The philosophy of unintended consequences
What is the best moral framework to use in deciding how to do good?
This was the central question in a recent series of episodes of Hidden Brain, a podcast dedicated to discussing what drives human behaviours.
In one of the episodes, they interviewed the philosopher Peter Singer, perhaps most known for his controversial views on morality. Singer argues that the best decision-making guide on how to do the most good is not feelings or emotions, but instead logic. He is a utilitarian who believes that we may be able to calculate the lives we can save.
His famous thought experiment goes like this: if you are wearing an expensive suit — say, worth $200 — on your way to an important occasion, and you happen to pass by a pond in which there is a drowning child, would you choose to save the child at the expense of ruining your suit?
One can imagine what most people would say.
But this moral calculation is made more difficult when it’s not just a suit that is at stake. The now-viral trolley problem illustrates this: would you choose to kill one man in order to save five? And in the other episode of the Hidden Brain series, they make it even harder: what if that one person you’d have to kill is your loved one?
Singer’s brand of utilitarianism, which argues that the moral thing to do is sacrifice your loved one to save five other people, is compelling. At the heart of it is a belief that is both elegantly simple but also incredibly radical: all lives are equal.
In fact it is so compelling that it has led to the growth of a movement called Effective Altruism (EA), whose vision is to do the most good in the world in large part through performing similar moral calculations as above. Their empirical and often quantitative approach in determining how “most good” looks like has led to the emergence of organisations like GiveWell, who calculate the effectiveness of non-profit organisations mostly via cost per lives saved.
Applied to international development and humanitarianism, such a framework can be powerful arguments for supporting the interventions of organisations like GiveDirectly, whose programme of unconditional cash transfers has been credibly shown to alleviate poverty.
I have a deep appreciation for this type of utilitarianism and for EA. Their unflinching belief that all lives are equal — no matter whether it’s your cousin or a boy in Cote D’Ivoire — and their insistence on rigorous methodologies to demonstrate impact and cost-effectiveness is something to be admired.
Many of their critics might say they are too logical, and are unable to account for human realities and human emotions. When asked in the podcast whether he would sacrifice his girlfriend to save five other people, Dylan Matthews, who is a journalist and a prominent Effective Altruist, said he would. Offering an alternative to this kind of seemingly heartless moral framework, Harvard psychologist Josh Greene says it’s better to exercise “deep pragmatism,” which encourages doing the most good according to a utilitarian calculation, but stops short at advocating for the sacrifice of our loved ones.
My main frustration with Singer’s and Matthews’s moral framework, however, is not its emotional lacuna, but its deficiency in logic.
The podcast talked about Singer’s views on how the US is handling the coronavirus crisis. Singer wrote in an article:
“It pains us to say it, but US President Donald Trump is right: ‘We cannot let the cure be worse than the disease.’ Lockdowns have health benefits: fewer will die of COVID-19, as well as other transmissible diseases. But they have real social and economic costs: social isolation, unemployment, and widespread bankruptcies, to name three. These ills are not yet fully apparent, but they soon will be. Some people insist that there is, in practice, no trade-off: lockdowns are better for saving lives and the economy. This seems to be wishful thinking.”
I actually agree.
The concept of trade-offs and unintended consequences is not new to the development and humanitarian sector. One of the best scholarly works around this is the book Imposing Aid by Barbara Harrell-Bond, which investigates the negative unintended consequences of humanitarian assistance to Ugandan refugees. The book was published in 1986.
Like the US government’s intervention to enforce a lockdown to prevent the spread of coronavirus, the international aid sector’s interventions — typically carried out by rich states, the UN, and non-governmental organisations — to, for instance, stamp our poverty or provide relief to those affected by disasters in low-income settings carry the risk of unintended adverse consequences.
There is, in fact, a rich body of literature around this — a comprehensive summary of which may be found in Harvard economist Nathan Nunn’s article, “Restraining Ourselves: Helping by not hurting.” Within it Nunn writes a devastating conclusion about such interventions:
“Overall, it is unclear whether interactions with the West are, on the whole, helpful or detrimental to developing countries. We may have our largest and most positive effects on alleviating global poverty if we focus on restraining ourselves from actively harming less-developed countries rather than focusing our efforts on fixing them.”
Let’s take a look at a more concrete example: health care. In a recently published paper, Northwestern University economist Erika Deserranno, Yale economist Nancy Qian, and researcher for the organisation BRAC Aisha Nansamba show how:
“…in poor rural communities where government workers provide basic health services, the entry of an NGO that aims to provide similar services reduces the supply of government workers and total services….The decline in health services is most pronounced in villages where the NGO hires the government worker, and is accompanied by an increase in infant mortality.”
In this case while NGOs’ interventions initially resulted into lives saved now, it also led to lives lost later.
Some of the most “effective” charities according to GiveWell are Against Malaria Foundation and Deworm the World, both of which provide assistance to households to improve their health outcomes. But in the same way that Singer pointed out how lockdown interventions may have unintended consequences, is it possible that these health interventions could also have negative unintended consequences, like weakening a low-income country’s health system?
