I've quote the following from the utilitarianism.net study guide with their permission.

Introduction

Peter Singer's 'Famine, Affluence, and Morality'¹ is widely regarded as one of the most important and influential texts in applied ethics. This study guide explains Singer's central argument, explores possible objections, and clarifies common misunderstandings.

The Argument

Singer argues that most of us in affluent societies are making a terrible moral mistake. When we look at distant suffering—such as results from global poverty, famine, or disease—we tend to think that helping is morally optional, or what philosophers call "supererogatory". Even if we could very easily give more to effective charities to help, we think that to do so would go "above and beyond the call of duty". It would be generous to give more, we think, but hardly required. We assume it is perfectly fine to instead spend our money on expensive clothes, travel, entertainment, or other luxuries. But Singer argues that this assumption is mistaken. Instead, he argues, it is seriously morally wrong to live high while others die.²

Singer's argument for this conclusion is straightforward, resting largely on a key moral principle that we will call Singer's Rescue Principle. The argument may be summarized as follows:³

P1. Suffering and death from lack of food, shelter, or medical care are very bad.

P2. We can prevent such suffering and death by donating to effective charities (in place of consumer purchases).

P3. Many of our consumer purchases are morally insignificant: we could give them up without thereby sacrificing anything morally significant.

P4. The Rescue Principle: If it is in our power to prevent something very bad from happening, without thereby sacrificing anything morally significant, we ought, morally, to do it.⁴

Therefore,

C. We ought, morally, to donate to effective charities rather than purchasing unnecessary luxuries.

Note that, although Singer is a utilitarian, this argument does not rely upon utilitarianism as a premise. P1 – P4 are all claims that non-utilitarians (and even non-consequentialists) could accept.

This is a really striking argument. The four premises each seem perfectly plausible. The conclusion logically follows from these premises. Yet the conclusion is radically at odds with how almost all of us live our lives. Every time we purchase something unnecessary, Singer’s argument implies that we ought instead to donate that money to effective charities. When you think about how this would apply to your own life, it could well turn out that the majority of the purchases you make in your everyday life would be considered morally wrong. Most of us could probably live significantly more frugally, without sacrificing anything morally significant, and use the savings to relieve suffering or even prevent several untimely deaths. According to Singer's argument, that is then precisely what we are morally required to do. (Note that similar arguments could also apply to one's choice of career.⁵)

Could such a radical conclusion really be true? You are probably already trying to think of ways to dismiss it. But it is not enough to simply reject the conclusion. Since the conclusion logically follows from the premises, to reject it you must show one (or more) of the premises to be false.

Assessing the Premises

Premise 1: Badness

The first premise claims that suffering and death are very bad. That is hard to deny. Any plausible ethical theory—whether utilitarianism, deontology, virtue ethics, etc.—will agree that, all else being equal, suffering and death are bad,⁶ especially suffering that is extreme, involuntary, and uncompensated. 

Premise 2: Preventability

The second premise is similarly secure: We can prevent suffering and death by donating to effective charities. Some “aid skeptics” are critical of foreign aid programs. This might suggest that we just do not know whether a given charity actually does any good. Some charitable interventions, on closer examination, even turn out to be counterproductive. However, while many charities have little impact, the most effective charities do a remarkable amount of good. Fortunately, finding effective charities is easy by consulting reputable sources such as GiveWell's in-depth charity evaluations. Even prominent aid skeptics do not deny that GiveWell's top-rated charities are genuinely effective. So there is no real question that well-targeted donations can be expected to prevent a lot of suffering and death. (Of course, the argument will not apply to anyone who lacks the resources to be able to make any such donations. It is exclusively directed at those of us who do, at least sometimes, make unnecessary purchases.)

Premise 3: Insignificant Sacrifice

The third premise claims that we could give up many of our consumer purchases "without thereby sacrificing anything morally significant." Could one reasonably deny this? One could insist that all interests are morally significant in the sense that they count for something, so you always have at least some reason to make any consumer purchase that would bring you the slightest bit of extra happiness. But of course Singer does not mean to deny this. His use of "significant" here is not meant to distinguish interests from non-interests (that count for literally zero), but rather to distinguish especially weighty or important interests from relatively trivial ones. And it cannot plausibly be denied that some of our consumer purchases are relatively trivial, or not especially important to our lives.

