A better formalism for interpreting confidence intervals

June 10, 2017

When we take a sample mean, we should think of it as a random variable, and our measured sample mean as a realisation of that random variable. The sample mean is a random variable because it is the result of random sampling. Repeated sampling involves observing repeated realisations of the random variable.

We should think of confidence intervals around this mean as realisations of a random interval, an interval whose bounds are random variables rather than real numbers. This is an attractive formalism because it resolves many confusions around the interpretation of confidence intervals.

Suppose the true population mean is the number . The mean of a random sample from this population is the random variable . Then the random interval

has an approximately 95% probability of containing .

Suppose in our sample takes the realisation and takes the realisation . So an instance of the above random interval is the confidence interval:

The confidence interval either contains or does not contain .

In full, my proposed interpretation schema is:

,

is a realisation of

,

and the probability

.

This formalism has several advantages:

  • robustness: distinguishing random intervals from confidence intervals means it’s much harder to get confused into making an incorrect probabilistic statement about the non-probabilistic object .
  • parsiomy: we express everything we want using only probabilities, random variables, and intervals, three well-understood notions.
  • relevance: our interpretation only involves the objects we actually have (a random interval and a confidence interval). We need not make reference to (hypothetical) repeated sampling.

The ugly and the bad

Unfortunately, my preferred formalism does not appear to be popular. Let me show some of the alternatives I have seen and explain their downsides and how my proposal does better.

1

Oxford department of statistics:

The interval is random, not the parameter. Thus, we talk of the probability of the interval containing the parameter, not the probability of the parameter lying in the interval.

This is the worst example, and is admittedly rarely seen in print. But in speech I’ve seen it used often, even by academics who were trying to explain the correct interpretation of confidence intervals! The problem with this of course is that once you write it down in mathematical language, the probability of the interval containing the parameter is exactly the same object as the probability of the parameter lying in the interval. In our example it is simply . It is equal to 1 or 0.

2

Quantitative Economics lecture notes for Oxford undergraduates:

“Were this procedure to be repeated on multiple samples, the calculated confidence interval (which would differ for each sample) would encompass the true population parameter 95% of the time.”

I don’t like this because:

  • It invokes the clunky counterfactual “were this procedure to be repeated”. What if it’s impossible to take repeated samples? We still want to be able to make statements about our confidence interval.
  • It doesn’t have a clear mathematical formalisation. how do I write “95% of the time” in terms of probabilities?
  • The actual confidence interval we have is nowhere mentioned. For what is supposed to be an interpretation of that object, that’s a little confusing.

My formalism solves these three problems.

3

Wikipedia:

“There is a 90% probability that the calculated confidence interval from some future experiment encompasses the true value of the population parameter.”

Similar complaint here: why do we need to refer to future experiments? We want an interpretation of the confidence interval we actually have.

4

Harvard University:

For this reason, for a 95% CI, we say that we have 95% confidence that the interval will cover the true population mean. We use the term ‘confidence’ instead of probability because although the sample mean is random, the single interval we calculate is fixed. We also cannot talk about the probability that the population mean will lie within a certain interval, since it is also fixed.

This needlessly introduces the new concept of ‘confidence’, which is bound to cause confusion. It’s much better to use probabilities, a concept we already understand and for which we have a formal notation.

Consistent Vegetarianism and the Suffering of Wild Animals - Journal of Practical Ethics

May 25, 2017

A revised version of the essay I wrote for the Uehiro Prize has been published in the Journal of Practical Ethics.

Abstract:

Ethical consequentialist vegetarians believe that farmed animals have lives that are worse than non-existence. In this paper, I sketch out an argument that wild animals have worse lives than farmed animals, and that consistent vegetarians should therefore reduce the number of wild animals as a top priority. I consider objections to the argument, and discuss which courses of action are open to those who accept the argument.

Qu'est-ce que l'altruisme efficace ?

July 24, 2016

Evénement de lancement de l’association Altruisme Efficace France, à Paris le 5 juillet 2016

Qu’est-ce que l’altruisme efficace ?

Quelles sont les meilleures façons d’aider les autres ? Bien entendu, cette question fait débat. Y répondre avec sérieux requiert des prises de positions morales, et des recherches empiriques approfondies. Les désaccords sont inévitables, et ils sont légitimes. Mais ce n’est pas pour autant qu’une affaire d’opinion. L’altruisme efficace est basé sur l’idée que toutes les manières d’aider les autres ne se valent pas.

Cette volonté de dépasser le domaine de l’opinion est tout à fait compatible avec une appréciation de la difficulté de la tâche et de l’incertitude de nos hypothèses. Une analogie avec la science pourrait être utile ici. Dans les sciences naturelles, les chercheurs sont souvent en désaccord. Cependant, malgré cette diversité de points de vue, il y a aussi beaucoup de consensus : il est démontré que certaines propositions sont incorrectes. Et au fur et à mesure que la communauté scientifique continue de récolter plus de données et de les analyser de manière critique, elle rejette les mauvaises théories, affine ses hypothèses, et progresse.

C’est ainsi que nous concevons l’altruisme efficace - une tentative de répondre à une question difficile, à laquelle nous pourrons apporter des réponses de plus en plus précises à mesure que nous y investissons plus d’effort. L’altruisme efficace, c’est avant tout cette question, et un engagement à tenter d’y répondre de manière rationnelle et scientifiquement argumentée. Ainsi, l’altruisme efficace est une démarche qu’il convient de séparer des conclusions particulières auxquelles l’on peut arriver en la suivant.