Here lies the problem with an Effective Altruist approach to development and humanitarian aid. I call it epistemic arrogance (in contrast to epistemic humility).
Epistemic arrogance is the illusion of certainty and perfect knowledge of the full consequences of one’s actions. It’s the mistake of thinking that one can easily account for all possible variables in calculating how to do the most good. And it’s the assumption that, once the numbers are crunched and the solution is implemented, the problem goes away and that’s the end of it.
In medicine its relatively more straightforward (but still incredibly difficult and itself full of ethical conundrums) to come up with a formula to calculate for utility and effectiveness. Utility or “lives saved” can be measured in Quality Adjusted Life Years (QALYs). This provides medical professionals with a framework in deciding, for instance, to which coronavirus patient they would give the remaining ventilator in a hospital — an 85 year old man with underlying health conditions, or an otherwise healthy 30 year old woman? And effectiveness, say, of a vaccine can be tested via a randomised control trial (RCT).
Meanwhile in aid calculating for utility and effectiveness is messier. If all lives are equal, which one holds more utility — the lives saved now, or the lives saved later? (This is the same line of argument used by those in the environmental justice movement.) And, sure, RCTs — which many also uphold as the gold standard of evaluating aid programmes’ effectiveness — have significantly contributed to discovering high impact interventions. But given the evidence of how interventions, in the bigger picture, could do more harm than good, isn’t it worth asking the same question Singer posed about the lockdown for coronavirus?
That is, in the context of imperfect knowledge of our action’s full consequences, could the cure be worse than the disease?
I suspect epistemic arrogance is in part caused by seeing the world through the lens of experiments — whether its an actual experiment like RCTs or a philosopher’s thought experiment. In the podcast Singer provides an example to illustrate the rationale behind his brand of utilitarianism: suppose in a town where majority of residents are from group Z, someone from a minority group X commits a crime — although there is no evidence as to who exactly did it. Out of their anger, a mob from group Z wants to kill five people from group X. Singer argues that from a utilitarian perspective, it’s more moral to sacrifice one person from group X (even if they are innocent) rather than have five people killed.
In a thought experiment, that could be a reasonable answer. But we don’t live inside a thought experiment. We live in the real world that is ever changing and dynamic. If that one innocent person from group X is killed, the world doesn’t stop there. There will be reactions to the action, and then reactions to the reaction, ad infinitum — just like how there are unintended consequences to seemingly benign and moral actions, such as enforcing a lockdown or providing aid. And once you realise that Singer’s example is, in fact, about how white Americans lynched Black Americans in the past, it’s even more disagreeable — particularly at a time when ‘reactions to the reactions’ are now culminating in a global protest against violence towards Black people.
Singer says logic is a lot better than feelings or emotions in determining what is the moral thing to do. He argues that using feelings or emotions in moral decision-making arose out of our evolutionary drive to survive in pre-historic times but are now outdated in the modern world (e.g., caring for someone in your tribe more than someone outside of it).
There are still many people who believe in the infallibility of such calculative approaches in aid — and they include not only utilitarians and Effective Altruists, but also aid workers who so resolutely believe in the effectiveness of their own programmes. (Not all utilitarians, Effective Altruists or aid workers, believe this, of course, but many do.) Could it be that such people are suffering from the very cognitive bias that Singer pointed out? Are they led to uphold their perspective because of feelings or emotions (e.g., the warm fuzzies they get with the sense of belonging to a particular community, or the pride they could plausibly feel for having what Matthews describes as an unusual view compared to mainstream norms), instead of logic and evidence?
In the podcast Matthews discussed about how his moral framework has led him to decide to donate his kidney to a stranger. In fact many within the EA movement advocate for this (“Donating a kidney gets ~14 QALYs at a cost of something on the order of ~0.1 QALYs,” states a post in the EA forum). For this act, Matthews and other kidney donors were invited to a fancy gala and presented with an “American Hero” award. At their table comprised mostly of men, they started talking about how they’re now thinking of donating their liver. “It felt like skateboarders talking to someone who is trying a new trick they’ve never tried before,” Matthews says, “That’s sweet man!”
I can’t help but sense a bit of hubris in their exchange. After all, they’ve just been given the title “American Hero.” But I see a similar kind of hubris in aid — in the form of “white saviourism” that prefers to leap into action to save the day, instead of choosing to act carefully at the risk of causing unintended consequences.
At least in the case of Matthews and his fellow kidney donors, their intervention surely saved lives. Those with a white saviour mentality, on the other hand, along with aid workers who fail to deeply reflect on the risks carried by their development and humanitarian programmes, could use a little more epistemic humility in deciding whether or not to intervene, lest the cure be worse than the disease.
Aid Re-imagined’s mission is to help usher the evolution of aid towards effectiveness and justice through deep, radical, and evidence-based reflection and research — unafraid to venture beyond the realm of development and humanitarianism, using insights from philosophy, economics, politics, anthropology and sociology, as well as management. Aid Re-imagined stands for a more effective and just aid for our new, ever-changing world.
