Here I present an argument against the ‘Effective Altruism’ (EA) movement. First, I argue that the philosophy of the EA movement is predicated upon a contested utilitarian ethical framework which the movement makes insufficient efforts to justify, and seems to lead to positively perverse conclusions when taken to extremes, as can be seen in the ‘repugnant conclusion’. Second, I argue that the myopic EA focus on empirical evidence of charitable efficacy is misguided, as it leads to worthwhile interventions being neglected simply because they are too difficult to measure. Third, I argue that in focusing excessively on empirical evidence and cost-benefit calculations, EA ignores many longer-term systemic flow-on effects, which in many cases can be far greater than short-term charitable outcomes. Finally, I argue that the EA movement’s lofty standards for empirical evidence and rational decision making are so onerous that even the EA movement itself cannot live up to them, meaning that EA fails in its endeavor to extirpate personal intuition and subjective judgements from charitable analysis.
There have been a number of critiques of effective altruism put forward, for example here, here, here, here, and here, but at least in my view, most of them aren’t very compelling or coherent. This piece is attempt to remedy the situation by providing a more robust case against EA. For purposes of this piece, I define effective altruism as a social movement, drawing particular inspiration from the work of Peter Singer, the core characteristics of which is a focus on achieving maximum charitable impact by targeting donations to causes which have been empirically demonstrated to yield the most cost-effective outcomes, and shaping one’s life and career so as to maximize the aggregate impact one can make through such programs. Note that the views expressed in this piece do not necessary reflect those of the author of this piece.
EA is based upon an underlying presumption of utilitarian ethical theory. Indeed, it often seems such utilitarianism is treated as an axiom not worthy of further discussion. But the trouble is, not everyone shares such a utilitarian ethic. In particular, deontologists and virtue ethicists will not necessarily agree with the hard-nosed EA utilitarian that we should let the blind man down the road go without a guide dog if doing so means that we are able to instead repair the eyesight of fifty Africans. Nor is there any clear reason for such people to accept this utilitarian viewpoint, aside of course from extended philosophical discussion, which given the past two millennia of philosophical history seems unlikely to yield much consensus anyway. Given such differing views on ethics, why is EA so dogmatically confident in their utilitarianism, taking it as so obvious and basic that all others should just automatically agree?
Indeed, there are some fairly simple yet powerful arguments against the sort of utilitarian ethic championed by EA. Consider, for instance, the general EA antipathy towards donations to art or cultural organizations, or even ‘community building’ sorts of activities like donating to the local scouting group or the Make-A-Wish Foundation. The standard argument against such donations is that, whilst they may be ‘nice’ and improve our communities, these benefits are overwhelmed by the utility of the many lives that could be saved if the same resources were redirected to health and education interventions in the Third World. The trouble with an argument like this, however, is that taken to its natural end, it leads to the repugnant conclusion: namely that it is better to have a world comprised of a very large number of people whose lives are just barely worth living, compared to a world comprised of a much smaller population of healthier, happier, flourishing persons. Virtue ethicists in particular would likely object to the notion that saving lives is the overriding ethical consideration that we face – what also matters is what we do with those lives, for example, to build happy, supportive, flourishing communities. But EA utilitarianism seems to tell us that more lives is always preferable.
EA supporters might retort that this abstract philosophical problem has little bearing on the actual real-world decision of whether we should donate to an art gallery or to AMF, but this seems like a highly unsatisfactory, and indeed positively evasive, response from a movement so founded on intellectual rigor and philosophical clarity. Analysis such as these demonstrate how EA rests on a much shakier philosophical grounding than its proponents care to consider or admit.
The Limits of Empirics
EA advocates careful measurement of the cost-effectiveness of donations, and generally recommends against donations to charities unable to demonstrate measurable positive impacts. Consider for instance GiveWell’s recommended charities: these are the charities for which GiveWell was able to find sufficient empirical evidence of cost-effectiveness. The ‘non-recommended’ charities are such not because they have been found to be ineffective, but only because they have not been found to be effective. The problem with this is approach is that, in general, absence of evidence is not evidence of absence. The question one must consider is: would we expect to have quality evidence of efficacy for charities that were effective? If so, then absence of evidence would be evidence of absence. But there seems to be no reason whatever to think this, and many reasons to think otherwise. In particular, social interventions of the kind carried out by charities are exceptionally complex, multi-faceted, and difficult to measure. It is not at all clear, therefore, that the absence of the evidence of efficacy tells potential donors anything at all about the charity in question. But EA treats such absence of evidence as if it were positive evidence that the charity is not worth donating to.
