Unit VIII Journal

Utilitarianism and risk

Morten Fibieger Byskov

Department of Politics and International Studies, Northern University of Warwick, Coventry, United Kingdom

ABSTRACT In day-to-day life, we are continuously exposed to different kinds of risk. Unfortunately, avoiding risk can often come at societal or individual costs. Hence, an important task within risk management is deciding how much it can be justified to expose members of society to risk x in order to avoid societal and individual costs y – and vice versa. We can refer to this as the task of setting an acceptable risk threshold. Judging whether a risk threshold is justified requires normative reasoning about what levels of risk exposure that are permissible. One such prominent normative theory is utilitarianism. According to utilitarians, the preferred risk threshold is the one that yields more utility for the most people compared to alternative risk thresholds. In this paper, I investigate whether and the extent to which utilitarian theory can be used to nor- matively ground a particular risk threshold in this way. In particular, I argue that there are (at least) seven different utilitarian approaches to setting an acceptable risk threshold. I discuss each of these approaches in turn and argue that neither can satisfactorily ground an acceptable risk threshold.

ARTICLE HISTORY Received 28 February 2018 Accepted 10 July 2018

KEYWORDS Philosophy of risk; ethics; utilitarianism; equality

In day-to-day life, we are continuously exposed to different kinds of risk. These risks may range from the longer-term and potentially life threatening, such as climate change, to the mundane, such as catching the flu or being involved in a car accident. Unfortunately, avoiding risk can often come at societal or individual costs. We may, for example, restrict access to public areas during a disease outbreak or impose mass surveillance measures to prevent terrorist attacks. Hence, an important task within risk management is deciding how much it can be justified to expose members of society to risk x in order to avoid societal and individual costs y – and vice versa. We can refer to this as the task of setting an acceptable risk threshold. Judging whether a risk threshold is justified requires normative reasoning about what levels of risk exposure that are permissible. One such prominent normative theory is utilitarianism. Utilitarians hold that pref- erable action in a certain situation is the one that maximizes the most utility for the most peo- ple. Hence, according to utilitarians, the preferred risk threshold is the one that yields more utility for the most people compared to alternative risk thresholds.

Although the cost-benefit calculus of utilitarianism has often been invoked within risk man- agement and assessment (Guehlstorf 2012, 45–47), little philosophical literature has investigated whether utilitarianism can be applied to the task of setting an acceptable risk threshold.

CONTACT Morten Fibieger Byskov morten.byskov@warwick.ac.uk Department of Politics and International Studies, University of Warwick, Coventry, CV4 7AL, United Kingdom � 2018 Informa UK Limited, trading as Taylor & Francis Group

JOURNAL OF RISK RESEARCH 2020, VOL. 23, NO. 2, 259–270 https://doi.org/10.1080/13669877.2018.1501600

Moreover, as I shall argue, the most prominent rejection of a utilitarian approach to risk – that of Hansson (2003, 2013, 2014) – does not address all possible utilitarian positions. In this paper, I aim to fill this lacuna by investigating whether and the extent to which utilitarian theory can be used to normatively ground a particular risk threshold in this way. In particular, I argue that there are (at least) seven different utilitarian approaches to setting an acceptable risk threshold. I dis- cuss each of these approaches in turn and argue that neither can satisfactorily ground an accept- able risk threshold.

The paper is structured as follows. In the first section, I provide a short introduction to the concept of risk and the task of setting an acceptable risk threshold. In the second section, I intro- duce the normative theory of utilitarianism and how it can be applied to the context of risk assessment. I discuss two general utilitarian positions – the actualist and the expected utility positions – and the arguments raised against them by Hansson (2003, 2013). In the third section, I identify five more specified utilitarian positions, discuss whether they are suitable for the task of setting acceptable risk thresholds, and conclude that they are not. In the fourth section, I address two objections to this conclusion, while I reflect on the broader consequences of my argument in the fifth section and how to take the research on ethics and risk forward.

1. Acceptable risk

It is difficult to avoid exposure to risk. Consider, for example, how you exposure yourself to the risk of being hurt in a car accident every time you enter into traffic, whether as a driver, passen- ger, or pedestrian. Now, that does not mean that you will necessarily be in a car accident in your lifetime, of course, even if you expose yourself to this risk on a daily basis and you may even take precautions against this risk materializing, for example, by driving within the speed limit, looking both ways before crossing an intersection, or by putting on your seatbelt.

Technically speaking, we can define risk (R) as the ‘expectation value of an unwanted event which may or may not occur’ (Hansson 2014, section 1; italics in original). The ‘expectation value’ of an unwanted event is calculated as a function of the probability (p) that the event will bring about the expected consequences and the severity (s) of these consequences (Zinn 2009, 5).1 To illustrate, consider how, in the above example, the probability of getting hurt in a car accident can be lowered by driving carefully while the severity of such an incident occurring can be miti- gated by wearing one’s seatbelt.

In many cases, however, we have little direct control, as individual actors, over the extent to which we are exposed to certain risks and their consequences. The risk of being in a car acci- dent, for example, to a large degree depends on the actions of other drivers who may not be as careful behind the wheel. For this reason, it often falls upon public officials to manage the extent to which the public is exposed to risks, for example, through the implementation and enforce- ment of traffic regulations. The problem, however, is that the safest way to minimize the risk of people getting hurt in car accidents is to outright ban cars, which, for obvious reasons, would have enormous societal and economic consequences.

