Moral responsibility, blame and risk

Those who are morally responsible for an outcome are proper subjects of praise or blame for that outcome. So, for our purposes, to say that someone is morally responsible for an outcome O is equivalent to saying that they are worthy of (ie, deserve) praise or blame for O. The project of analysing the necessary and sufficient conditions for moral responsibility, however, remains contentious. We do not hope to offer a complete account here. Instead, we propose three necessary conditions for moral responsibility. The conditions will be specified only roughly, but with enough precision to ensure that they can be uncontroversially applied to the vast majority of cases of harm caused by refusal to vaccinate.

An agent A is morally responsible for outcome O only if

• (Cause) Agent A performs an action Φ that is a relevant part of a causal explanation of O.

• (Salience) Outcome O is morally salient—it is a good or bad outcome.

• (Foreseeability) Outcome O is a predictable or foreseeable outcome of action Φ.

Evidently, refusal to vaccinate frequently satisfies the salience condition: the harms caused by infectious diseases are highly significant. Further, we see no difficulty meeting the foreseeability condition. No doubt, when inoculation was a novel invention and mechanisms of transmission were less widely understood, refusal of a vaccination would not have always satisfied the foreseeability condition. But now, at least in those countries where vaccination is widely available, it is practically common knowledge that non-vaccination can have dangerous consequences for oneself and for others. The most serious source of potential resistance is the causal condition. Although it is undeniable that some individuals' refusal to vaccinate is part of the causal explanation of some infections, it might be disputed whether the ‘relevant’ sort of causal connection required for moral responsibility is present. For the moment, we put this objection to the side: it will be addressed in a section below.

These conditions provide a natural way to identify possible sources of objection to our conjecture that non-vaccinators are responsible for harms that result from their resultant lack of immunity. In the following sections, we consider and address the following purported objections:

1. Non-vaccinators lack the relevant sort of agency to be considered responsible for harms that result from a failure to vaccinate. The harms do not arise from anything that non-vaccinators do.

2. Harms due to a failure to vaccinate are due to collective behaviours of many individuals, and consequently the causal condition for responsibility is not satisfied.

3. Non-vaccinators may be morally responsible for the decision whether or not to vaccinate, but there is no duty to vaccinate. Vaccination is supererogatory, and so failure to vaccinate is not blameworthy.

Objection: non-vaccination and agency

To legitimately consider someone morally responsible for an outcome, it is necessary that the person had a relevant form of agency, and that this agency was involved in the appropriate way in the outcome for which they are attributed responsibility. It is for this sort of reason that it does not make much sense to consider someone responsible for the fact that 2 is a prime number; nor for the actions of their ancestors of 20 generations earlier. These states of affairs do not stand in the appropriate causal relationship to the agency of an allegedly responsible person. No one can causally influence what happened 20 generations earlier, nor can one influence a mathematical fact.

In non-vaccination cases, it might be objected that the harms that result from non-vaccination are not harms that the non-vaccinator has caused to occur; they are, rather, harms that the non-vaccinator has merely failed to prevent. Or a somewhat different claim might be made: that the agent has not intended the harm. For one or both of these reasons, it might be argued, the agency of the non-vaccinator is not sufficiently involved in the harm for us to hold the agent responsible.

Considered generally, this type of objection relies on two claims. First: a conceptual claim is required, distinguishing two types of harm, such that stereotypical wrongful harms fall in one category, while harms suffered as a causal consequence of non-vaccination fall in the other. As already suggested, all manner of distinctions might be used in this way: we can distinguish between harms that are a result of doing versus allowing; or harms that we commit rather than those that we omit to prevent; or harms that we intend versus those that we do not intend.18 Call this first part of the response the conceptual claim.

Second, it must be argued that the distinction identified in the conceptual claim separates states of affairs for which we may be held responsible and those for which we cannot. “We can be held responsible for the harms that we inflict on others, but not for those that we merely fail to prevent”, or something to this effect. Call this the moral significance claim.