Enfin, l’altruisme efficace n’est pas qu’une analyse académique. Nous souhaitons identifier la meilleure manière d’aider, puis passer à l’action en la mettant en oeuvre.

Comment trouver une cause efficace ?

Parmi tous les moyens d’aider les autres, comment trouver celui qui permettrait d’avoir le plus grand impact ? Nous tentons de répondre à cette question en suivant certains principes fondateurs.

Faire le meilleur usage de nos ressources

Nos ressources sont limitées. Quelle que soit notre générosité, il est impossible de résoudre immédiatement tous les problèmes constatés dans le monde. Notre situation est similaire à celle d’un médecin en zone de guerre : il y a des centaines de blessés, mais le personnel médical est limité. Il met alors en place un système de triage médical, qui assigne un degré de priorité à chaque blessé, dans le but de sauver le maximum de victimes. De même, nous avons chacun une quantité limitée de temps et d’argent, et nous ne pouvons pas aider toutes les victimes de souffrances dans le monde. La meilleure chose à faire est alors de se concentrer sur les actions qui aideront le plus possible, celles qui auront le plus grand impact par euro donné ou heure investie.

Empathie Globale

Prendre en considération toute vie consciente, sans exclusion liée à l’appartenance à un groupe donné, qu’il soit fondé sur la nationalité, l’ethnie, la croyance ou l’espèce..

Dans un des textes majeurs de la philosophie morale contemporaine, Famine, Affluence and Morality[1], Peter Singer nous invite à étendre notre cercle d’empathie au-delà de notre propre pays et à reconnaître qu’une vie vaut la même chose, qu’elle soit vécue dans un pays développé ou un pays en développement. Nous souhaitons aider ceux auxquels il est possible d’apporter le plus grand bénéfice, et s’il s’agit d’habitants des pays en développement au lieu de nos compatriotes, c’est cela qu’il faut préférer. La plupart des personnes engagées dans l’altruisme efficace étendent ce raisonnement jusqu’à son aboutissement logique, qui est de prendre en considération les vies non-humaines ainsi que les vies futures de ceux qui ne sont pas encore nés. Certains se concentrent donc sur la souffrance des animaux d’élevage ou tentent d’améliorer la trajectoire future de l’humanité à très long terme.

Ouverture d'esprit et "agnosticisme de cause"

Considérer toutes les causes et actions possibles, puis agir de la manière qui produit le plus grand impact positif.

Il est habituel de commencer par choisir une cause pour des raisons personnelles ou émotionnelles, puis de se demander quelle serait l’action la plus efficace au sein de cette cause. (“Je souhaite soutenir la recherche contre le cancer car mon père est mort d’une tumeur cérébrale. Quel est le meilleur organisme de recherche contre le cancer ?”) Mais dans l’altruisme efficace, le choix de la cause lui-même fait l’objet d’une recherche rationnelle, car les plus grandes différences d’efficacité se trouvent entre différentes causes et non au sein d’une cause. Il s’agit donc de partir d’une position “agnostique à la cause” afin de pouvoir choisir la plus efficace parmi toutes les causes possibles, et de pouvoir changer de cause en réaction à de nouvelles informations.

Se concentrer sur l'action la plus efficace

Le choix optimal est probablement 10 ou 100 fois supérieur à la moyenne.

La taille chez les humains suit une distribution normale : les plus grands humains ont une taille au plus 60% supérieure à la moyenne. Par contre, le classement de popularité des sites web suit une loi de distribution très asymétrique (loi de puissance) : Google, le site le plus populaire, reçoit des milliards de visiteurs, alors que la vaste majorité des sites sont très peu visionnés. Certains éléments de preuve suggèrent que les actions altruistes suivent, comme la popularité des sites web, une distribution très asymétrique. En utilisant la base de données DCP2, le philosophe Toby Ord a montré que dans le domaine de la santé, le rapport coût-efficacité de différentes interventions suit une distribution extrême : l’intervention la plus efficace produit 15 000 fois plus de bénéfice que la moins efficace, et 60 fois plus que l’intervention médiane[2]. Au-delà de la base DCP2, les interventions de santé les plus efficaces sont particulièrement exceptionnelles : l’éradication de la variole en 1979 a prévenu plus de 100 millions de morts, pour un coût de 400 millions de dollars.

Ainsi, il apparaît probable que la meilleure action possible soit non pas 30% plus efficace ou 3 fois plus efficace, mais bien 10, 100 ou même 1000 fois plus efficace que la moyenne. Il est donc essentiel de concentrer nos ressources non pas seulement sur une action très efficace, mais sur celle qui est la plus efficace. Cela ne veut pas dire que tous ceux qui agissent selon ce principe doivent forcément travailler au service de la même cause. Il existe bien sûr des incertitudes considérables et des différences de valeurs, qui amènent différentes personnes à choisir des projets différents.