Hence, a problem with risk management is that avoiding or minimizing risk often involves measures that are if not unrealistic in a functioning society will at least often involve societal or individual costs. As a consequence, in such cases the task of risk management is not necessarily to avoid risk altogether but rather to minimize risk exposure while limiting the negative societal consequences of risk avoidance. We can refer to this task of balancing concerns of risk exposure with larger societal concerns as the task of setting an acceptable risk threshold.

Setting an acceptable risk threshold involves balancing two types of error (Hansson 2014, section 3). The first type of error (Ei) stems from safeguarding against some perceived outcome to an extent that is not proportional to the actual probability or severity of that outcome. As such, the first type of error is associated with what we usually call a ‘false positive’: assuming

260 M. F. BYSKOV

that a risk is more probable and/or more severe than it actually is. The second type of error (Eii), is what we usually call a ‘false negative’ and occurs when the probability and/or severity of an expected event is underestimated. The problem here is the opposite than for a false positive, namely that we fail to safeguard against an actual unwanted event. An acceptable risk threshold can be defined as a function of comparing and balancing the probability and severity of these two types of error.

There are three different ways in which we can set an acceptable risk threshold by balancing Ei and Eii. First, we can simply choose to minimize the error that has the highest expectation value. Thus, if Ei (as a function of pi x si) is larger than Eii (as a function of pii x sii) then we should choose to minimize Ei. However, this does not solve the issue, as it would allow us to run a risk with a high expectation value, for example if both types of error have a high expectation value, yet where Eii has an only slightly less severe outcome than Ei.

Second, the acceptable risk threshold could be set at the point at which the combined expect- ation value of the two types of error (i.e. pi x siþ pii x sii) is the lowest. If we face two inevitable risks this solution aims to minimize the overall risk. The problem, however, is that by aggregating expectation values for different people in this way might lead to a widely unequal distribution of risk among affected individuals. That is, if the lowest combined expectation value is the result of one high expectation value of Ei, which is cancelled out by a very low expectation value of Eii, then the burden of this risk threshold falls disproportionally on those who are vulnerable to the consequences of Ei.

Third, we can set the risk threshold at the point in which the expectation value of the two types of error is equal (i.e. pi x si¼ pii x sii) or, more realistically, approximately equal (i.e. pi x si � pii x sii) . This would solve the issue above as it would prohibit the unequal distribution of risk among affected individuals. The problem with this solution, however, is that it could endorse a high threshold for acceptable risk in cases where there is no equal distribution among low-risk versions of Ei and Eii. That is, in some cases, lowering the threshold for one type of risk, thereby protecting one group of individuals from risk-exposure, may actually require a trade-off with other types of risks, leading to an increase in risk-exposure for a different group of individuals.

A major issue in setting an acceptable risk threshold concerns how to make such trade-offs between different risks. How can we weigh different kinds of risks against each other? What kinds of risks do we find more acceptable and which do we want to avoid at all costs? The prob- lem, as Zinn (2009, 13; emphasis in original) argues, is that it is unclear what the losses are with each risk and how much weight they should be given: ‘Central is the sociocultural dimension of values. Risk questions are never just a question of the rational application of objective, value-free problems. Even in the highest technical application, values concerning the acceptability of a spe- cific level of risk or uncertainty are involved’. In this sense, risk assessment and the task of set- ting an acceptable risk threshold is inherently normative (Cranor 1997).

In the following sections I investigate whether, or the extent to which, one influential norma- tive theory – namely, utilitarianism – can resolve this issue of how to trade-off different types of risks. Utilitarianism has often been invoked in risk management due to its cost-benefit calculus. Do the benefits, utilitarians ask, outweigh the costs?

2. Utilitarianism and risk

Consequentialism is the ethical view that an action should be evaluated based on the conse- quences that it brings about (Darwall 2003; Scheffler 1988). Now, this does not tell us much. According to which standard(s) should we evaluate the goodness or badness of an action’s con- sequences? Utilitarianism is the most popular version of consequentialism. According to utilitar- ians, the goodness or badness of an action’s consequences should be measured in terms of the

JOURNAL OF RISK RESEARCH 261

amount of utility they bring about. Usually, utilitarianism is said to hold the maxim that the pref- erable action is the one that maximizes the most utility for the most people.2

How can we take a utilitarian approach to setting an acceptable risk level?3 Hansson (2003, 293–97, 2013, 23–28) has influentially identified and criticized two major utilitarian approaches to risk, namely the actualist position and the expected utility position. In this section, I summar- ize Hansson’s critiques of these two positions before I turn, in the followings section, to argue that we can identify five additional utilitarian position that can avoid the shortcomings of the actualist and expected utility positions.

The first utilitarian position holds that we should focus on the utility of the actual consequen- ces that are brought about by a certain action. That is, according to the actualist position, whether we are justified in running a particular risk, depends on the actual outcome. For example, imagine that we judge that a medical procedure has 50% risk of leading to complica- tions for the patient. If we perform the procedure and no complications arise, the actualist pos- ition holds that we made the right decision in performing the procedure. We can state the actualist position in the following way.