This response is implausible because there are numerous examples—unrelated to vaccination—which illustrate the credibility of considering individuals responsible for harms that they both (i) did not intend and (ii) did not cause directly by their actions, but merely failed to prevent. Consider:

• Intoxicated driver. A driver is driving while intoxicated at a safe and reasonable speed. Due to her intoxication, however, she blacks out briefly. During the blackout, she runs a red light, and her vehicle collides with a pedestrian, killing him.

In such cases, the relevant agent does not intend the harm: the harm comes about as a result of failures to act. Indeed, at the crucial time of the red light signal, the driver is not capable of acting in the required way because she is unconscious. Perhaps in some cultures, and in some historical periods, this driver would not be considered responsible for her careless behaviour; but in many cultures and jurisdictions she would be, and it is not credible to suppose that some conceptual error is occurring in such cases. If it is coherent to consider individuals responsible in cases like this, then there is no principled basis for objection that non-vaccinators cannot be considered responsible because they lack appropriate agency for the harms that they transmit to others. That said, it remains very plausible that intentionally caused harms warrant more blame than unintentionally caused harms. We might think that the person who accidentally forgets to get vaccinated deserves less blame than the person who actively decides against vaccination. In addition, in some countries vaccination is difficult to obtain, and where it is not universally and freely available (or heavily subsidised), some people may have difficulty paying the financial costs. Cases like these would seem to be much less culpable, especially where vaccines are not available, and blame may be attributable to local or international public health agencies. But this is no objection to our thesis that non-vaccinators (with easy access to vaccines) can legitimately be considered responsible.

Objection: collective action and individual responsibility

Typically, one person being unvaccinated is extremely unlikely to cause harm to others because the probability of having direct contact with both a contagious person (to become infected) and a vulnerable person (infecting them) is low, at least in populations where vaccine uptake levels are high. Actual transmissions to vulnerable individuals will typically involve a chain of transmission involving dozens of unvaccinated individuals.

Is the collective nature of the harm-causing process compatible with considering individuals responsible? We might have concerns about overdetermination (where multiple chains of transmission within an outbreak are sufficient to cause a given harm) or because of more general concerns that our ordinary assumptions about individual responsibility are not appropriate to collective causation.

To address the concern about overdetermination first: this will depend to a large extent on disease epidemiology. Take two extreme cases:ⁱⁱⁱ (i) endemic malaria: each sick person's clinical case of malaria is likely to be overdetermined since they will be infected with parasites transmitted to them by many other people, and each episode of transmission is itself potentially sufficient to cause clinical malaria; and (ii) measles in a country with a high vaccination rate: here each sick person's clinical case of measles is much more likely to have been transmitted to them by just one other person, and each sick person is unlikely to have had sufficient contact with more than one infectious case.^iv

In cases like the former (endemic malaria), our ordinary ascriptions of responsibility might seem inappropriate. Blaming someone for a bad outcome usually implies that, had the blamed party behaved otherwise, the bad outcome would not have occurred. This is precisely what fails to hold in overdetermination cases.

Legal doctrine has evolved ingenious strategies to address this sort of concern so that parties who participate in multiple, simultaneous wrongs (eg, Jones burns down the factory from the outside, Smith blows it up from the inside; had one of them not acted, the factory would have been destroyed anyway) cannot benefit from the overdetermination of the damage they collectively do. In dealing with overdetermination of risks, the measures become even more conceptually sophisticated.19 ,20 We suggest something similar should happen here: we should adjust our understanding of causation so as to ensure that overdetermined harms are still caused in the relevant sense. Blind adherence to a counterfactual test of causal responsibility will give an absurd answer to questions of blameworthiness. That suggests we should reject the counterfactual test rather than denying that anyone should be blamed for overdetermined harms.^v