Privilégier les causes indûment négligées par d'autres

Pour déterminer l’impact d’une action, il faut appliquer un raisonnement à la marge : quel est l’effet supplémentaire de l’effort que j’apporte ? Il ne suffit pas d’observer l’effet moyen d’une action, il faut considérer son effet marginal. Imaginez que le grenier d’un village ne prenne feu. Les villageois ont le temps de sauver des flammes quelques réserves de nourriture avant que le feu n’engloutisse tout le bâtiment. Le plus important est de sauver le blé qui permettra de survivre l’hiver. Pourtant, si tous les autres villageois se concentrent sur les sacs de blé, ils réussiront à en sortir assez. Plutôt que de prendre encore plus de blé, votre meilleure action altruiste serait de sortir le sac de sel. Dans cet exemple, le sel a plus de valeur à la marge, malgré le fait qu’en moyenne le blé soit essentiel à la survie. De même, si l’on investit dans une cause négligée, le bénéfice marginal de chaque action sera supérieur, car les meilleures opportunités d’aider les autres n’auront pas encore été exploitées. Au contraire, il sera plus difficile d’avoir un grand impact dans une cause qui reçoit déjà beaucoup de ressources. Par exemple, apporter des vaccinations de base à ceux qui en ont besoin est extrêmement efficace : de nombreux cas de maladies graves peuvent être évités à faible coût. Mais soutenir ces vaccinations n’est généralement pas conseillé pour un donateur individuel, car les besoins les plus essentiels sont déjà couverts par les gouvernements ou les grandes fondations.[3] Si par contre une cause est négligée, cela peut donc constituer un indice en sa faveur.

Prendre en compte ce qui adviendrait si vous n'agissiez pas

Il ne suffit pas de considérer les effets directs de votre action, il faut prendre en compte son effet contra-factuel, c’est à dire la différence entre les conséquences réalisées suite à votre action et les conséquences qui seraient advenues si vous n’aviez pas agi. Imaginez que quelqu’un subisse un arrêt cardiaque devant vous. Vous venez de recevoir un entraînement de secourisme, et vous avez tant de zèle à aider que vous poussez de côté le médecin qui s’apprêtait à s’occuper de la victime. Vous parvenez à réanimer la victime, mais vous n’avez pas pour autant sauvé sa vie. Si vous n’étiez pas intervenu, le médecin l’aurait fait. Votre action était donc remplaçable, elle n’a pas eu d’impact contra-factuel. Ce raisonnement est essentiel pour évaluer rigoureusement nos actions altruistes : par exemple, devenir enseignant bénévolement dans une école d’un pays en développement pourrait avoir un impact négatif au lieu de positif, si le bénévole prenait la place d’un enseignant local expérimenté qui aurait fait un meilleur travail.

Mettre en balance la probabilité de succès avec la taille du bénéfice

La plupart de nos actions n’apportent pas un bénéfice garanti, mais augmentent la probabilité qu’un certain bénéfice advienne. Par exemple, militer contre les conditions de l’élevage intensif n’a aucune garantie de modifier le comportement des éleveurs ou d’inciter un gouvernement à légiférer, mais rend ces éventualités plus probables. Il est tout aussi fallacieux de choisir des objectifs très importants sans penser à la probabilité de les atteindre, que de rejeter toutes les actions à faible probabilité de succès sans prendre en compte la taille de ce succès potentiel. Il s’agit plutôt d’estimer l’espérance mathématique de notre action, en multipliant la probabilité des succès par la valeur du succès. On peut donc être amené à soutenir des actions qui garantissent un succès modeste, comme un don à une ONG distribuant des moustiquaires, ou des actions qui ont une faible probabilité de produire des bénéfices exceptionnels, telles que la recherche scientifique ou l’action politique.

Nos hypothèses : quelques exemples de causes efficaces

Appliquer ces principes de manière rigoureuse et pragmatique n’a rien d’automatique ni de simple. C’est pour cela que l’altruisme efficace est indissociable d’une réflexion critique sans cesse renouvelée. Cependant, les causes suivantes apparaissent prometteuses :

  • Lutter contre les problèmes de santé de base dans les pays en développement.[4] Environ 10 millions de personnes meurent chaque année à cause de maladies dont la prévention est simple, telle que la diarrhée, le paludisme, ou la tuberculose. Ces maladies ne tuent pratiquement personne dans les pays riches. Il est clairement possible d'améliorer significativement la santé dans les pays en développement, avec des interventions à faible coût dont l'efficacité est prouvée. Par exemple, distribuer des moustiquaires imprégnées d’insecticide en Afrique Sub-Saharienne permet d'empêcher un enfant de mourir du paludisme pour un coût de moins de 3 000 $[5]. Améliorer la santé crée aussi des opportunités économiques[6]: la maladie ralentit le développement des enfants, augmente fortement l'absentéisme scolaire, et les empêche de réaliser leur potentiel.
  • Aider les animaux victimes de l'élevage.[7] Chaque année environ 100 milliards d'animaux sont élevés pour que des humains les mangent : 15 par personne en moyenne[8]. L'écrasante majorité des experts en neurologie animale estiment que les cochons, vaches et poules sont capables de ressentir la souffrance d'une manière similaire aux humains. Malheureusement, ces animaux sont en grande majorité élevés dans des conditions de forte souffrance. Dans les élevages industriels, les poules ont le bec coupé sans anesthésie et sont confinées à de minuscules cages. Les truies mettent bas dans des cages où elles ne peuvent effectuer aucun mouvement, et en France les cochons sont castrés à vif. Parmi les personnes qui travaillent sur cette cause, certains tentent de développer des alternatives aux produits animaux, de convaincre d'autres de devenir végétariens ou véganes, ou d'influencer la réglementation de l’élevage et de l’abattage.
  • Contrôler les risques d'échelle mondiale.[9] De nombreux événements pourraient être catastrophiques pour la terre entière : une guerre mondiale, une nouvelle pandémie, le changement climatique, ou des nouvelles technologies à haut risque. Malheureusement, ces risques sont le problème de tous à la fois et la responsabilité de personne en particulier, et sont donc souvent négligés. Pourtant, une telle catastrophe pourrait non seulement causer beaucoup de souffrance, mais aussi réduire le potentiel de l'humanité à long terme. Par ailleurs, certains biais cognitifs nous empêchent d'estimer correctement les risques lorsqu'il s'agit d’événements très rares. Souvent, les électeurs sont très inquiets au sujet de risques peu importants, et n'accordent aucune attention à des risques que les experts considèrent comme extrêmement inquiétants.