1. Actualist utilitarian position. We are justified in relaxing precautionary measures if (and only if) we thereby actually bring about an equal or increase in utility output as compared to retaining or further strengthening existing precautionary measures.

According to the actualist position, whether or not we ought to relax precautionary measures against exposure to a particular risk can only be judged retrospectively: did our decision actually bring about an increase in utility output, all things considered, or did it actually make things worse? There are two major issues with the actualist position, however, according to Hansson.

First of all, the actualist position only tells us ex post whether we were justified in running a particular risk. Yet, when setting an acceptable risk level, we want to know beforehand whether we are justified in relaxing the precautionary measures (or not). Thus, although the actualist pos- ition might be well suited to evaluate actual actions, it is ill suited for providing action-guidance where it is necessary to predict the justifiability of running certain risks.

A second major issue with the actualist position is that it may actually justify – ex post, of course – taking quite severe risks. Imagine, for example, that there is a 50% risk that relaxing precautionary measures against a disease outbreak will lead to a (significant) increase in mortal- ity from infections and related complications. As above, if we relax the precautionary measures and no such increase in mortality occurs, the actualist position hold that we were justified in running this risk. However, we would not usually say that we justified in taking a risk where someone has a 50% risk of dying, such as flipping a coin for someone’s life. Thus, the actualist position does not sit well with our moral intuitions.

The second version of the utilitarian approach thus holds that we should focus on the expected utility that an action will bring about, rather than the actual outcome. According to the expected utility position, we should choose the action that has the highest probability of yielding the highest utility outcome. Consider, for example, two alternatives. While both alterna- tives have an expected increase in utility outcome of two units, the first alternative has a 50% chance of leading to this increase in overall utility compared to the second alternative’s 75% chance of achieving the same increase in utility output. On the expected utility position, we should prefer the latter alternative. The expected utility position is better suited to deal with risk since it takes into account the aspect of probability. We can restate the expected utility position in the following way:

2. Expected utility position. We are justified in relaxing precautionary measures if (and only if) there is a greater probability that it will lead to an equal or higher expected utility output than retaining or further strengthening existing precautionary measures.

What the expected utility position holds is that, when managing risk, we need to focus on the probability that relaxing (or strengthening) precautionary measures will bring about a better

262 M. F. BYSKOV

state of affairs (i.e. a better utility output). To exemplify, let us imagine that we can relax, strengthen, or retain the current precautionary measures in some unspecified context. While the current control measures yield a utility output of 5 units out of ten, both strengthening and relaxing the control measures will lead to an increase in utility output up to, say, 7 units. (The utility output will naturally be distributed differently between groups, whether we decide to relax or to strengthen the control measures.) Let us further imagine that there is a 75% probabil- ity that relaxing control measures will lead to this increase in utility output, while the probability that strengthening the measures will have the same result is a mere 50%. In this case, according to the expected utility position, we are justified in relaxing the precautionary measures because the chance that this decision will lead to a better state of affairs, measured in terms of expected utility output, is higher than the alternative options.

A major problem, however, is that a commitment to evaluating actions and decisions in terms of their expected utility output allows us to overrule the interests of a few individuals in favor of the many if this leads to an increase in overall utility. Consider, for example, how the alternative of strengthening control measures might have 75% probability of yielding a two-point increase in overall utility output from five units to seven units. However, we can further imagine that while the five utility units of the status quo were equally divided among different groups of stakeholders, strengthening the precautionary measures would lead to an unequal distribution of utility among different groups. We can for example imagine that in this case the public would derive more utility from stricter precautionary measures (say, six units) than those who are tar- geted – for example, car owners or carriers of an infectious disease – by these measures and who will only derive one utility unit from this new control strategy. Thus, while a strengthening of the precautionary measures might lead to an overall increase in utility output, it would argu- ably put an unreasonable burden on those targeted by the measures. Hence, one of the prob- lems with the expected utility position is that allows for the aggregation of utility output at the expense of a few suffering individuals.4

Based on these arguments, Hansson concludes that utilitarianism is poorly suited to deal with risk and the task of setting an acceptable risk threshold. However, as I will argue in more detail in the following, by only focusing on the actualist and expected utility versions of utilitarianism, Hansson risks arguing against a straw man. That is, as I show in the section, there are at least five alternative utilitarian positions that might be better at dealing with risk.

3. Five alternative utilitarian approaches to risk

One of the main objections to utilitarianism, as Hansson’s highlights in his criticism, is the prob- lem of aggregation. The problem of aggregation arises from the utilitarian maxim that overall utility output should be maximized for as many people as possible. Consequently, utilitarians who advocate this maxim would endorse sacrificing the well-being or safety of a few people as long as it is outweighed by an increase in well-being or safety to the majority of society.

To illustrate, consider the case of a potential nuclear meltdown (Hansson 2003). We can avoid this meltdown by sending in one engineer to repair the nuclear power plant. The engineer, how- ever, will most likely die in the process due to radiation poisoning. Is it justified to sacrifice the life of one engineer in order to protect many more members of society from radiation poison- ing? Utilitarians who endorse the utility maximization maxim would answer yes to this question as the overall utility is clearly increased by sacrificing the lone engineer. Yet, to most people, including many moral philosophers of deontological or contractualist persuasions, it is not per- mitted to simply sacrifice the life one human being like that.

