Should we always give to the most effective charities?

By Teresa M-D, for University College London, 2023

Introduction

Effective altruism takes the idea that we ought to do as much good as possible to entail that we “apply evidence and reasoning to work out the most effective ways to help the world”.[1] It is thus a highly demanding moral theory; suggesting that we not only have an obligation about whether and to what extent we give, but also about where to do so. I interpret giving ‘most effectively’ to mean that for any amount X, one must donate X to the charity that will comparably produce the most good.[2] I further assume that we could theoretically possess sufficient and reliable knowledge to discern what giving the most effectively would be - even if in practice such judgements may be fallible if not impossible. I adopt the putative conception of a charity as a non-profit organisation dedicated towards producing some benefit or good. Furthermore, I note that we can interpret ‘always’ in either a strong or a weak sense. The first is comprised of two claims: “You should always give and this should be to the most effective charity”. The second is conditional: “If you are going to give, then you should do so to the most effective charity.” This is interestingly compatible with there being no obligation to give in the first place. 

This essay will see me assessing an argument for the weak conclusion, given that this does not involve answering a further question of an obligation to give. Inspired by Horton, I term this “The All or Nothing [Argument]”[3]. Exploring what I take to be the strongest challenge - the problem of supererogation, and (briefly) how this bears on moral permissibility, I ultimately conclude: we should not always give to the most effective charities. 

1. The All or Nothing Argument (AoN)

    a) Collapsing Building 

Imagine that you are a bystander watching a building on the verge of collapse. Unfortunately, two children will be crushed if you take no action. Sacrifice both your arms, however, and you can save at least one, if not both. Hence there are “three options: do nothing, save one child by allowing your arms to be crushed, or save both children by allowing your arms to be crushed.”[4]  These can be reframed in terms of three choice types (in order): 

    (1) Inaction 

    (2) Non-optimific action 

    (3) Optimific action

Horton intuits that it would be morally wrong to choose (2), although morally permissible to do nothing. This is because it is of no extra cost to you to perform the optimific action - that is, the one that brings about the most good. Option (2) therefore “constitutes a deliberate refusal to do something much better at no extra cost”[5]; in Collapsing Building, the moral agent actively decides not to save one child, despite this bringing about double the benefit for the same cost. Underpinning this is Pummer’s ‘Avoid Gratuitous Worseness’ (AGW) principle: 

[Weak] “It is wrong to perform an act that is much worse than another, if it is no costlier to you to perform the better act, and if all other things are equal.”[6]  

AGW may be reformulated to be more demanding to varying degrees, the strongest of which exchanges the first clause for “even if it is much costlier to you to perform the better act”.[7] Since the project of this essay is to assess an argument for the weak conclusion (cf. introduction), I will stick to the weakest form as given above.[8]  

b) Optimific Giving 

As Horton and Pummer contend, although AGW is brought out via an example that does not concern charitable giving, it grounds an argument in favour of doing so optimifically:[9]   

P1: It is much worse to save one child than to save two. 

P2: It is wrong to perform an act that is much worse than another if it is no costlier to you to perform the better act, and if all other things are equal (AGW). 

P3: It is wrong for you to select option (2) (save one child). 

P4: Given the cost of either (2) or (3) to you, it is morally permissible to choose [1]. 

P5: But if you are willing to perform (2) or (3), you ought to choose (3). 

The case is then substituted for that of giving to charity. Since P2-P5 are presented in general terms they remain the same. The analogous argument thus follows: 

P1*: It is much worse to donate X to the charity(s) that will use it non-optimifically. 

P2 - P5 

C: If you are willing to give to a charity (ie. perform (2) or (3)), then you “ought to give to one of the charities that would use [y]our gift to do the most good.”[10]  

Whilst I consider P1 to be a fairly uncontroversial assumption it’s analogue (P1*) may not be so intuitive. In Collapsing Building, the difference between (2) and (3) is a whole life out of a possible two lives that can be saved. Hence, (3) is, in quantitative terms, twice as good as (2). This grounds (2) being deemed ‘much worse’ than (3). Therefore, to establish the analogy, I read P1* as implying that the non-optimific outcome is significantly less beneficial than the optimific one. For example, think of donating X to either a) a soup kitchen that could provide a hot meal for 100 people, or b) a homeless charity that provides shelter for 20. The first is five times ‘better’, at least in terms of the quantity of individuals who receive the benefit. Thus, I think it is plausible to suggest that option (2) in this case could indeed be ‘much worse’ than (3). I grant this at least for the sake of setting up the argument. 