Agir

L’altruisme efficace est intellectuellement ambitieux, mais ne fait pas de la rigueur académique son but final. Si nous prenons ces questions au sérieux, c’est car nos actions en dépendent. Les méthodes de l’altruisme efficace peuvent paraître abstraites, mais les vies des personnes que nous aidons sont pleines de souffrances et de joies réelles. Les personnes engagées dans l’altruisme efficace mettent en pratique leurs projets avec le même enthousiasme qui les pousse à analyser leur plans avec précision. Un grand nombre d’entre eux donnent 10% de leur revenu aux causes qu’ils soutiennent. D’autres ont changé leur carrière en profondeur[10].

Pour en savoir plus

Le livre Doing Good Better[11] de William MacAskill apporte un traitement plus approfondi des idées centrales de l’altruisme efficace.

Notes

[1] utilitarianism.com - Traduit de l’anglais par Fanny Verrax (archive)

[2] givingwhatwecan.org (archive)

[3] GiveWell.org (archive)

[4] Pour en savoir plus: GiveWell tente d’identifier les ONG qui aident les habitants des pays en développement le plus efficacement. Les rapports de GiveWell sont accessibles au public et se basent sur plusieurs décennies de recherches rigoureuses dans les domaines de l’économie et de la santé publique.

[5] GiveWell.org (archive)

[6] Gallup et Sachs, 2001 (archive)

[7] Pour en savoir plus: Animal Charity Evaluators tente d’identifier les opportunités de dons les plus efficaces pour aider les animaux et étudie certaines questions empiriques et théoriques cruciales pour la cause animale. La fondation Open Philanthropy Project soutient financièrement des projets pour la réduction de souffrance des animaux d’élevage, et donne un accès public aux résultats de ses recherches.

 

[8] 50 milliards d’animaux terrestres (Organisation des Nations Unies pour l’alimentation et l’agriculture, archive) et environ 80 milliards d’animaux issus de l’aquaculture (A. Mood et P. Brooke, 2012, archive).

[9] Pour en savoir plus: le Future of Humanity Institute à l’Université d’Oxford étudie les questions considérées comme les plus déterminantes pour la trajectoire future de l’humanité. L’article Existential Risk Prevention as Global Priority (Nick Bostrom 2012, Global Policy, archive) est un bon point de départ.

 

[10] 80000hours.org - voir en bas de page la section “people we’ve helped”.

[11] Amazon.fr

Oxford Uehiro Prize in Practical Ethics, “How should vegetarians actually live?"

July 4, 2016

In 2016, I was the joint winner in the Undergraduate Category of the Oxford Uehiro Prize in Practical Ethics. Read my essay here, or below.

Ethical vegetarians abstain from eating animal flesh because they care about the harm done to farmed animals. More precisely, they believe that farmed animals have lives so bad they are not worth living, so that it is better for them not to come into existence. Vegetarians reduce the demand for meat, so that farmers will breed fewer animals, preventing the existence of additional animals. If ethical vegetarians believed animals have lives that are unpleasant but still better than non-existence, they would focus on reducing harm to these animals without reducing their numbers, for instance by supporting humane slaughter or buying meat from free-range cows.

I will argue that if vegetarians were to apply this principle consistently, wild animal suffering would dominate their concerns, and may lead them to be stringent anti-environmentalists.

If animals like free-range cows have lives that are not worth living, almost all wild animals could plausibly be thought to also have lives that are worse than non-existence. Nature is often romanticised as a well-balanced idyll, so this may seem counter-intuitive. But extreme forms of suffering like starvation, dehydration, or being eaten alive by a predator are much more common in wild animals than farm animals. Crocodiles and hyenas disembowel their prey before killing them[1]. In birds, diseases like avian salmonellosis produce excruciating symptoms in the final days of life, such as depression, shivering, loss of appetite, and just before death, blindness, incoordination, staggering, tremor and convulsions.[2] While a farmed animal like a free-range cow has to endure some confinement and a premature and potentially painful death (stunning sometimes fails), a wild animal may suffer comparable experiences, such as surviving a cold winter or having to fear predators, while additionally undergoing the aforementioned extreme suffering[3]. Wild animals do experience significant pleasure, for instance when they eat, play or have sex, or engage in other normal physical activity. One reason to suspect that this pleasure is outweighed by suffering is that most species use the reproductive strategy of r-selection, which means that the overwhelming majority of their offspring starve or are eaten shortly after birth and only very few reach reproductive age.[4],[5] For instance, ‘in her lifetime a lioness might have 20 cubs; a pigeon, 150 chicks; a mouse, 1000 kits’,[6] the vast majority of which will die before they could have had many pleasurable experiences. Overall, it seems plausible that wild animals have worse lives than, say, free-range cows. If vegetarians think the latter are better off not existing, they must believe the same thing about wild animals.