Is it possible to construct a utilitarian position that does not suffer from the problem of aggre- gation? It might be argued that we can avoid the problem of aggregation, which persists with the two utilitarian positions discussed in the previous section, by demanding that the

JOURNAL OF RISK RESEARCH 263

distribution of the expected utility should also be the most equal. We can call this the equal dis- tribution of expected utility position. This position holds that we should choose the option that has the highest probability of delivering the most equal distribution of utility among different (groups of) individuals. Let us restate this equality-weighted version of the expected utility pos- ition as follows:

3. Equality-weighted expected utility position. We are justified in relaxing precautionary measures if (and only if) there is a greater probability that it will lead to a more equal distribution of utility among affected stakeholders.

The problem with strict egalitarian approaches like this is that they are vulnerable to what has been called the leveling down objection (Gosepath 2011, section 5.1; Nozick 1974, 229; Raz 1986, chapter 9; Temkin 1993, 247–48). According to the leveling down objection, distributive principles that favor a perfectly equal distribution would endorse lower overall utility output as long as it is distributed equally among stakeholders. Recall the example in the previous section: we can choose to strengthen precautionary measures, which would increase the overall utility output from five to seven measures. Yet, whereas retaining the status quo would distribute the five utility units equally, the benefits of implementing stricter precautionary measures would be distributed unequally. We can, for example, imagine sacrificing one group, so that they only receive one utility unit, in order to protect another group who will receive six utility units. According to the equality-weighted expected utility position, we should decide to retain the existing precautionary measures, but that would lead to a worse overall outcome in terms of util- ity (i.e. five utility units rather than seven).

Proponents of the equality-weighted expected utility position could of course argue that their approach protects some people from being much worse off by implementing certain precaution- ary measures or by exposing certain people to very high risks in order to protect other people a smaller risk, such as in the case of the engineer sent in to prevent the nuclear meltdown. Yet, even in cases where all affected stakeholders would be (much) better off than the status quo, the equality-weighted position would still endorse a lower overall utility output as long as the output is distributed equally among the affected stakeholders.

Consider, for example, how we can protect against some risk by implementing either of two precautionary measures. The first precautionary has a good probability of leading to an increase in utility output from five to eight utility units. Let us further assume that all affected stakehold- ers would benefit equally from this increase, such that if there are two groups of stakeholders each would receive four measures of utility from the implementation of the first precautionary measure. The second precautionary measure would lead to an even larger increase in overall util- ity output, namely from five to nine utility units. However, while the second precautionary meas- ure protects one group equally to the first precautionary measure, the other group receives even better protection from the second precautionary measure than its alternative. Consequently, while the first group receives four utility units from the second precautionary measure, the second group receives five utility units. According to the equality-weighted expected utility pos- ition we should prefer the first precautionary measure as it yields the most equal distribution of utility. However, this conclusion seems odd since the worst off in this scenario – the first group – does not experience any loss between the two alternative precautionary measures. In fact, the second precautionary measure could provide even better protection for some people without sacrificing others.

In order to avoid the leveling down objection, we might further choose to add a clause to the equality-weighted expected utility position that prohibits such leveling down:

4. Equality-weighted expected utility position, with leveling down-prohibition. We are justified in relaxing precautionary measures if (and only if) there is a greater probability that it will lead to a more equal distribution of utility among affected stakeholders – as long as this does not also lead to a decrease in overall utility output.

264 M. F. BYSKOV

Even this amended expected utility position does not get us very far in answering whether we are justified in relaxing precautionary measures, however, as it is also subject to the second problem that we can identify with the expected utility position. The second problem states that the expected utility position (and, by extension, the equality-weighted expected utility position) prohibits cautious decision-making that avoids low-probability risks that have high-stakes out- comes (Hansson 2003). For example, we would usually avoid taking any actions that would increase the risk of a serious outbreak of Ebola, even if the probability of such an outbreak were quite low. This is contradictory to the expected utility approach’s position that we should choose the alternative with the highest probability of yielding the highest utility output. This is so in two ways.

First of all, it is contradictory to the idea that we should always choose the alternative with the highest probability of an increase in utility output. That is, if this alternative also carries a higher probability of some serious consequence, such as an Ebola outbreak, we would usually be justified in choosing an alternative with a lower probability of an increase in utility output if this alternative does not also increase the probability of this serious consequence happening. Second, it is contrary to the idea that we should choose the alternative with the highest expected utility. That is, if the alternative with the highest utility output also involves an increase – whether negligible or not – in the probability that some serious consequence, such as an Ebola outbreak, will happen, then we are usually justified in choosing for an alternative with a lower overall utility output.

Thus, taking into account the need to be risk-averse in terms of certain serious consequences, we cannot make use of the expected utility positions. The fifth version of the utilitarian approach demands that for each alternative we set a ‘security-threshold’ for the probability that the worst- case scenario will happen. This position is usually referred to as the maximin utilitarian position as it aims to ‘maximize the minimal outcome’ (Hansson 2003, 295–96). The maximin position can be stated as follows:

5. Maximin utilitarian position. We are justified in relaxing precautionary measures if (and only if) it does not lead to an increase in the probability of the worst-case scenario happening.