It's an interesting question precisely how to distinguish significant interests from comparatively trivial ones. The two extremes seem intuitively clear enough: luxury goods like designer clothes seem fairly unimportant, while providing a good life for one's own child is obviously of genuine importance. In intermediate cases where it is unclear whether an interest qualifies as deeply "morally significant", it will be similarly unclear whether Singer's argument requires us to be willing to sacrifice that interest in order to prevent grave harm.⁷ But it's important to note that an argument can be sound and practically important even if it is sometimes unclear how to apply it.

Premise 4: Singer's Rescue Principle

Finally, we come to Singer's Rescue Principle: If it is in our power to prevent something very bad from happening, without thereby sacrificing anything morally significant, we ought, morally, to do it. If we are to reject the argument's conclusion, we must reject this premise. But can you really believe that it is morally okay to just sit back and watch something terrible happen, when you could easily (without sacrificing anything important) prevent it?

Some may claim that our only duty is to do no harm.⁸ On this view, it would be wrong to steal from the global poor, and it would be generous to help them, but we have no obligation to help in any way—it is never wrong to simply mind one's own business. This minimal view of morality (as limited to the duty not to harm others) meshes nicely with common views about charity. But it turns out to be unacceptable when we consider a broader range of cases, as Singer brings out with his famous drowning child thought experiment.

The Drowning Child

Singer writes:

If I am walking past a shallow pond and see a child drowning in it, I ought to wade in and pull the child out. This will mean getting my clothes muddy, but this is insignificant, while the death of the child would presumably be a very bad thing.⁹

In a case like this, when you can easily prevent something very bad (like a child's death), it seems clear that doing so is not only morally praiseworthy, but morally required. That remains true even if saving the child comes at a cost to yourself, as long as the cost is insignificant in comparison to the value of the child’s life. Since the cost of ruining your clothes (even an expensive suit that costs several thousand dollars to replace) is insignificant compared to the child’s life, you ought to wade into the pond to save the child.

What does the minimal “do no harm” view of morality imply about the right way to act in the thought experiment? Well, it is not your fault that the child is drowning; you did not push them in. If you were to walk by and let the child drown, you would not be causing any additional harm—they would be just as badly off if you were not there in the first place. So the minimal view implies that it would be morally fine for you to just walk by (or even sit and eat some popcorn while watching the child drown). But that strikes most people as obscenely immoral. So the minimalist's response to Singer's rescue principle fails. 

If applied consistently, Singer’s principle has radical implications in the real world. He writes that:

We are all in that situation of the person passing the shallow pond: we can all save lives of people, both children and adults, who would otherwise die, and we can do so at a very small cost to us: the cost of a new…shirt or a night out at a restaurant or concert, can mean the difference between life and death to more than one person somewhere in the world.¹⁰

If you accept that saving the drowning child is morally required, even at the cost of ruining your expensive suit, then morality may equally require you to donate an equivalent amount of money to save a child’s life via other means. And you can save a child's life, just by donating a few thousand dollars to GiveWell's top charities. (Or you can save a quality-adjusted life year by donating $100 or so.) If you would not think it okay to let a child drown, moral consistency requires that you likewise refuse to let children die unnecessarily from poverty or preventable disease.

Aside from helping to address objections to Singer's rescue principle, the drowning child thought experiment clarifies which of our interests are sufficiently "insignificant" that we may be called upon to sacrifice them to prevent grave harm to others. For example, someone might initially think that wearing designer clothes is a vital part of their identity, but if pressed on whether they would sooner watch a child drown than give up this expensive lifestyle, they might change their mind.¹¹

Objections

Not everyone accepts Singer’s radical conclusion about our moral obligations to donate to those in need. Yet, Singer’s conclusion is difficult to avoid since the standard objections no longer seem plausible when applied to corresponding variations of the pond case.¹² For example:

(1) You may believe charitable donations are uncertain to help. Would that remove the moral requirement to donate? But suppose you are similarly uncertain about whether the child in the pond is truly drowning (maybe they are just playing a game). Even so, mere uncertainty does not justify doing nothing. So long as the chance that your actions would help is sufficiently high (relative to the costs or any associated risks from attempting aid), you may still be required to wade in and offer assistance, just in case. Similarly, uncertainty about the impacts of your donations does not justify keeping the money to yourself, as long as the expected value of your donations is sufficiently high (relative to the costs).