A second important empirical fact is that wild animals far outnumber farmed animals. Using figures from the FAO, Tomasik estimates that the global livestock population is 24 billion (including 17 billion chicken)[7]. I restrict my count of wild animals to those at least as complex as chicken or small fish, which vegetarians clearly believe do have moral weight. Using studies of animal density in different biomes, Tomasik estimates conservatively that there are at least 6*1010 land birds, 1011 land mammals, and 1013 fish. Animals in each of these categories alone are several times more numerous than livestock.

If wild animals’ well-being is negative and the above numbers are remotely correct, the scale of wild animal suffering is vast. As Richard Dawkins writes, ‘During the minute it takes me to compose this sentence, thousands of animals are being eaten alive; others are running for their lives, whimpering with fear; others are being slowly devoured from within by rasping parasites; thousands of all kinds are dying of starvation, thirst and disease.’[8] If they accept the premises so far, consistent vegetarians should focus on preventing the existence of as many wild animals as possible, since even a small reduction in the global number of wild animals would outweigh the impact of ending all livestock production. For example, they could reduce animal populations by sterilising them, or by destroying highly dense animal habitats such as rainforests. This would place them directly at odds with environmentalists who try to preserve nature from human intervention. It may even be the case that vegetarians should react to this argument by eating more meat, since feeding the livestock requires more surface area for agriculture, and fields contain far fewer wild animals per square kilometre than other biomes such as forests.

An intuitive response to wild animal suffering can be that cycles of predation and starvation are natural, and therefore they must be neutral morally. But what is natural is not necessarily what is good, for instance, humans will routinely use technology to remove diseases which are natural.

It is important to emphasize that the claim ‘wild animal suffering is bad’ does not imply a guilt claim of the form ‘predators are morally guilty’. A lion’s instinct is indeed natural and does not deserve our moral condemnation. However, we can avoid much confusion if we remember to keep separate the concepts of guilt of an agent and wrongness of an action. It is perfectly possible to claim that X is harmful and should be prevented while also holding that the direct cause of X is not a moral agent. The fact that we are so used to thinking about cases of human behaviour, where guilt and wrongness are largely aligned, may partly explain why arguments about wild animal suffering seem counter-intuitive.

Underlying some of these principled arguments is the intuition that harmful acts, like killing livestock, are worse than harmful omissions, like failing to avert wild animal suffering. I cannot begin to give a full treatment to the act/omission debate here, but one thought experiment suggests harmful omissions matter at least somewhat. Imagine you see a fire spreading in a forest and, while walking away from the fire, you see an injured fawn: a broken leg prevents her from fleeing. You carry a rifle and could instantly kill the fawn at no cost to yourself, preventing her from the extreme suffering of being burned alive. In this situation, for vegetarians who care about harm to animals, it is clear that it would be immoral to omit to act and allow wild animal suffering to happen. So the general principle ‘allowing wild animals to suffer is morally neutral’ cannot hold.

A second set of counter-arguments are empirical: they concede that consistent vegetarians are morally obliged to reduce wild animal suffering, but attack various empirical claims made above.

It may be objected that we cannot reduce the number of animals by sterilising them, because as soon as fewer animals are born, more resources (like food and territory) become available, which increases the evolutionary payoff of producing more animals. If we sterilise some deer, there will at first be fewer fawns, so there will be more nuts and berries available, which allows other deer (or other species) to have more offspring, until we are back to the original equilibrium. The existence of such evolutionary pressures towards an equilibrium population seems plausible, but it remains an unsolved empirical question. It may be the case that the population takes several years to reach its equilibrium again, in which case much animal suffering would be averted in the meantime. Regardless, this is only an objection against one particular method for reducing wild animal numbers, and it only tells us that sterilisation would be ineffective, not harmful. If we reject sterilisation on these grounds, habitat destruction, for instance, evidently does reduce animal numbers for the long run.

A frequent objection against intervening in nature is that we are uncertain about the consequences: for instance, culling predators might cause an ecological catastrophe. While our uncertainty is a good reason to do more research in order to reduce it, it is not in principle an argument for inaction. If we are so uncertain, inaction towards predation could also be causing vastly more suffering than we currently estimate. In order to make sure our aversion to intervene is not caused by status quo bias, we can use the reversal test,[9] an elegant instance of which is provided by the reintroduction of wolves in Scotland, where they had been hunted to extinction in the 1700s.[10] If we are more worried about the uncertain effects of reintroducing wolves than we are about the uncertainty of inaction towards wolf predation, this may be due to status quo bias.