The problem with this position, though, is that it remains indeterminate between cases that do not lead to an increase in the probability of a worst-case scenario. Consider, for example, how both retaining and strengthening current precautionary measures against an Ebola outbreak might not lead to an increase in probability of an outbreak happening. The maximin position does not provide us with any way of deciding between these two alternatives since it is only concerned with the worst-case scenario, even if it would be reasonable to retain the measures so as to avoid incurring further societal and individual costs.

Moreover, the maximin position seems to biased against taking a slightly greater risk with higher benefits. Consider, for example, how getting to work in the morning – whether by private or public transportation – increases the risk of the worst-case scenario happening, namely get- ting killed in a traffic accident. On the maximin position, as it is stated above, going to work would be prohibited due to the increase in the probability of the worst-case scenario occurring. But surely the benefits outweigh the risk in this example: on a household level, being able to work provides an income that can pay for accommodation, food, and leisure activities; on a soci- etal level, employees contribute to the overall economy and productivity. If we are only con- cerned with not increasing the probability of the worst-case scenario happening, we ignore that running some smaller risks can be justified by being traded off against some larger societal and/ or individual benefits.

Hence, for these two reasons, we need to add a comparative element to the maxi- min position:

6. Comparative maximin utilitarian position. We are justified in relaxing precautionary measures if (and only if) it has a lower probability of the worst-case scenario happening than the alternative options.

JOURNAL OF RISK RESEARCH 265

The comparative maximin position allows us to choose between alternatives that have differ- ent, yet insignificant, probabilities of the worst-case scenario happening. On this position, we should choose the alternative that has the lowest probability of leading to the worst-case scen- ario. A major problem with this position, however, is that it will remain indeterminate between alternatives that have an equally low probability of leading to a worst-case scenario, yet might differ in terms of overall output or equality of distribution. In the first case, as in the example above, both relaxing and strengthening precautionary measures might carry the same risk of the worst-case scenario happening, such as an Ebola outbreak, yet be expected to lead to an increase in overall utility output. (Recall, retaining the current measures had a utility output of five units, while relaxing and strengthening the precautionary measures both had an expected utility output of seven units.) In the second case, as in the example above, we can imagine that both retaining and strengthening the current precautionary measures carry the same risk of an outbreak happening while yielding the same expected utility output (say, five units), yet distrib- ute this utility output in different ways. For example, while retaining the precautionary measures would distribute the utility output equally between stakeholders, the option of strengthening the measures would lead to an unequal distribution between stakeholders.

In neither case, does the comparative maximin position provide any guidance on how to choose between these different alternatives. Hence, as with the expected utility position, we need to insert a couple of clauses, namely a leveling-down prohibition (in order to decide in the first case between alternatives with different overall utility output) and an equality-weighting (in order to decide in the second case between alternatives with different distributions of util- ity output).

7. Equality-weighted comparative maximin utilitarian position, with levelling-down prohibition. We are justified in relaxing precautionary measures if (and only if) it (a) does not lead to an increase in the probability of the worst-case scenario happening, (b) has a lower probability of the worst-case scenario happening than the alternative options, (c) has a greater probability of an equal distribution between affected stakeholders, and (d) does not lead to a decrease in overall utility output.

Although this seems like a sensitive approach, in practice there will often be contention between clause (a), on the one hand, and clauses (c) or (d), on the other. That is, taking worst- case scenarios into consideration will often involve violating a commitment to equality (c) or overall utility (d). In the first case, current precautionary measures within infectious disease con- trol illustrate how we might prefer an unequal distribution of utility by imposing quarantines, isolation, mandatory tests, and mandatory eradication treatments on potentially infected individ- uals for the benefit of the greater public health. In the second case, imposing strict hygiene requirements for both healthcare personnel and patients might lower the overall utility output, yet be justified out of a concern with preventing an outbreak.

While we might prioritize the different clauses, this only leads us into the same issues as the other versions of the utilitarian approach. Giving priority to clauses (c) or (d) leads us into the same issues as versions 3 (i.e. the leveling-down objection) and 4 (i.e. prohibition of cautious decision-making). If, however, we give priority to clause (a), it requires us to be extremely cau- tious to an almost indefensible degree. As the two cases above show, a concern for avoiding the worst-case scenario will often lead to choosing an alternative with a much lower overall utility output and/or a more unequal utility distribution. To wit, the maximin utility position will sacri- fice any other concern at the altar of protecting against the worst-case even if the risk of it hap- pening will be vanishingly small.

Thus, in conclusion, although it is possible to identify five additional utilitarian positions than the two rejected by Hansson (2003), neither of these alternative positions are better suited to deal with the important task of setting an acceptable risk threshold.

266 M. F. BYSKOV

4. Addressing two objections

It is possible to raise at least two objections to the arguments that I have presented in the previ- ous sections. In the following I address both in turn.

First, it could be argued that the five utilitarian positions that I identify (and reject) in section 3 are not actually utilitarian positions because they do not merely evaluate actions based on their consequences but also according to additional normative concerns, especially equality and priority for the worst-off. The commitment to take these additional normative concerns, this objection holds, is at odds with the utilitarian commitment to maximize utility for as many peo- ple as possible. However, it is important to distinguish between, on the one hand, the measure by which we determine the goodness or badness of an action and, on the other hand, the princi- ples for the distribution of the units of measurement. In the first case, we can identify different measures for the evaluation of actions. An action can be evaluated in terms of not only the amount of utility that it produces, but also in term of subjective well-being (i.e. the pleasure that individuals derive), the amount of choices that it enables individuals to choose from, or some other objective measure of well-being, such as economic growth. The five approaches discussed in section 3 are utilitarian because they hold that utility should be used as the ultimate unit of measurement of whether a risk threshold is acceptable.