Overdetermination aside, the question of how to apportion individual responsibility in cases where one group (those who opt out of vaccination) collectively causes harms to another group (those who are vulnerable to vaccine-preventable disease) remains. Even for cases like measles in high-income countries, one individual alone opting out of vaccination is rarely sufficient to cause infection in another. Such transmission will most commonly involve a causal chain leading back to many other susceptible people, some of whom have chosen to remain susceptible by opting out of vaccination.^vi

Suppose a chain of transmitters (T1 → T2 → ··· → T100) infect an immunosuppressed victim V. Suppose all T1–T100 opted out of vaccination, knowing that this might cause harm to others. Suppose further that they all engaged in similar risk-creating behaviours: they had similar exposure to immunosuppressed individuals and took similar precautions with washing hands and the like. Had T100 obtained a vaccination, V would not have become unwell.^vii But it is also the case that had T99 obtained a vaccination, V would not have become unwell. And so on, for every member of the chain. So who is responsible?

At this point, it is helpful to stress the distinction between moral responsibility for conduct versus moral responsibility for consequences of our conduct. It is possible to think that someone is blameworthy for their conduct without holding them blameworthy for all the bad consequences of their conduct. The distinction is philosophically important because the outcomes of our actions are never entirely under our control, whereas our conduct is—at least to a much greater degree. If we consider individuals responsible to different degrees because their conduct, though similar, led to different outcomes, then we introduce an element of luck into our system of moral judgement. This ‘moral luck’ is a pervasive feature of normal moral practices involving praise and blame, but it is controversial among theorists whether it can be justified.21

So, in the present case, with respect to their conduct, ex hypothesi, each of the 100 individuals is equally blameworthy. They all imposed similar risks and had similar foreknowledge and control over those risks. Their moral responsibility for harm to others will be affected by moral luck because not everyone who behaves in a similarly blameworthy way will cause the same harms. While all non-vaccinators are guilty of equally blameworthy conduct (ie, they have all behaved equally badly), only those who are actually involved in chains of transmission to vulnerable individuals end up being morally responsible/blameworthy for harm because causation is a necessary condition for blameworthiness; and the causal difference between these groups is a matter of sheer luck. With regard to moral responsibility for harm, therefore, moral luck is inescapable.

We claim that those non-vaccinators involved in a chain causing harm are blameworthy for avoidable harms arising from non-vaccination and, in practice, this means that they could legitimately be blamed for these harms.^viii We do not claim that there is a uniquely appropriate way, in theory or practice, to assign blame for the eventual harms, at least in part because of the element of moral luck that influences the outcomes of risky behaviours. With regard to theory, there may be multiple plausible views about the degree of blame that the 100 individuals in question deserve for harm to V. With regard to practice, indeterminacy could be resolved by adopting particular conventions or laws. There are numerous examples of stable moral practices in which we hold people blameworthy for the outcomes of their collective risky behaviour and we see no reason to think that an appropriate custom could not be devised for non-vaccination.

To illustrate, consider three ways (not exhaustive) in which we might ascribe blame to those involved in chains of transmission:

1. Nobody among T1–T100 is blamed for the harm to V.

2. Only T100 is blamed for the harm to V (proximate cause).

3. All individuals, T1 to T100, are blamed to an equal degree for the harm to V.

With regard to the first option: is it plausible to think that no one is blameworthy for the harm? A crucial requirement for any credible account of individual and collective responsibility is that individuals are not left with perverse incentives to (collectively) impose large risks on others for little or no benefit to themselves. Allowing non-vaccinators to escape all responsibility for subsequent deaths fails this basic test.22

The second of these policies (blaming the proximate cause T100) might be appealing in practice because of its convenience: it requires only one individual be found to receive the warranted blame. Theoretically, however, it is implausible to consider the proximate cause more blameworthy (for harm) than any other member of the chain. Of 100 individuals who imposed similar risks and had similar causal involvement in the death, it was sheer luck that made T100 the ultimate transmitter of the harm to a vulnerable person.23 There would thus be a very strong element of moral luck, making it, on some views, unfair to blame only T100, given that T1–T99 were all similarly blameworthy. Nevertheless, a moral community could have practical or policy reasons (eg, deterrence) to blame only the proximate transmitter—even if this blame were out of proportion to her causal role in the harm.^ix

