According to the inefficacy argument there are collective impact cases in which the outcome will occur whether or not you act in the relevant way, and because of this you have no collective-outcome-related reasons to act in this way. The conclusion of this argument seems to be false. For it seems that you might have collective-outcome-related reasons to act in the relevant way. So, we need to scrutinise this argument further. There are two ways in which it can be resisted. One is to deny the implication, or in other words assert that you might have a collective-outcome-related reason to act in a particular way even if the outcome will occur whether you act in the relevant way or not. The other is to deny the description, and thus hold that the occurrence of the outcome depends on whether you act in the relevant way or not, or at least that this might be the case.
I have considered the first possibility, and argued that you might indeed have a collective-outcome-related reason to act in a particular way even if the outcome will occur whether you act in the relevant way or not. In particular, you might have such a reason if your act makes the occurrence of this outcome more secure. In the next three chapters (Chapters 7 through 9), I will consider the second possibility: that the inefficacy argument fails because it is untrue that the outcome will occur whether you act in the relevant way of not, or at least that there is some chance that this is the case.
Philosophical discussion of this issue has centred primarily on how, when and whether the expected utility approach can explain why you have collective-outcome-related reasons. According to the standard view, expected utility can quite straightforwardly explain why you sometimes have such reasons in threshold cases, that is cases where there is a threshold such that if n acts or more of a certain kind are performed some normatively relevant outcome will occur, but if only n – 1 acts or less of this kind are performed this outcome will not occur. In such cases, the occurrence of the outcome might depend on whether you act in the relevant way or not. In THE LAKE,1 for instance, an expected utility theorist would say that each boat owner has a reason not to use the hazardous paint because there is a risk that life in the lake will come to an end if he, or she, proceeds to use the paint. It is a matter of choice under uncertainty. Still, also according to the standard view, the expected utility approach cannot straightforwardly explain why you might have
outcome-1 This case is introduced on p. 66.
related reasons in non-threshold cases, that is cases where no single act makes a difference for whether some normatively relevant outcome occurs or not. To see the problem, consider again DROPS OF WATER. In this case, there are 10,000 people suffering from thirst in a nearby desert, and you are one of 10,000 pint holders each of whom could pour a pint of water into a cart that will be driven to the desert, where the water will then be evenly distributed to those suffering from thirst. If you donate your pint, each person suffering from thirst will get one drop more than they otherwise would have done. If we accept (A) that a single drop of water never can make a perceptible difference to a person’s thirst, and (B) that only perceptible differences can matter morally, the expected utility approach entails that you do not have an alleviation-of-suffering-related reason to add your pint of water to the cart.
There is no chance that doing so will make a morally relevant difference. More generally, if (A) no act of j-ing makes a perceptible difference to anyone’s harm, and (B) only perceptible differences can matter morally, the expected utility approach says you do not have an outcome-related reason to j.2
Although the discussion of non-threshold cases often revolves around the expected utility approach, non-threshold cases also pose a problem for other theories purporting to explain how you should act on a particular occasion. In particular, they pose a problem for any account of reasons which, like the expected utility approach, relies on SIMPLE.3 Take objective consequentialism, for instance. That is, consider the principle that you ought always to do what will in fact have the best consequences, where consequences are understood along the lines of SIMPLE (i.e., an event is a consequence of what you did if and only if it would not have occurred if you had acted differently). According to this principle, you lack a reason to pour your pint into the cart in DROPS OF WATER, at least if (A) and (B) hold. Adding a pint will not make things better for anyone. But this seems incorrect. It seems that you do have a reason to pour your pint into the cart. So, one or the other of the assumptions we have made must be mistaken. If you want to hold on to objective consequentialism, you must show that either (A) or (B) is false. The discussion presented in the upcoming chapters is framed around expected utility, but most of what I say would apply to any account of reasons that relies on SIMPLE.
Whether any particular real-life collective impact case is a threshold case or not will depend on the details. Consider climate change, for example. Climate change does involve thresholds of a sort – at some point, as the sea level rises, a house that is situated on the sea front will be flooded; at some point as temperatures rise and malaria spreads to new areas, a certain individual will be infected by this disease, and so on. However, the question is whether these thresholds are fine-grained enough. Could a single drive in a fossil fuel powered car make any difference to
2 In Chapter 8, I call (B) THE PERCEPTIBILITY PRINCIPLE.
3 This principle is introduced on p. 76.
whether a particular house is flooded, or to whether a particular individual is infected by malaria?
Some have argued on empirical grounds that climate change is a threshold case, much as THE LAKE is(e.g. Broome 2012; Lawford-Smith 2016; Broome 2019).