Possibly the strongest counter-argument is that we are extremely uncertain about whether wild animals’ lives are worth living. How much pain or pleasure animals feel in response to certain stimuli is dependent on facts about their neurology which is not well understood. While we may make some reasonable extrapolation from our human experience (being eaten alive is very painful), animal subjective experience may differ significantly. While animals might experience hedonic adaptation[11] to their circumstances, encounters with predators produce lasting psychological damage similar to post-traumatic stress disorder in humans.[12] There is some evidence that domesticated animals are less stressed,[13] but measures of stress hormones may not coincide with animals’ revealed preferences[14]. Clearly, I do not pretend to have solved this difficult empirical question. However I note that these considerations should also make us uncertain about the subjective well-being of farmed animals; and I have already offered reasons why wild animals plausibly have worse lives than free-range animals.

Even if vegetarians still reject this argument, and believe that wild animals’ lives are better than the lives of farm animals, to the extent that they are worth living, this does not imply they should do nothing. They should not reduce animal numbers, but they should still reduce the suffering of existing animals. Because there are so many animals and the suffering they undergo can be so extreme, this consideration would likely still dominate concern about farmed animals. One could vaccinate animals against diseases: rabies has already been eliminated from foxes for human benefit[15]. After elephants’ teeth wear out, they are no longer able to chew food and eventually collapse from hunger, after which they may be eaten alive by scavengers and predators. Fitting elephants with artificial dentures, which has already been done on captive animals, would significantly increase their healthspan[16]. Or one could cull predator populations by allowing more of them to be hunted.

A possible concern with this type of intervention may be that any advantage given to a particular individual by reducing their suffering would increase the suffering of others. For instance, if elephants can eat for longer, more other herbivores will starve; or if we kill predators, their prey will proliferate and their competitors will starve. If we think that ecosystems lie on such a razor-sharp evolutionary equilibrium where all animals are strongly competing for every piece of resource, this objection is plausible. But crucially, if we accept this, then it is becomes plausible that wild animals actually do have lives that are not worth living: if evolution produces so many animals that each can just barely survive, it is likely that they endure much suffering and little pleasure. So it seems like we must either accept that some interventions can reduce extreme wild animal suffering, or concede that animals’ lives are plausibly not worth living.

Some may choose to treat this outlandish conclusion as a reductio against vegetarianism (either against the idea that farm animals matter morally or against the belief that we should prevent them from coming into existence). Perhaps vegetarians who still reject the conclusion should increase their confidence that buying free-range meat is a good thing. For those who accept it, the question of how most effectively to reduce wild animal suffering is left open. As I have repeatedly emphasised, we are still very ignorant about many relevant empirical questions, so immediate large-scale intervention will not be very effective. In addition, intervention may have significant backlash effects and reduce sympathy for the anti-speciesist message. The best immediate action is probably to produce more research on wild animal suffering, in order to make future action more likely to be effective.

[1] Dawrst, Alan. "The predominance of wild-animal suffering over happiness: An open problem." Essays on Reducing Suffering (2009): 255-85.

[2] Michigan Department of Natural Resources. "Salmonellosis." Quoted in Tomasik, “The Importance of Wild Animal Suffering”

[3] Tomasik, Brian. “Intention-Based Moral Reactions Distort Intuitions about Wild Animals.” Essays on Reducing Suffering (2013)

[4] Horta, Oscar. "Debunking the idyllic view of natural processes: population dynamics and suffering in the wild." Télos 17.1 (2010): 73-88.

[5] Ng, Yew-Kwang. "Towards welfare biology: Evolutionary economics of animal consciousness and suffering." Biology and Philosophy 10.3 (1995): 255-285.

[6] Fred, Hapgood. Why males exist: an inquiry into the evolution of sex. 1979. Quoted in Tomasik, “The Importance of Wild Animal Suffering”.

[7] Tomasik, Brian. "How Many Wild Animals Are There?." Essays on Reducing Suffering (2014).

[8] Dawkins, Richard. River out of Eden: A Darwinian view of life. Basic Books, 1996.

[9] Bostrom, Nick, and Toby Ord. "The Reversal Test: Eliminating Status Quo Bias in Applied Ethics*." Ethics 116.4 (2006): 656-679.

[10] "Wild Wolves 'good for Ecosystems'" BBC News. BBC, 31 Jan. 2007. Web. 25 Jan. 2016.

[11] Frederick, Shane, and George Loewenstein. "Hedonic adaptation." (1999).

[12] Zoladz, Phillip R. An ethologically relevant animal model of post-traumatic stress disorder: Physiological, pharmacological and behavioral sequelae in rats exposed to predator stress and social instability. Diss. University of South Florida, 2008.

[13] Wilcox, Chritie. "Bambi or Bessie: Are Wild Animals Happier?" Scientific American Blog. N.p., 21 Apr. 2011. Web. 20 Jan. 2016.

[14] Dawkins, M. S. "Using behaviour to assess animal welfare." Animal welfare-potters bar then Wheathampstead- 13 (2004): S3-S8.

[15] Freuling, Conrad M., et al. "The elimination of fox rabies from Europe: determinants of success and lessons for the future." Philosophical Transactions of the Royal Society B: Biological Sciences 368.1623 (2013): 20120142.

[16] Pearce, David. "A Welfare State for Elephants?." RELATIONS 3.2. November 2015-Wild Animal Suffering and Intervention in Nature: Part II (2015): 153.