The third policy would regard each individual in the chain (who had reasonable opportunity to act otherwise) as blameworthy for the harms that arise. By considering each morally responsible, we acknowledge that each is implicated in causing avoidable harm.^x

That said, there remain many unresolved questions about how to apportion blame in collective action cases. In our contrived example, the causal roles of T1–T100 are all similar. But in cases where the contributions are heterogeneous (eg, some better foresee harms than others, some take more/better infection control measures than others, etc), we may need to discriminate different degrees of blame deserved by the various agents. For instance, one fundamental issue is whether it is possible to meaningfully quantify the degree of causal contribution that an agent (or chance) makes to an outcome. Intuitively, we can at least make a partial ranking of contributions in multiagent cases: some contribute more than others, though some contributions may be incomparable. While we may take some guidance from this partial ranking information, it will leave some issues indeterminate. For instance, if A makes a greater contribution than B, this may constrain us to ascribe a greater quantum of deserved blame to A. But suppose that a third agent C is involved, and neither A's contribution nor B's contribution is comparable to the contribution of C. (To illustrate with an example: Transmitter A may have made a greater contribution than transmitter B, because although both refused vaccinations, transmitter A made more reckless visits to hospitals and other environments in which immunosuppressed individuals reside. Transmitter C took the most stringent hygiene precautions of all, but C was also the blogger who spread ‘antivaccination’ ideas to A and B, such that they would have been much more likely to get a vaccination had C not written the blog. We might think that C's contribution is so different from A's and B's that it cannot obviously be rated as greater, less than or equivalent in magnitude to either of them.) How much blame does C deserve relative to these others? Any particular answer will involve some arbitrariness.

At least one further question remains: does the blame deserved by one agent change, depending on whether other agents (or chancy processes) are involved in causing the harm? Suppose that an agent's causal contribution to an outcome (including the degree to which he foresaw and intended the outcome) is the same in two scenarios S1 and S2, but that in S1 the outcome is facilitated (or hindered) by another agent, whereas in S2 he achieves the outcome alone. Does the agent deserve the same or a different degree of blame for the outcomes in these alternative scenarios? One possibility is to treat the total quantum of blame deserved for a given harm as constant, regardless of the number of agents, and to distribute it among the agents who causally contribute (eg, if three play similar roles in collectively causing a death, they may each deserve a reduced share of the blame that would ordinarily be ascribed to one individual for that death). This is what typically happens with liability to pay compensation. The amount of compensation required is constant, and mechanisms like joint and several liability allow that constant sum to be distributed in a variety of ways over the liable parties. On this approach, the presence of an additional agent will lead to a reduction in degree of blame deserved. Another possibility is that the blame deserved by the agent is identical in the two cases: despite the fact that our agent had an accomplice in his blameworthy deed, the degree of blame he deserves for the harm remains the same.^xi

Regardless of how we resolve these issues, it should be stressed again that any such proposal that makes causation a necessary condition of blame for harm will be affected by moral luck (just as in the case of blaming the proximate cause T100). Given this, we may wish to consider whether it is preferable (in practice) to adopt other policies—such as vaccine mandates—which would arguably involve a fairer distribution of risk and avoid the case where an unlucky subset of risk imposers would be blamed for their chance involvement in harms.

While it may not be possible to prescribe one uniquely preferred approach to apportioning moral responsibility to individuals in cases of collective responsibility, we believe that at least one of the above methods of attributing blame is defensible, and that is all we require for the defence of our thesis. While the incompleteness of our account may leave some degree of arbitrariness in the choice of rules to ascribe responsibility in theory and practice, this is no worse than the degree of arbitrariness in extant norms and laws: we punish attempted murderers less harshly than successful ones; we blame reckless drivers who cause injury much more harshly than those who luckily avoid injuring anyone; and we are of course resigned to blaming much less harshly those who simply evade detection.