(2) What if other, wealthier people could give more instead? Surely, they are under an even greater moral obligation to give, since doing so is less of a sacrifice for them. But suppose that other people stand around the pond, watching the child drown but refusing to help. They ought to help, so ideally your help would not be needed. But given that they are not helping, so your help is needed, it sure seems like it would still be wrong for you to do nothing and let the child drown. Likewise, it would be wrong not to save others’ lives by donating, even if there are more affluent people who could help but refuse to do so. (The bystander effect suggests that waiting for others to help first could easily result in no-one helping at all.)

(3) What about geographical proximity as a factor? Does it make a moral difference that the drowning child is right in front of you whereas the beneficiaries of your donations are far away? Yet, imagine that the pond with the drowning child was actually located far away from you, say, in another country, and you could rescue the child’s life by simply pressing a button. Surely, you would be required to press the button, even if you had to pay some money to do so. What matters morally is your ability to prevent the child from dying at a low cost to yourself, so that is what you should do—regardless of how far away the child is. Many moral theories (including utilitarianism) explicitly deny that geographical proximity is inherently morally relevant.¹³

(4) Or, perhaps you think it is enough to do your "fair share": to just give as much as would be needed if everyone else did the same (perhaps 5% of one's income). But suppose that after saving one child from drowning, you notice three other drowning children. Two bystanders are just watching the children drown, though you are relieved that one other adult is on track to save two of the remaining three children. Would it be okay to watch the last child drown on the grounds that you have already done the "share" (saving one out of four) that would have sufficed if everyone had done likewise? Or should you step up and do the share that is required to actually save all the children given what others are—and are not—doing?

It is unfair when some do not do their share. It is unfairly demanding on us to have to do more than our ideal share would have been. But it would be even more unfair on the child to just let them drown. Losing their life would be a far greater burden than the extra cost to us of helping more. So, while some unfairness is inevitable when some do not do their share, concern to minimize unfairness should still lead us to step up and do more when needed.

None of these responses seems successful in establishing a morally important difference between the drowning child thought experiment and charitable giving. But considerations of salience, repeatability, or emergency may prove more significant. We address these in the next three sections.

Salience

Although geographical distance by itself does not seem to make a moral difference, it may make a psychological difference to us by affecting the salience of the different needs at stake. The visible suffering of a child right before our eyes has a very different emotional impact than merely abstract knowledge of distant suffering. And this difference in emotional impact plausibly explains why most of us would be so much more strongly motivated to save the drowning child than to relieve distant suffering by donating to charity. But what is the moral significance of this difference in psychological salience?

Plausibly, greater salience can help to bring to our attention genuine reasons to act that are there regardless, but that we might otherwise mistakenly neglect. After all, it is not as though a suffering child suddenly becomes objectively more important once they enter our visual field. But we certainly become more aware of them (and how vitally important it is to help them). If this is right, it seems there is just as much moral reason to help those in need who are far away; we just tend not to notice this so much, and so we (understandably) make the moral mistake of failing to do as much as is objectively warranted in order to aid them.¹⁴

On this analysis, the difference in salience does not affect the strength of our moral reasons—it is just as important to save a distant child as it is to save one right before our eyes. But it does make an important difference to how we should evaluate the failure to act. Intuitively, failing to save the child from drowning would be morally monstrous, whereas failing to donate does not reflect so badly on you, even if it is a serious moral mistake. We can explain this difference in terms of one's quality of will. One is blameworthy to the extent that one acts from malicious motivations, or acts in a way that reveals an egregious lack of concern for others. To neglect more salient suffering reveals a greater lack of altruistic concern, even holding fixed the magnitude of the suffering in each case. So it is more blameworthy. As Chappell & Yetter-Chappell put it: "A child drowning before our eyes shocks us out of complacency, activating whatever altruistic concern we may have, whereas the constant suffering of the global poor is easier to ignore, meaning that inaction does not necessarily imply [such] an egregious lack of concern."¹⁵

We can thus accommodate the intuition that failing to donate to effective charities is not as blameworthy as watching a child drown (since only the latter reveals an extreme lack of altruistic concern), without this providing any reason to deny that aid in either case may be equally morally important.

