, there are also more extreme concentrations of poverty, with the additional factor of race and socio-economic disadvantage. The policy trajectory, however, has followed a similar course, over time moving to a broader view of disadvantage that encompasses the value of social mix. Thus, in 1992, a US Congressional Commission set up what came to be known as the Hope VI programme, which allocated $7.5 billion to ‘severely distressed’ public housing areas over ten years. Whilst this programme first concentrated on improving life-chances, without seeking to change the composition of these neighbourhoods, it soon went on to adopt measures designed to introduce social mix and ‘de-concentrate’ poverty (Berube, Reference Berube2005).

Some Hope VI projects also went on – in marked contrast to any form of British neighbourhood intervention – to pursue a policy of relocation, diversifying poor areas by moving some of the poor out. The positive aim was to provide those who did move out with the opportunity to live in less disadvantaged areas. This approach to de-concentration was also taken up by the Moving to Opportunity (MTO) programme, instigated by the Department of Housing and Urban Development in the US. The most striking feature of this programme is its explicitly experimental nature: 4,604 low-income families in five US cities were split into three groups: a group given vouchers and freedom to move anywhere they chose; a group given vouchers but constrained to live in deprived areas; and a group given no assistance.

If the proximity thesis is correct, these are clearly the type of structural interventions that we should be pursuing. What we do not yet know with certainty, however, is the extent to which neighbourhood social mix does in fact increase feelings of empathy or compassion between different types of household. The key reason for this is the MTO ‘experiment’, as well as other systematic attempts to promote social mix, has not actually attempted to measure the impact of proximity on these virtues, focussing instead on the popularity and quality of the neighbourhood and on specific outcomes for individuals (typically health, education and ‘worklessness’). At the level of the neighbourhood, some of the evidence is encouraging; the stigma of deprived neighbourhoods declines when they become more mixed (Dean and Hastings, Reference Dean and Hastings2000; Fitzpatrick, Reference Fitzpatrick2004; Allen et al., Reference Allen, Camina, Casey, Coward and Wood2005), and the neighbourhoods can be popular with would-be house buyers (Rowlands et al., Reference Rowlands, Murie and Tice2006; Tighe, Reference Tighe2010).

With one crucial exception – the role of ‘peer effects’ in changing behaviours and expectations in workless households – very little consideration is given to the social dynamics created in mixed communities. To the extent that it is, there is a certain moral asymmetry: the poor and workless are to learn the virtue of self-sufficiency from the employed and independent. Of course, there is a degree of caricature here, and much of the literature touched upon in this section argues that concentrated worklessness does have an impact on attitudes to work. The expectation of peer effects has also diminished with the growing body of evidence that finds no such positive effect on labour market outcomes (Tunstall and Fenton, Reference Tunstall and Fenton2006; Ludwig et al., Reference Ludwig, Duncan, Gennetian, Katz, Kessler, Kling and Sanbonmatsu2012). But the important normative point is that the issue of mix and neighbourhood renewal – and the body of empirical evaluation accompanying it – is framed in a way that assumes that the cultivation of virtue must be one-sided. The ‘teachers’ have nothing to learn.

As Collins et al. rightly insist, we do not want ‘compassion’ to be a thin veil for patronising pity (2012: 251), or for ‘mix’ to be an excuse for self-satisfied ‘grittiness’. It is therefore important to stress that the proximity advocated here does not require active interaction (a common misinterpretation of the case for social mix (Morton, Reference Morton2012)) or an insistence that all become friends. Observation may well be enough, as in the case of the struggling mother on the bus. Broader awareness of local labour markets, transport issues, costs of housing could all play a part as well. More direct interaction observation – and a degree of real interaction – can take place in shared public services, most notably in schools and nurseries. Sure Start, the UK's version of the US Head Start, not only sought to improve the early development and long-term educational attainment of children in deprived areas, it was also seen as a vehicle of mix and interaction – a role that it has played with some success (DCLG, 2007). The key point is that these sites of interaction are not seen by users as an experiment in social engineering, but as a part of normal, everyday life.

This process of mix and compassion will clearly be imperfect. Some will be poor observers, and some will maintain preconceptions no matter what they observe. Indeed, this variability in individual capability is precisely why the language of individual virtue (with judgement and the tacit ideal of perfectibility) is so fitting in this context. And, of course, it may prove to be the case that the Humean assumption is wrong: the psychological predisposition to sympathy or compassion witnessed at close quarters may not be as strong as we might hope. This would have significant normative implications for the proximity thesis, and indeed for the status of compassion as a virtue in social policy. But it is also striking that, at present, the answer to this question is largely left to intuition. If the value (or otherwise) of social mix is to be taken seriously by policy makers and social scientists, we need to see a broader approach to the ways in which welfare institutions shape our beliefs and attitudes to different social groups, including those most in need of care and compassion.Footnote ¹ ‘Distance’ then, both spatial and psychological, may well be just as important as the more standard universal-targeted axis found in the current study of welfare ‘regimes’ – and possibly more important than the preoccupation with poverty outcomes in the literature on neighbourhood effects. If we are really to take seriously the structure of social identity more broadly, and the place of compassion more specifically, we need to find out if it is.