Objection: vaccination is supererogatory

The final objection we consider concedes that non-vaccinators are responsible for imposing risks on others, but denies that they are blameworthy. The idea is that while it might be praiseworthy to obtain a vaccination, it is strictly supererogatory to do so. Refusal to vaccinate is permissible either because vaccination comes at a significant cost to oneself or because the magnitude of risk imposed is not considered morally significant. This latter argument might proceed either by asserting that the absolute risk is small or by asserting that the overall risks and benefits of some social practice are equitably shared across the population.

Alternative duty assignments

It might be wondered what concrete assignments of rights and obligations would be morally acceptable ways to address the risk of infectious diseases. In the absence of mandatory vaccination, should non-vaccinators be required to take other steps to reduce the risks they impose on others? Or should systems of liability hold all non-vaccinators to account for their risk imposition, whether or not specific harms arise?20

An argument that the present assignment of obligations is inadequate can be made from considerations of economic efficiency. The risks imposed by non-vaccinators constitute a cost that is not consented to. In the jargon of economics, this is an externality: a cost that falls on people not participating in the relevant activity. And where there are externalities, there will often be inefficiency: people will choose a balance of activities that does not optimise welfare.29 In this case, the cost to the immunosuppressed of being exposed to a risk of infection is greater than the value to the unvaccinated of remaining unvaccinated. To identify how to remove the inefficiency, we can imagine various thought experiments, regarding what would happen if it were possible to negotiate a mutual agreement without excessive transaction costs, assuming a determinate division of rights.

On one salient alternative, imagine that the unvaccinated have a right to remain unvaccinated, but that the immunosuppressed can band together to purchase the compliance of the unvaccinated with a universal vaccination scheme. Given the very significant gains that the immunosuppressed stand to achieve, it is plausible that they would be willing to pay more than enough to obtain the compliance of at least a very large proportion of the unvaccinated community. The result would be to improve the position of the immunosuppressed, who are now better protected from immunity, and to improve the position of the unvaccinated, who have been adequately compensated for frustrating their preference for non-vaccination. This is a Pareto superior outcome, and the inefficiency has been removed. A scheme like this, on the other hand, could create perverse incentives for people to cultivate or imitate a preference not to be vaccinated.

Another salient possibility is to allocate the immunosuppressed an enforceable right not to be exposed to (avoidable) infection risks. In this case, the immunosuppressed could demand that those who chose not to be vaccinated (i) remain absent from public life, (ii) provide monetary compensation for the risks they impose on the immunosuppressed, (iii) undertake controlled infection^xii or (iv) become vaccinated. It is very plausible that (iv) is the most preferred option for the non-vaccinators, given the very high costs of the other options. So, again, the inefficiency will have been removed: the immunosuppressed will enjoy the benefits of universal vaccination, and a scheme like this would not obviously create perverse incentives.

Both of these scenarios give very clear support to the intuitive claim that it is unreasonable for some to impose the risks of non-vaccination upon the immunosuppressed. It is unreasonable because it is possible to bring about other arrangements that make some better off, and nobody worse off.

Both of the above scenarios are admittedly unrealistic because it would be intolerably costly to engage in all the relevant transactions involving everyone with whom we might need to interact over the course of our lives. But a more realistic, centralised proposal is that those who refuse vaccination might merely be given the choice between (i) providing monetary compensation for the risks they impose on the immunosuppressed or (ii) becoming vaccinated. Monetary compensation could be achieved in practice via imposition of a new (non)vaccination tax. Those who are vaccinated could receive a rebate for the tax in question, so only those who remain unvaccinated end up paying. The revenue raised could be paid out in individual compensation settlements for infection-related harms or could simply be paid directly to vulnerable individuals as compensation for exposure to risk itself.