Repeatability

A notable difference between the two cases is that it is very rare to come across drowning children, whereas the needs of the global poor are constant and unrelenting. The significance of this fact is that a policy of helping those nearby in need of direct rescue would not be expected to prove especially costly. But a policy of helping anyone in the world in desperate need of aid would soon take over your life. A better analogy would then seem to be a limitless line of ponds containing drowning children. And when we consider such a case, it may no longer seem so wrong to at least sometimes take a break, and thereby let a child drown.¹⁶

Of course, to save as many lives as possible over the long term, it would likely be optimal to take strategic breaks for self-care. At a minimum, you need to eat and sleep. But you may also save more lives in the long run if you take care to avoid burnout, taking extra breaks to spend time with friends and pursue hobbies that help you to de-stress. If so, taking such strategic breaks is morally justified by Singer's principles. (There is no virtue in being counter-productively self-sacrificial in one's altruism.) So this is not yet a counterexample to Singer’s view.

Still, this optimal route is highly demanding since it involves significant personal sacrifice. Suppose that in order to save the most lives you had to forsake your plans to become a parent, and cut down time spent with friends and hobbies to the bare minimum required to maintain your sanity and productivity. That is a big ask, and one that involves the loss of many morally significant goods in life. If Singer's principle required us to pursue this optimal route, it might not seem so plausible after all.

But does it require this? Unlike maximizing utilitarianism, P4 only asks us to give up things that are not morally significant. Since the above sacrifices are clearly morally significant, it seems that they would be excluded from the list of P4's possible demands.¹⁷

One difficulty is that it is not immediately clear how to apply the rescue principle to cases of repeated actions. Consider: giving up any one second of life might seem trivial. But repeated enough times, you would eventually give up your entire life, which is certainly significant. This suggests that repeatedly making an insignificant sacrifice might add up to an extremely significant sacrifice. To apply the rescue principle sensibly, then, it is not enough to ask whether the immediate sacrifice in isolation is morally significant. We must further ask whether it is part of a pattern that, in context, adds up to a morally significant sacrifice. If interpreted in this way, Singer's rescue principle would seem to allow broad leeway for reserving substantial time and resources to pursue the personal projects that are most important to us.¹⁸

Still, the conclusion of Singer's argument remains strikingly revisionary. Even if we may reserve the majority of our spare time and resources for personal projects, we are still required to do much more for others than almost any of us actually do. Even when we need not entirely give up some expensive (or time-intensive) pastime, we may be morally required to economize, if by doing so we could (perhaps over several years) save many lives without significant lifetime loss to our own well-being. Many hobbies plausibly exhibit diminishing marginal utility: the more time and money we plow into them, the less additional value we gain from further investment. In such cases, we may be able to cut our personal investment by, say, half while still retaining most of the well-being we gain from the hobby. And of course many of us also spend time and money on entirely frivolous things that, upon reflection, do not significantly contribute to our lives at all. If we reflect carefully and honestly, most of us would likely find significant opportunities to do more to help others, without needing to sacrifice anything truly important. If Singer's rescue principle is right—and it seems hard to deny—then we really ought to pursue these opportunities.

Emergencies

The last major challenge to Singer's argument comes from the idea that special ethical norms apply in emergency cases that cannot be broadly generalized. The drowning child scenario is a paradigmatic emergency. So perhaps common sense could be restored by combining a minimal view of our everyday obligations with ambitious positive obligations to assist in cases of emergency?

The difficulty for this view is to provide it with a principled basis. Why should emergency deaths be treated as inherently more important than equally preventable deaths from ongoing causes?

Sterri and Moen propose to explain this in terms of an "Informal-Insurance Model".¹⁹ Their basic idea is that emergency ethics can be understood as a mutually-beneficial agreement amongst all in the moral community to informally insure each other against rare, unexpected risks of grave harm. That is, we undertake to help others out in emergency situations, on the understanding that they would do the same for us. Since emergencies are rare, the comfortably-off can agree to participate in such a scheme without expecting to be bled dry by all the world's needs. And since emergency situations can befall anyone, it is in their enlightened self-interest to do so. We are all better off informally insuring each other against disaster in this way, than if we all were left to fend for ourselves.