There are deeper issues with this overarching focus on the empirical. An obsession with specific measurable outcomes inevitably leads to a focus on those outcomes to the exclusion of other considerations. Even specific exhortations to the contrary are often ineffective, as the formal and informal incentives of the relevant organisations are all focused towards the specific goal that is actually being measured: for corporations this is profits, leading to neglect of environmental and other considerations. For charities, this problem could manifest itself in other ways. For example, if a program’s efficacy is measured by the number of bednets placed, this produces an incentive to supply large numbers of low-quality nets to easily-accessible populations who may not necessarily need them. One can consider many similar examples where a focus on one or two easily-measurable outcomes results in a neglect of other important aspects of a program, catastrophically undermining its efficacy.
The retort to these sorts of problems is usually the assertion that one simply needs to adopt better metrics. But that is exactly the point – often there are no good metrics, or the only good metrics would be too complicated and expensive to collect given limited resources and infrastructure. So either the program goes ahead with lousy metrics, in which case EA over-emphasis on empirical outcomes can lead to adverse consequences owing to the inadequacy of the metrics being used, or the program just doesn’t go ahead at all, owing to the lack of any ability to demonstrate its efficacy. EA philosophy seems impotent to deal with this deep dilemma.
Excessive focus on empirically-proven charitable interventions also means that the EA movement has a tendency to ignore other causes and consequences which are harder to quantify or even define with any precision, but are nonetheless no less real or important. One example would be ‘flow-on effects’, or second and third-order consequences of charitable programs. What effects does a program have on community cohesion in the long-run? How does this particular initiative alter the incentives faced by local officials, charity workers, and donors? What unintended consequences may this action have? Such indirect consequences are by definition hard to foresee or measure, and therefore tend to be neglected in the sort of outcome-based analyses favoured by EA advocates.
Consider another example. If EA had existed in the 18th century, its advocates would presumably have argued for the upper classes and emerging bourgeoisie of Western Europe to invest their time and energies in poverty reduction, promoting basic health care and education for the masses, etc. Much as contemporary EA supporters decry donations to art galleries as being comparatively ineffective, our imaginary 18th century EA predecessors would presumably have opposed the use of time and resources for speculative research in physics, chemistry, and biology, or work in areas that obviously have no practical value for improving peoples lives, such as archaic mathematical concepts like calculus and statistics, or inventing new fields of study like ‘economics’, or new ethical philosophies like ‘utiliarianism’, given that lack of any evidence at all that this sort of work would yield any practical benefits at all for the needy. And yet, without these pioneering developments, the modern EA movement would be without the technical ability to carry out its approved interventions, the statistical and modelling tools necessary to gauge their effectiveness, or even the very philosophical and theoretical framework with which to articulate their position and analyze opposing views. Examples such as this show that, at best, EA recommendations about cause prioritization seem to be missing a great deal of importance.
Impossible Standards to Meet
This issue of the limits of empirics brings me to a final point about the impossible standards that EA sets, standards which even it is unable to meet. The EA movement calls for charitable efforts to be prioritized on the basis of objective cost-effectiveness analyses. But there are numerous ‘meta-level’ questions surrounding EA which simply cannot be decided on the basis of such criteria, either because of lack of data, lack of agreement about key concepts, lack of time for analysis, or some combination of these factors. Consider the debate over the question of whether to give now or give later. What is the evidence-based, objective basis for deciding one way or the other on this question? Or consider the debate about the relative importance of ameliorating global poverty vs tackling existential risks. Again, where is the evidence or objective criteria for arriving at a position on this matter? Similar questions could be raised regarding disputes concerning earning to give, environmental ethics, population ethics, animal suffering, and many other matters. What EA supporters end up doing, of course, is make a decision based on intuition, emotion, personal experiences, and vagaries of their particular situation – precisely the decision making methods which the EA movement decries in charitable giving more broadly. This comes about, as I have said, because the standards that EA sets for itself (and others) are simply impossible to meet. We don’t have enough data, we don’t have enough agreement about core concepts, and we don’t have enough time or ability to weigh up and analyse all the various considerations.
EA supporters might respond by saying “yes but at least we’re trying to use evidence and rationality as much as we can, so our donations will on average be at least somewhat more effective, even if we make some mistakes and leave out some factors”. It is not at all clear, however, that even this claim is true. Since EA supporters cannot agree about whether to give now or later, or whether to support malaria eradication or friendly-AI research, and since no objective rational basis exists for supporting one over the other (at least by EA standards, given the lack of any clear evidence), the effectiveness of these different approaches could well vary just as much as the effectiveness of traditional charities. As such, EA giving falls prey to the very same criticism it levels against traditional charities – one might be doing good, but we can’t really tell because we can’t measure it.