 

A Path Appears Book Review

July 1, 2015

This my book review of Nicholas Kristof and Sheryl WuDunn’s ”A Path Appears”. (Also available on scribd here). In 2015, the review won the Sciences Po - Books prize, a book review competition organized by my University and the magazine Books. It was published in French translation in the June 2015 edition of Books, of which I’ve scanned the relevant pages.

In May 2013 TIME magazine’s cover called millennials –those born between 1980 and 2000– the “Me Me Me Generation”. It featured a young woman taking a selfie, the emblematic act of a generation of smartphone-toting, selfish narcissists. Yet statistics such as those by the Pew Research Center describe an empathetic age group that votes less often, but is more likely to do volunteer work than their parents were. Most likely, Millennials are simply doing things differently, both shaping and reacting to the ways altruism and citizenship are being reinvented. That, at least, is the hopeful message captured in the title of Nicholas Kristof and Sheryl WuDunn’s new book A Path Appears. A generation ago, they say, “’giving back’ was what we did in December, hunched over a checkbook and relying on guesswork. In recent years, advances in neuroscience and economics –and a flowering of carefully monitored experiments- have given us much greater insight into what works to create opportunity worldwide, and much greater prospects for personal satisfaction from giving”. The sunlit path towards creating a better lives for others, they say, appears before us clearer than ever. And they want you to take it: the book is sold as a “galvanizing narrative about making a difference here and abroad—a road map to becoming the most effective global citizens we can be.”

The first chapter of the book is representative of its approach – and of its paradoxes. We meet Rachel Beckwith, an American girl who decided to celebrate her ninth birthday by asking friends and relatives to donate to charity:water, an organization that drills wells in impoverished villages around the world and lets people set up their own fundraisers online. Rachel’s target sum was not reached. Six weeks later, Rachel tragically died in a highway accident, but the story of her generosity rippled through social media, and her page eventually raised over a million dollars. Cut to Lester Strong - who left his job as a news executive to run Experience Corps, an organization bringing in older Americans to tutor students in public schools across the country. Next we learn about Dr. Gary Slutkin, whose Cure Violence program combats urban violence in the United States by applying methods of epidemiology. After these uplifting stories, the second half of the chapter focuses on numbers and scientific studies designed to show how much more we know about creating opportunity; and how much more efficiently, therefore, we who live in privilege can employ our wealth and skills for causes larger than ourselves. Esther Duflo, an MIT economist, has pioneered the use of randomized controlled trials, which apply the rigorous methods of pharmaceutic drug testing to policies designed to help the world’s poorest people. She found that for 50 cents a year, you can deworm a child in Kenya, increasing their school attendance and cognitive development. Once an adult, that child will earn 20% more compared to those in a control group. So although two thirds of Americans donate an average of $1000 to charity each year, they rarely give money away as intelligently as they make it. As Esther Duflo says, “Worms have a little bit of a problem grabbing the headlines. They are not beautiful and don't kill anybody”.

This book, like its first chapter, constantly meshes hard evidence with inspiring anecdotes. As regular New York Times columnists and Pulitzer prize winners, the authors have clearly understood that if they want to produce a compelling book they must combine storytelling with rational argument. In fact, as they keep piling up story after story of activists saving the world, you may begin to see the strings. But the real challenge of this 656-page tome is to coherently bridge the gap between the rational and the emotional, the Esther Duflos and the Rachel Beckwiths. After all, these motivations for altruism are quite different, and they may point to quite irreconcilable courses of action. Can Kristof and WuDunn successfully appeal to both the calculating utilitarian and the impulsive altruist?

The book is structured into three parts, the first of which collects stories and evidence about programs that work best in providing opportunity. Microfinance institutions have been all the rage in recent decades. They lend small amounts to developing-country entrepreneurs who lack access to traditional banks, without requiring any collateral. The idea is that a poor Kenyan farmer has exceptionally high-return investment opportunities, such as doubling his crop yield by using fertilizer; but is never able to get started because he does not have enough money to buy that first unit of fertilizer. When Esther Duflo and her team conducted a randomized trial of microlending in Hyderabad, they found that 7 percent of those who had received a loan had successfully started a small business, compared to 5 percent in the control group. The economists called themselves “quite pleased with these results” because they demonstrated that the main goal of microfinance had been achieved. But nearly everyone else in the development community felt a huge letdown. Having believed microfinance to be a silver bullet, they found those numbers depressing. In fact, some microfinance institutions reacted by trying to cast doubt on the study. In politics as in development aid, then, great narratives are not always compatible with the numbers. Though not a magical solution, supporting a microfinance institution is still an efficient donation. The average loan repayment rate is over 90%, which allows your donation to be reinvested many times into the local economy.