That does not mean, however, that additional normative concerns, such as equality and prior- ity, cannot or should not be invoked in order to determine how an acceptable risk threshold should best distribute these units of measurement among affected individuals and groups. Mill (1863), for example, argues that maximization of utility should work for the greater social good, while Rawls (2001) holds a position in which the maximization of utility, measured as happiness or pleasure, should be weighed against other normative concerns, such as the protection of cer- tain rights and freedoms. In this perspective it is perfectly consistent to hold, for example, the position that utility should be maximized as long as it is equally distributed among stakeholders or as long as the distribution prioritizes the worst-off as long as this leads to a greater social good or protection of individual rights and freedoms.

Second, it could be argued that, while none of the utilitarian positions are ideally suited to deal with the issue of risk, it may be that one or more are better suited for this purpose than others. If this is true, it does not necessarily invalidate utilitarian approaches to setting accept- able risk thresholds. As indicated, adding additional clauses, such as equality-weighing and level- ing-down prohibition, can ameliorate some of the more egregious problems with simple utility maximization positions. We should thus expect these to arrive at a more justifiable risk threshold than the more ‘traditional’ utilitarian approaches that were rejected in section 2 (i.e. the actualist and the expected utility positions).

The main issue with the alternative utilitarian positions discussed in section 3, however, is not whether or not they are better suited to deal with risk than the ‘traditional’ utilitarian positions, but whether they are suitable to make any decisions on risk at all when faced with competing normative concerns without abandoning the commitment to evaluate risk decisions in terms of utility. That is, although utilitarians might theoretically hold that an acceptable risk threshold should lead to a (roughly) equal distribution of utility, not make the worst-off worse off, or not be overly risky/cautious, as argued in response to the first objection, in practice taking such con- cerns into consideration will often lead to a much lower utility outcome. This leads utilitarians into a dilemma: on the one hand, they can reassert the priority of utility, which leads to the problem of aggregation or, on the other hand, they can trade utility off with one or more of the other concerns, which leads to a problem of incommensurability and incomparability (Chang 1997) because utilitarians have no way of assigning priority to non-utility concerns without aban- doning a commitment to evaluating decisions on risk in terms of utility.

JOURNAL OF RISK RESEARCH 267

5. Taking the agenda on ethics and risk forward

What are the broader consequences of the arguments that I have presented in this paper? How should we proceed if utilitarian approaches are unsuitable to normatively ground acceptable risk thresholds? In this section I briefly reflect on the consequences for risk decision-making and the questions that ethicists of risk should investigate.

Decision-making on risk has often followed cost-benefit analyses of benefits and disadvan- tages of running certain risks. As MacIntyre (1992, 180–181) argues, however, the ethical frame- work for cost-benefit analysis is provided by utilitarianism and, accordingly, the main consequence of my rejection of utilitarianism as an ethical framework for setting acceptable risk thresholds thus seems to be that cost-benefit analysis is undermined as a decision-making tool for risk. That does not mean, though, that we cannot and should not weigh and trade off differ- ent kinds of considerations against each other. In this paper, I have repeatedly alluded to at least four such normatively relevant considerations, namely equality, the priority for the worst-off, and the need to avoid both extremely cautious and extremely risky decisions.

This raises at least three questions that ethicists working on risk could address in future research. First, what other ethical frameworks for risk are there and are they better suited to deal with the complex issues posed by risk than utilitarianism? The problem is that Hansson (2003, 2013) has similar objections to the two other prominent ethical approaches, namely deontology and contractualism. While deontology is the idea that human beings have some rights that can never be violated by imposing certain risks on them, contractualism is the ethical position that an action is acceptable or justified if it can be subject to a consensus among all affected parties or, somewhat weaker, if none of the affected parties can reasonably object to it (Scanlon 1998). Several authors have argued that contractualism may be the more suitable ethical position when dealing with risk (Frick 2015; John 2014; Lenman 2008; Millar 2013). I do not here have space to discuss the merits of these frameworks nor of Hansson’s objections. There is relatively little litera- ture that engages with alternative ethical approaches to risk and it should thus be subject to fur- ther research – including, as I have done in the case of utilitarianism in this paper, the question of whether and the extent to which there are alternative versions of deontology and contractual- ism that can avoid Hansson’s objections.

Another question that risk ethicists need to ask is how these normative concerns that are rele- vant for risk evaluations, how they should be assigned value, and how they should be weighed and traded off against each other. As argued in section 1, a major challenge to setting accept- able risk thresholds concerns how to make trade-offs between different societal and individual losses. While this task has often been solved through cost-benefit analysis, the main problem with this approach is that it is often unclear what the losses are and how much weight they should be given. Consider, for example, how the construction of floodgates might pull funding and resources away from the protection of other important societal values, such as education, healthcare, and security. Risk ethicists ought to set out clear-cut principles for assigning weight and value to these different options and for how to trade them off against each other, for example by investigating the ethical potential and challenges of the psychometric approach to risk assessment (Slovic 1987, 2000). The psychometric approach aims to capture people’s percep- tions about risk by tracking people’s expressed preferences to ‘reveal patterns of “acceptable” risk trade-offs’ (Slovic 2000, xxii) and to identify other aspects of risk, such as benefits of running certain kinds of risk and other characteristics that plays into one’s perception of risk, including ‘dread, familiarity, control, catastrophic potential, equity, and level of knowledge’ (Slovic 2000, xxiv).

