Some time ago, I lived near a family of brown rats (Rattus norvegicus) who spent their days foraging in the streets. The rats were friendly and harmless creatures, and the young rats were nearly tame. But for reasons that I have never fully understood, my human neighbours thought the rodents should be killed. In fact, they were so certain that the rats lacked moral standing, that killing them did not strike them as in any way morally problematic.

But just how certain were they? Suppose that my neighbours were as much as 99.9% certain that the rats lacked moral standing. Even so, killing the rats might still turn out to be both unreasonable and irrational, even from their own point of view. To see this, note that these people, who were otherwise kind and considerate, would clearly judge it morally wrong to place a poisoned bait that had a small risk to mistakenly kill a human child. But at the same time, these situations seem comparable, such that there is a sense in which:

• (i) performing an act that has a 0.1% risk to kill a creature with moral standing

is just as morally problematic as

• (ii) performing an act that kills a creature with a 0.1% risk of having moral standing.

Cases like these suggest that when we make moral decisions, we need to account not just for empirical uncertainty (such as whether a poisoned bait might kill a child) but also for moral uncertainty (such as whether rats might have moral standing). But that leaves us with the further question of how, more exactly, we should act when we are confronted with moral uncertainty. Up until recently, relatively little work in this area had been conducted, in contrast to the more discussed question of how to act in the face of empirical uncertainty. In their book Moral Uncertainty, William MacAskill, Krister Bykvist and Toby Ord begin to address this gap in the literature.

Moral Uncertainty has nine chapters, as well as an introduction and a conclusion. In chapter 1, the authors start out by arguing that questions about moral uncertainty should be taken seriously. They then state their own position in chapter 2, holding that in the face of moral uncertainty we should maximize expected choice-worthiness (p. 48). This is just like the view that we should maximize expected utility but applied to moral instead of empirical uncertainty. On this view, an option's expected choice-worthiness is its probability-adjusted choice-worthiness; or, more specifically, its choice-worthiness according to different moral theories and adjusted by the probabilities assigned to these theories being true (i.e., our ‘credence’ in them). Choice-worthiness, in turn, is used to represent the strength of reasons for choosing an option (p. 4).

A practical implication of the authors’ account is that for the purpose of moral decision-making, it is not enough to consider the moral theories that we believe are true. We must also consider theories that we believe are false, provided that we assign some likelihood to their being true. The authors own preferred view, Maximize Expected Choice-worthiness, therefore contrasts with the rival view known as My Favourite Theory. According to the latter view, which the authors argue against, we should act according to the theory that we judge most likely to be correct (and which would have my human neighbours kill the friendly rats).

This is where things start to become more complicated. For the theory Maximize Expected Choice-worthiness is only meant as a ‘default’ position about how to act under moral uncertainty, as it can only be used when we are actually able to determine the expected choice-worthiness of our options. But as the authors point out, sometimes this is not possible. To begin with, some theories provide us with ‘merely ordinal’ rankings of choice-worthiness, but no information about the size of the difference in choice-worthiness between options. For example, a Kantian theory may not tell us whether the differences in choice-worthiness between murder and lying is larger than, equal to, or smaller than the difference in choice-worthiness between lying and failing to aid someone in need (p. 60).

Moreover, even for theories such as utilitarianism that provide this information, we may lack a way of comparing their verdicts about choice-worthiness to each other. As the authors point out, if a rights-based theory tells us that it is wrong to kill one person to save fifty others, and utilitarianism tells us it is right, it is hard to say for which theory more is at stake (p. 57). In other words, we may in such cases not have a clear idea of how much utility a given rights-violation is ‘worth’.

In chapters 3 and 4, the authors propose two ways to account for these difficulties, which they call the ‘Borda Rule’ and ‘variance voting’. These methods are framed as natural extensions of the view that we should maximize expected choice-worthiness. In chapter 5, they also discuss to which extent moral theories are in fact comparable to each other. Wrapping up this part of the book, chapter 6 offers a discussion about how to handle tiny credences in theories that postulate options with infinitely or extremely high stakes, as well as on how to deal with incomparability of values.

What remains of the book turns to other interesting questions about moral uncertainty. In chapter 7, it is argued that non-cognitivism in meta-ethics is hard to reconcile with moral uncertainty. Then, in chapter 8, the authors focus on the implications for practical and applied ethics. They convincingly argue that taking moral uncertainty into account will have wide-reaching ramifications for moral decision-making, yet does not thereby make ethics ‘easy’. Even if we must account for moral uncertainty, it remains important to discuss traditional first-order questions about the merits and implications of theories like utilitarianism.

Finally, chapter 9 offers an eye-opening discussion of the value of learning more about morality to reduce one's moral uncertainty – just as someone might want to learn more about the world in order to reduce one's empirical uncertainty. This chapter also contains a delightful discussion of the value of studying moral philosophy.

To conclude this summary of Moral Uncertainty, I was highly impressed by this clearly written and excellent introduction. It answered many of the questions that I had on the topic. Moreover, as good books in philosophy often do, it leaves the reader not just better informed but also with new questions to ask about the subject matter.

As for myself, one question I had was about what type of moral information we should consider when deciding under moral uncertainty. For example, consider the claim that:

I think it is fair to say that when taken in isolation, this piece of moral information would be hard to incorporate into one's decision-making. It does not constitute a fully-fledged moral theory, and it does not even help us order a set of options in terms of their choice-worthiness.

Now, consider a person who has credence in such an ‘incomplete’ piece of moral information. Does this pose a problem for the view that we should maximize expected choice-worthiness? Presumably, it does not. In these cases, the proper response is that one needs better information. So either you find something more to say about this topic, or you can ignore it for any practical considerations. More generally, one would think it is not the job of our theory about how to act under moral uncertainty to accommodate such moral information. Instead, it is the information itself that, in some sense, must ‘become’ sufficiently clear and useful so as to deserve proper consideration.

Since this strikes me as a natural thing to say, it made me wonder if the default view about moral uncertainty really has to be extended to account for merely ordinal theories. Why not instead require that these merely ordinal theories are revised to fit with what is, from what I can tell, a very reasonable view of how to act under moral uncertainty (i.e., to maximize expected choice-worthiness)? Such a move might also be justifiable on pragmatic grounds, as any extension to the view that we should maximize expected choice-worthiness will make the resulting position more complicated, as well as harder to apply in real-life decision-making.

In closing, I am grateful for the opportunity to learn from this book, and I can highly recommend it to moral philosophers with interests in normative and applied ethics.