As it turns out, for governments in the rich world, creating opportunity is also a great investment. A black man in America is more likely to spend time in prison than college. If that statistic can be turned around, the taxpayer benefits, too. In one of the best chapters of the book, entitled “The land of opportunity – if you catch them early”, the authors argue that the best way to make the lottery of birth less unfair is to intervene in the earliest stage of life. Actually, it’s better to start before that: during pregnancy, the fetal brain is being shaped by the uterine environment in ways that will affect the child for the rest of his or her life. A simple program to encourage women to stop smoking during pregnancy costs about $30 per woman counselled. Randomized trials have shown that each of these $30 save about $800 in averted neonatal costs. After birth, returns to society quickly decrease, but it’s still not too late: pre-kindergarten nursing visits for low-income unmarried mothers produce $5.7 in state and federal government savings for each dollar invested. James Zimmermann, CEO of Macy’s and an advocate for this kind of intervention, puts it succinctly: “investing in early childhood achieves the best return on investment for our country. Currently more than 90% of our education dollars are spent after age five, yet 85% of a child’s core brain structure is developed before age five.” There are three words in that quote you may have glossed over: for our country. Yet they represent one of the main paradoxes of this book. The authors keep talking about effective giving, but they miss the biggest effectiveness gap of all: between poor and rich countries. If you read this book trying to decide how best to make a difference, forget about nurse visits: 50-cent-a-year deworming in Africa beats anything you can do in the rich world. The authors repeatedly show they are aware of this discrepancy, yet they brazenly ignore its implications. It may feel good to say that poverty at home and abroad are “too important to be pitted against each other”, but upon reflection that turns out to be as fallacious as claiming that college scholarships and nurse visitation programs cannot be compared. If this book bridges the gap between rational analysis and gut feeling, it is only at a huge cost in consistency. As a result, A Path Appears lacks the moral force of Peter Singer’s resolutely utilitarian manifesto for helping the global poor, The Life You Can Save.

In part two, the authors discuss not on-the-ground programs, but rather how the art of helping itself has been transformed by new approaches. Here, they wade into more controversial terrain, but that may be where their insights will most surprise you. As an example of the “social business” model that is blurring the lines between for-profit and non-profit, they hold up an enterprise by Danone in producing a Yogurt containing micronutrients to fill nutritional deficits of children in Bangladesh. The yogurt is sourced locally, providing business and employment opportunities. This venture is often criticized as a public-relations ploy and, to some, smacks of cultural imperialism (Bangladeshis had never heard of Yogurt before Danone started advertising). The authors don’t dwell on these considerations; instead they make the larger point that social businesses, which operate on a “double bottom line” model combining profits and social impact, have helped bring the efficiency and scalability of business to charitable causes.

One of the most interesting characters of this book is Dan Pallotta, an ambitious consultant who in 1994 launched Pallotta TeamWorks, a company which organized bike rides and other events to raise money for AIDS and breast cancer prevention. Pallotta knew that the reason his events were successful was that they were fun, cool, and well-organized. He poured money into marketing and logistics, saying “We advertised our events the way Apple advertises iPads.” He also paid himself $394 000 in 2001. All this meant that only about 60% of the money raised went to the charities.  Increasingly criticized for this high overhead figure and his somewhat glitzy approach to charity, Pallotta was dropped by his sponsor in 2002. That sponsor tried to run the events themselves, keeping costs down. But though overhead was lower, the amounts raised for charities dropped 70% in one year, from $6 million under Pallota’s management to only $1.6 million. People started to realize that what had been called overhead was actually essential to the fundraising effort. In an inspiring TED talk, Pallotta argues that “the way we think about charity is dead wrong”, and urges donors to ignore the “depressing label of overhead” and instead focus on what charities are getting done per dollar. Meanwhile, established charities have quickly realized that they can exploit our emotional responses through advertising: in 2002, UNICEF  experimented with putting a nickel in its mailings, visible through a glassine window on the envelope. This triggered people’s reciprocity instinct, and the response rate doubled. Sounds like manipulation? In the for-profit world, it’s called marketing. If we think it’s okay for a multinational to use those tactics to sell hamburgers, why do we frown upon charities doing it to ‘sell’ education for Ethiopian girls?

The last third of the book is dedicated to showing that far from being a Gandhi-style sacrifice, altruism can be a source of great fulfillment. The MRI scan, the grail of modern brain science, has been used in countless experiments to show that giving money to charities activates the same pleasure centers of the brain as eating fine food or having sex. In fact, things we think will make us happy, like winning the lottery, have almost no long-term impact; but helping others has been shown to increase long-term life satisfaction and health. While the studies cited are rather solid, this optimism hides the many ways in which our altruistic instincts actually work against us. The American charity Smile Train, which offers cleft palate surgeries to children in the developing world, uses the photo of a Russian boy in their mailings: randomized trials showed that the mostly white donor base responded best when shown a picture of someone with the same skin color. Spontaneously, we help those who resemble us. We also like identifiable victims: people are actually more willing to donate $300 000 to save one child than the same amount to save eight children. Perhaps we should not rely all too much on our instincts. Very often, actions that make you feel good crowd out actions that do the most good.

A Path Appears functions nicely as an inspiring catalogue of all the ways you can help others. It provides a treasure trove of studies and anecdotes that will delight anyone interested in altruism; its twenty chapters are actually so exhaustive that they can become exhausting. If you have a friend who might like to do some volunteer work, but doesn’t know where to start, buy her the book. If you care more about the impact than the narrative, however, I have a provocative suggestion for you. We’ve already seen how much good money can do, when given to the best causes. If you really want to make a difference, why not try something none of the people in this book have done: quit your job, but don’t go work for a charity. Instead, go to Wall Street, earn as much money as you can, then give it to the most effective charities. Join Goldman Sachs. Save the world.

Scan:

Click here for scan of French translation