The third research question that risk ethicists could ask is what ideal and non-ideal commit- ments can play in risk research. Risk research is inevitably a non-ideal endeavor: risks arise from non-ideal circumstances, such as the presence of natural hazards or a lack resources to build resilience, and thus often lack the possibility of ideal outcomes – there will almost always be

268 M. F. BYSKOV

some kind of loss involved when addressing risk. Risk ethicists should ask what role ideal and non-ideal theory play in risk research: To what extent to should decision-making take into account non-ideal circumstances? Should theories of risk aim to set out ideal normative princi- ples for dealing with risk, even if they are not feasible in practice? To what extent can it be justi- fied to set out less-than-optimal risk thresholds? These are all questions that are important for how ethical analysis can shape decision-making on risk and which might circumvent the need for ideal normative theory.5

In sum, the shortcomings of traditional ethical theories when dealing with risk presents a challenge for risk researchers to find normative grounding for risk theories and practices. I have here sketched out a few lines of inquiry that researchers on ethics and risk may follow in order to address these issues and challenges.

6. Concluding remarks

In this paper I have discussed utilitarianism as a normative theory for setting acceptable risk thresholds. Minimizing exposure to risk often involves incurring some societal and individual costs. Utilitarians hold that incurring these costs are justified if (and only if) it increases the prob- ability of an increase in the utility output. While Hansson has argued that this makes utilitarian- ism poor at dealing with risk because it would endorse sacrificing some individuals for the benefit of society at large, in this paper I have argued that Hansson’s conclusion is premature and that it is possible to identify five additional utilitarian positions that avoid this problem of aggregation. I concluded that none of these alternatives, however, are better at dealing with the task of setting an acceptable risk threshold. I finally considered three questions that should be addressed in future research on ethics and risk.

Notes

1. For this paper, I assume that all relevant risks can be identified and that their expectation values – that is, how probable they are to occur and how severe their outcome will be – can be known. This is of course not the case the case in practice where individuals and policy-makers will often have to make decisions under uncertain circumstances. For a discussion of decision-making under uncertainty, see Baron (2017), Fischoff et al. (2006), Hansson (2013, 2014, 2017), Loriaux (2017), Shrader-Frechette (2017), Welch (2017), Zinn (2009), and _Zuradzki (2017).

2. Classical utilitarians include especially Bentham (1972), J. S. Mill (1863), Sidgwick (1907), and G. E. Moore (2004). Within contemporary philosophy, utilitarianism has especially been defended by Goodin (1995), Hare (1981), Harsanyi (1977), and Singer (2000, 2011). For a broader discussion of the benefits and disadvantages of utilitarianism, see especially Smart and Williams (1973), and Sen and Williams (1982). For an excellent introduction to utilitarianism, see Mulgan (2014).

3. For a discussion of the utilitarian basis of risk assessment, see Guehlstorf (2012, chapter 3) and Thompson (1990).

4. See additionally Moore (1966, chapter 5, 18–21) who provides a more general rejection of the expected utility position. Moore (1966, 21; emphases in original) argues that “there is no conclusive reason against the view that [… ] the question whether an action is right or wrong always depends on its actual consequences” since moral blame can only be assigned when the actual consequences are known, even if we have to make decisions based on probable or expected consequences.

5. For a discussion of ideal and non-ideal theories, see Lawford-Smith (2010), Robeyns (2008), Simmons (2010), Stemplowska (2008), and Valentini (2011, 2012).

Acknowledgments

I am grateful to Marcel Verweij and two anonymous reviewers for their comments and suggestions on previous versions of this paper. Work on this paper was partially made possible by a Caroline Miles Visiting Fellowship to the Ethox Centre of the University of Oxford in October 2017.

JOURNAL OF RISK RESEARCH 269

References

Baron, Jonathan. 2017. “Uncertainty and Probability within Utilitarian Theory.” Diametros 53: 6–25. Bentham, Jeremy. 1972. An Introduction to the Principles of Morals and Legislation. Oxford: Oxford University Press. Chang, Ruth. 1997. Incommensurability, Incomparability, and Practical Reason. Cambridge, MA: Harvard University Press. Cranor, Carl F. 1997. “The Normative Nature of Risk Assessment: Features and Possibilities.” Risk 8: 123. Darwall, Stephen L. 2003. Consequentialism. Oxford: Blackwell Publishing. Fischhoff, Baruch, W€andi Bruine de Bruin, €Umit G€uvenç, Denise Caruso, and Larry Brilliant. 2006. “Analyzing Disaster

Risks and Plans: An Avian Flu Example.” Journal of Risk and Uncertainty 33 (1–2): 131. Frick, Johann. 2015. “Contractualism and Social Risk.” Philosophy & Public Affairs 43 (3): 175–223. Goodin, Robert E. 1995. Utilitarianism as a Public Philosophy. Cambridge, UK: Cambridge University Press. Gosepath, Stefan. 2011. “Equality.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta, Spring 2011.