Others maintain that non-threshold cases are impossible. If they are right, there is no need to consider the empirical facts of any specific case in order to know that the expected utility approach applies also to this case. Some of those who argue that non-threshold cases are impossible (e.g. Arntzenius & McCarthy 1997; Norcross 1997, 2004; Kagan 2011) reject (A). They hold it to be a conceptual truth that some act triggers a perceptible difference in harm in all collective harm cases. They would say, for instance, that one or other drop of water necessarily makes a perceptible difference to the sufferers’ thirst in DROPS OF WATER. Others (e.g. Parfit 1984;
Shrader-Frechette 1987; Barnett 2018; Broome 2019) reject (B). They argue that imperceptible differences in harm can also matter morally. For example, they would say that receiving an additional drop of water when you are suffering from thirst might make a morally relevant difference to the harm you are experiencing even if this extra drop makes no perceptible difference to the way you feel.
In this chapter, I discuss the empirical arguments against the inefficacy argument.
In the two following chapters, I discuss the conceptual arguments. In Chapter 8, I consider whether it is true that acts trigger perceptible differences in harm in all collective harm cases. In Chapter 9, which also is the last chapter of this part of the thesis, I consider the remaining questions: Can imperceptible effects be morally relevant? Does the expected utility approach really explain why we have reasons to act in a certain way in triggering cases?
I will argue that appealing to empirical evidence is fine as long as we can trust the expected utility approach, but that the mere possibility of non-threshold cases
Ways of denying the description
Empirical arguments Conceptual arguments
Rejecting (A)
There are no imperceptibility cases Rejecting (B)
Also imperceptible differences might matter morally
should make us hesitate to accept this approach; that we still lack a convincing argument showing that acts trigger perceptible differences in harm in all collective harm cases; that there is, however, a convincing argument showing that harms can be imperceptible (or, alternatively, showing that a set of imperceptible effects might be morally relevant), implying that acting in the relevant way in non-threshold cases might make a morally relevant difference after all, which in turn means that these cases pose no threat to the expected utility approach. Finally, I will argue that the expected utility approach faces another problem. It does not accurately capture all our intuitions about reasons in threshold cases. This becomes particularly clear when we consider ex-post judgements in cases of overdetermination.
Appealing to Empirical Evidence
Assuming that the expected utility approach gives the right verdict in triggering cases, one could, in some cases, resist the inefficacy argument by arguing that the empirical evidence indicates that doing your part in a collective harm case does make a perceptible difference to the harm. This would amount to arguing that the case at hand is not a collective harm case. Broome (2019), for instance, takes the empirical evidence to indicate that a single drive does make a difference to climate change and its related harms.
Given the atmosphere’s instability, we should expect global weather in a few decades’ time to be entirely different if you go joyguzzling on Sunday from what it would have been had you stayed at home. [… I]t will cause typhoons to form at quite different times and places, and it will lead to a completely different distribution of cholera outbreaks. Your Sunday drive will cause a completely different group of people to be exposed to cholera and other risks of death. Some who would have died will survive because of your drive, and others who would have survived will die. The total numbers who die, and the total amount of harm done in the world may also be greatly altered.
(Broome 2019: 113) If these empirical details are correct, the expected utility approach entails that you have a reason to refrain from going for a single drive in your car.4 Some drives make
4 Is Broome right about these empirical details? He builds his case on studies of the butterfly effect, particularly on Palmer, Döring, and Seregin (2014). I contacted the Swedish national weather service, SMHI (the Swedish Meteorological and Hydrological Institute) and asked whether you can draw the conclusions quoted in the main text from the theoretical papers Broome discusses.
They answered that, as they define it, “climate” is (roughly) statistics on the weather over a certain period of time, usually several decades, and therefore climate is not directly impacted by the butterfly effect. That is, on the SMHI definition, climate truly is an emergent phenomenon.
a difference for the worse, and others make a difference for the better. Because of the complexity of the situation, we cannot know which difference any particular drive will make. Still, on average, single drives makes a difference for the worse, as evidenced by the fact that when enough people drive cars, there is clearly a change for the worse. The expected utility of a single drive is negative.
Alternatively, it can be argued that in any particular case there will be empirical evidence indicating that while most acts make no difference at all, there is some triggering act such that if n acts of this kind are performed the relevant harm will not occur, and that if n + 1 such acts are performed the relevant harm will occur.