This line of argument faces two significant problems, common to efforts to ground ethics in enlightened self-interest. Firstly, the underlying logic of mutual benefit excludes from the moral community not just the global poor but also others (including infants, non-human animals, future generations, and the severely disabled) who are not in a position to reciprocate. But surely you still ought to rescue a drowning paraplegic, for example, even if he could not do the same for you.

The second problem is that even when the informal insurance model gets the right result (requiring that you help), it does so for the wrong reasons. It implies that you should help for the sake of playing your part in a co-operative scheme of mutual benefit, which does not seem remotely the right reason to save a child from drowning.

To see this, imagine extending the logic of the informal insurance model to a society including water-phobic robots who just want to collect paperclips but occasionally drop them in puddles. In order to secure the assistance of the robots in helping to free us from getting our feet caught on railroad tracks (or other non-water-related emergencies), we might reciprocate by rescuing their lost paperclips from puddles. If the informal insurance account of emergency ethics were correct, then your moral reason to save a drowning child would be of exactly the same kind as your reason to "save" a paperclip from a puddle in the imagined scenario. But this is clearly wrong. We have moral reasons to save lives and avert great harms for the sake of the affected individuals. These moral reasons are distinct from (and more important than) our reasons to participate in mutual-benefit schemes.

Conclusion

Singer identifies a logical tension in our ordinary moral thought. We tend not to think much about our power to prevent great suffering (and even save lives). Even when this fact is brought to our attention, we tend to assume that it is morally okay for us not to act on it (or to do little). Helping would be generous, we think, but not required.

However, Singer's rescue principle seems undeniable: if we can easily prevent something very bad—that is, without giving up anything morally significant—it sure seems that we ought to do so. And the drowning child scenario verifies this principle: we would not think it okay to just watch a child drown when you could easily save them at no risk to yourself. Differences in salience may explain why we find it easier to ignore more distant suffering; but it would also seem to suggest that we are morally mistaken to do so.

Considering repeatability means that we need to take our overall patterns of response into account: sacrifices that are small in isolation may add up to extreme sacrifices that are more than Singer's principle would require. But even taking this into account, there are likely to be many changes we could make to our lives in order to help others more, without causing any significant loss to our own well-being. If Singer is right, we are morally required to make these changes. It is no less important than saving a child who is drowning right before our eyes.

Discussion Questions

Many effective altruists now believe that you can do more good through pursuing a high-impact career than by donating (even generously) whilst working at a less impactful job. How does that affect your view of Singer's argument? Could you be morally required to consider a career change? Should someone in a high-impact career be expected to donate to charity in addition?

This study guide focuses on the more moderate version of Singer's rescue principle. But he also defends a stronger version, according to which we are morally required to prevent bad things from happening whenever we can do so without sacrificing anything of comparable moral significance. How much of a difference do you see between these two versions of the principle? Do you think the stronger principle is correct?

What would a scalar utilitarian think of Singer's principles? If there is no such thing as obligations, but just better and worse actions, how would that affect Singer's argument? Would it become any less important or worthwhile to save lives if it was no longer "obligatory" in addition? What do you think is added by saying that an act is (not only good but also) "obligatory"?

Imagine that you are going to donate money to an effective charity—enough to save two lives. But along the way, you see a child drowning in a pond. There is no time to set aside the cash in your pockets: if you jump in, the money will be destroyed, so you will be unable to make the donation after all. Should you still save the drowning child? Why / why not?

We tend to just think about the money or resources that people already have. But suppose that you could easily earn more, say by working overtime (or shifting to a more lucrative job). Might it be wrong to not earn more money (in order to then donate more)? How would you apply Singer's principles to this case?

Suppose that a wealthy friend tends to get defensive when confronted with moral challenges. Currently they donate a little bit to help others, but you know that if you told them about Singer's argument, they would clam up and stop donating altogether.

Would it be wrong for you to tell them about Singer's argument?

If so, would that mean that Singer's conclusion is false, and they are not obliged to donate more after all? Or could a moral claim be true even if it was not always a good idea to tell people about it?