Metaphysics Research Lab, Stanford University. https://plato.stanford.edu/archives/spr2011/entries/equality/. Guehlstorf, Nicholas P. 2012. The Political Theories of Risk Analysis. Dordrecht: Springer Science & Business Media. Hansson, Sven Ove. 2003. “Ethical Criteria of Risk Acceptance.” Erkenntnis 59 (3): 291–309. Hansson, Sven Ove. 2013. The Ethics of Risk: Ethical Analysis in an Uncertain World. Springer. Hansson, Sven Ove. 2014. “Risk.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta, Spring 2014.

Metaphysics Research Lab, Stanford University. https://plato.stanford.edu/archives/spr2014/entries/risk/. Hansson, Sven Ove. 2017. “Uncertainty and Control.” Diametros 53: 50–59. Hare, Richard M. 1981. Moral Thinking: Its Method, Levels, and Point. Oxford: Clarendon Press. Harsanyi, John C. 1977. “Morality and the Theory of Rational Behavior.” Social Research, 623–656. John, S. D. 2014. “Risk, Contractualism, and Rose’s ‘Prevention Paradox.’” Social Theory and Practice 40 (1): 28–50. Lawford-Smith, Holly. 2010. “Ideal Theory—A Reply to Valentini.” Journal of Political Philosophy 18 (3): 357–368. Lenman, James. 2008. “Contractualism and Risk Imposition.” Politics, Philosophy & Economics 7: 99–122. Loriaux, Sylvie. 2017. “Global Duties in the Face of Uncertainty.” Diametros 53: 75–95. MacIntyre, Alasdair. 1992. “Utilitarianism and Cost-Benefit Analysis: An Essay on the Relevance of Moral Philosophy

to Bureaucratic Theory.” In The Moral Dimensions of Public Policy Choice: Beyond the Market Paradigm, edited by John Martin Gilroy and Maurice Wade, 179–194. Pittsburgh, PA: University of Pittsburgh Press.

Mill, John Stuart. 1863. Utilitarianism. London: Parker, Son and Bourn. Millar, Michael. 2013. “‘Zero Tolerance’ of Avoidable Infection in the English National Health Service: Avoiding the

Redistribution of Burdens.” Public Health Ethics 6 (1): 50–59. Moore, G. E. 1966. Ethics. Oxford: Oxford University Press. Moore, G. E. 2004. Principa Ethica. Mineola, NY: Dover Publications. Mulgan, Tim. 2014. Understanding Utilitarianism. London and New York: Routledge. Nozick, Robert. 1974. Anarchy, State, and Utopia. Oxford: Blackwell Publishing. Rawls, John. 2001. Justice as Fairness. A Restatement. Cambridge, MA: The Belknap Press. Raz, Joseph. 1986. The Morality of Freedom. Oxford: Clarendon Press. Robeyns, Ingrid. 2008. “Ideal Theory in Theory and Practice.” Social Theory and Practice 34 (3): 341–62. Scanlon, T. M. 1998. What We Owe To Each Other. Cambridge, MA: The Belknapp Press. Scheffler, Samuel, ed. 1988. Consequentialism and Its Critics. Oxford: Oxford University Press. Sen, Amartya, and Bernard Williams. 1982. Utilitarianism and Beyond. Cambridge, UK: Cambridge University Press. Shrader-Frechette, Kristin. 2017. “Mischaracterizing Uncertainty in Environmental-Health Sciences.” Diametros 53:

96–124. Sidgwick, Henry. 1907. The Methods of Ethics. Hackett Publishing. Simmons, A. John. 2010. “Ideal and Nonideal Theory.” Philosophy & Public Affairs 38 (1): 5–36. Singer, Peter. 2000. Writings on an Ethical Life. Harper Collins. Singer, Peter. 2011. Practical Ethics. Cambridge University Press. Slovic, Paul. 1987. “Perception of Risk.” Science 236 (4799): 280–85. Slovic, Paul., ed. 2000. The Perception of Risk. Earthscan. Smart, John Jamieson Carswell, and Bernard Williams. 1973. Utilitarianism: For and Against. Cambridge University Press. Stemplowska, Zofia. 2008. “What’s Ideal About Ideal Theory?” Social Theory and Practice 34 (3): 319–40. Temkin, Larry S. 1993. Inequality. Oxford University Press. Thompson, Paul B. 1990. “Biotechnology, Risk and Political Values: Philosophical Rhetoric and the Structure of

Political Debate.” Biotechnology: Assessing Social Impacts and Policy Implications 260: 3. Valentini, Laura. 2011. “A Paradigm Shift in Theorizing about Justice? A Critique of Sen.” Economics and Philosophy

27 (3): 297–315. Valentini, Laura. 2012. “Ideal vs. Non-Ideal Theory: A Conceptual Map.” Philosophy Compass 7 (9): 654–64. Welch, John R. 2017. “Coping with Ethical Uncertainty.” Diametros 53: 150–66. Zinn, Jens O. 2009. Social Theories of Risk and Uncertainty: An Introduction. John Wiley & Sons. _Zuradzki, Tomasz. 2017. “Ethics and Uncertainty: The Guest Editor’s Introduction.” Diametros 53: 1–5.

270 M. F. BYSKOV

Copyright of Journal of Risk Research is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.