This is to argue that the case is a threshold case, much like voting. Lawford-Smith (2016) suggests that climate change fits this bill. Building on climate research, she suggests that there are both macro-level and micro-level thresholds. If a macro threshold is passed, the speed of climate change will increase drastically, and severe harms will ensue. For instance, if the ice sheet covering Greenland were to melt, or the Gulf stream to cease, severe consequences would result. As Lawford-Smith sees it, there are precise thresholds at which these harms are triggered.5 Micro thresholds are less severe but more common. As an example, she suggests that refraining from using the car one day might “delay a tornado by an hour or a day, or set it on a slightly different path, or cause it to occur with a slightly lesser intensity, all of which would cause slightly less damage to persons than otherwise” (Lawford-Smith 2016: 71). Appealing to the expected utility approach, Lawford-Smith then argues that since we have no way of knowing whether a particular greenhouse gas emitting act will take us past some macro or micro threshold, we have reasons not to perform such acts.6
The appeal to empirical evidence rests on two assumptions:
1. THE EMPIRICAL PREMISE: Empirical evidence shows that a single drive in a fossil fuel powered car on average makes a counterfactual difference for the worse to climate change and its related harms.
2. THE EXPECTED UTILITY APPROACH: If an act on average makes a difference for the worse, you have a reason not to perform it.
\ So, you have a reason not to go for a drive in a fossil fuel powered car.
That said, Broome may be right to say that a Sunday drive makes a counterfactual difference to climate-change-related harm in some important sense. The SMHI definition might be too coarse grained to capture all the details of reality.
5 I think we could question this claim. Still, I will not pursue this line of inquiry here. My focus lies elsewhere.
6 In his book Climate Matters, 2012, Broome argues for a similar view.
If Broome, Lawford-Smith and others are correct, the empirical premise is true. Still, if there are non-threshold cases, the expected utility approach itself is questionable, since non-threshold cases are outright counterexamples to it. It seems that you have an outcome-related reason not to flip the switch in HARMLESS TORTURERS, or to pour your pint into the cart in DROPS OF WATER, but the expected utility approach indicates the contrary. Unless we can show that there are no genuine non-threshold cases, we cannot rely on the appeal to empirical evidence.
To see the threat the mere possibility that there are non-threshold poses to the appeal to empirical evidence, consider what would have happened if the empirical evidence had indicated differently. Suppose that empirical research had revealed that there is no possibility that a Sunday drive in a fossil fuel powered car would make a difference to, or affect, who is exposed to cholera or other risks of harm. Although the atmosphere is unstable, it is not that unstable. And even if there are macro and micro thresholds such that if these thresholds are passed some climate-change-related harm will occur, the emissions from a single drive are incapable of triggering these harms. It is only the accumulation of emissions originating in many drives that could trigger them. It might have turned out that global warming is an emergent phenomenon. Just as one oil molecule never makes a difference to whether or not a small patch of oil is slimy, it might have turned out that one molecule of CO2 never makes a difference to average temperatures on earth. It might even have turned out that the quantity of molecules released from any single drive makes no difference to average temperatures on earth.7 If that were the case, the expected utility approach – likewise, other accounts incorporating SIMPLE – would imply that I have no climate-change-related reason to refrain from joy-guzzling. Such a drive has no chance of making a difference to the harm under consideration.
As we can see, if non-threshold cases are possible, the expected utility approach entails that whether or not I have a reason to decline a joy-guzzle depends on physical details of the way in which climate change works at the level of atoms.
This verdict, however, seems implausible. Surely, my reason could not depend on such details. If it had turned out that global warming, climate change and their related harms are emerging events, I would still have had a climate-change-related reason not to joy-guzzle.
The same point can be made in relation to HARMLESS TORTURERS. Suppose experiments show that it is true for most victims subjected to the treatment described in HARMLESS TORTURERS that the flipping of some switch triggers pain, but also that there are some victims with a rare neural configuration who are unable to perceive a difference between any adjacent pain states. Would this mean that the torturers have a reason not to flip the switch when a victim with normal neural
7 Kingston and Sinnott-Armstrong (2018) argue that this is the case. I challenge their reasoning in (Gunnemyr 2019).
configuration is connected to the torturing machine but lack that reason when a victim with the rarer condition is connected up? This would be an odd view to hold.
Surely, the torturers have a reason not to flip their switches even if the victim is unable to perceive a difference between any adjacent pain states in virtue of the fact the victim will be in excruciating pain when enough switches are flipped.
So, the mere possibility that there are cases in which no act makes a difference to the outcome poses a problem for the expected utility approach, and for any approach that require an act to make, or risk making, a difference to harm’s occurrence in order for us to have an outcome-related reason to refrain from performing it. Unless we can show that such cases do not exist, we have reasons to doubt such approaches.
This brings us to the question whether non-threshold cases are possible. This is the topic of the next two chapters.