The other and more apparently controversial assumption is P4. Although it follows from AGW (as given in P2), and must be allowed if we are going to support the weak conclusion11, when combined with P3 it generates a counterintuitive result - it is better to do nothing than to save one child. Mapped on to Optimific Giving, it is better not to give, than to do so non-optimifically. To mitigate this we may add a further premise: 

P4.5: “If you are not going to save both, you ought to do the next best thing, which is to save one”.[12]  

P4.5*: If you are not going to donate X optimifically, you ought to do the next best thing, which is to donate it non-optimifically. 

Although, given that you’re willing to do (2) means that you ought to do (3) (P5), choosing (2) is still intuitively better than choosing (1). In concrete terms, saving one child is better (overall) than saving none; analogously, giving to some charity is better than not giving at all. P4.5 thus rests on the assumption that the agent is not willing to save either child (or by extension both), and analogously, give to charity at all. This is captured in the antecedent of P5. 

To ground such reasoning, Horton suggests that moral permissibility is dependent on one’s ability to justify the action. Choosing (1) is morally permissible on the basis of the cost you will incur by choosing either (2) or (3). But if you are willing to perform (2) there is no further justification for not performing (3). That is, assuming one “do[es] not have adequate agent-relative reasons to favour this other outcome (2)”.[13] This highlights the non-consequentialist nature of AoN, in spite of its appearance. Such a theory of justification seems implicit in P4 and P5 in particular. I further take the ‘do not have adequate agent-relative reasons’ to map onto the ‘all other things being equal’ clause of P2. As given by the name, agent-relative reasons are reasons particular to the moral agent in a given case that may potentially ground a justification to choose (2). For example, if one of the children in Collapsing Building was my own, and so I had reason to prioritise her safety. Likewise, think of donating to a local hospice charity as an expression of gratitude for their care of a terminally ill relative.

2. Critical analysis

a) The problem of supererogation 

Whilst AoN argues against the idea that “optionality about whether to help at all entails optionality about which particular helpful act to perform”[14], it neglects to address why this may indeed be a “common assumption”.[15] This seems to lie in our view of charitable giving as supererogatory: actions which “instantiate ideals of the highest possible moral evaluation, although their performance is (...) considered [non]obligatory.”[16] On a stronger formulation, it is in virtue of the supererogatory action being (morally) non-obligatory that we morally praise the individual who performs it. That is, non-obligatory at least insofar as it goes beyond what is expected - “the minimally decent permissible option [moral minimum]”.[17] In both Collapsing Building and Optimific Giving this is framed as (1). Because the weak conclusion does not support a general duty to give (ie. to select (2) or (3)) it seems at a first glance to be compatible with the supererogatory intuition; one may even contend that AoN explicitly supports the idea through the term “gift” in C which conveys the ‘voluntariness’ bound up in the commonsense understanding of charity. As McMahan emphasises: “If the cost to her of giving her wealth away ma[kes] it supererogatory to give it to any charity, the additional cost to her of giving it [...optimifically] seems to exempt her from any duty to give it to the former.”[18] 

However, if the supererogatory intuition is correct, AoN faces two connected problems:

i) P5 can be undermined because both (2) and (3) are considered optional but P5 articulates a duty to perform (3) 

ii) (2) would denote an action with opposing evaluative properties (simultaneously morally impermissible and morally praiseworthy).

Underscoring i) is the idea that both (2) and (3) are morally praiseworthy to the extent that they go beyond the moral minimum. This is not to say that (2) and (3) are evaluated equivalently; as Benn outlines, “some supererogatory actions are better than others, and some of those who perform supererogatory actions are deserving of greater praise”.[19] Whilst we may recognise that donating to the soup kitchen (3) is significantly better than donating to the homeless shelter (2), we morally praise both actions (and to similar extents). This calls P1* into question, even in spite of my qualified account of it. It seems that the only cases in which donating X to one charity would indeed be ‘much worse’ than another are those in which the charities being compared pertain to very different causes, generally considered of equally different significance. For example, an arts foundation vs. a charity rescuing refugee children. 

However, given that Collapsing Building involves saving either one or two children, and not say, one child and a priceless Da Vinci, I have been assuming that the causes in Optimific Giving must have (broadly) the same aims[20]. If not, we are left with a disanalogy; and yet, on this reading, the supererogatory intuition is indicative of a disanalogy (presented through P1*). Hence, the proponent of AoN is stranded. They can use this to their advantage, however, claiming that this only serves to show a discrepancy in how we perceive the two cases - a fact which they acknowledge through the ‘common assumption’. Hence maintaining P1 and its analogue P1* (therefore that (3) is much better than (2)), they can claim that (2) only appears to warrant praise. This is something I shall return to later on. Regardless, challenge i) only requires that both (2) and (3) are framed as optional (whether or not they are both praiseworthy), and (3) is perceived as supererogatory. The tension thus arises because P5 suggests that the agent must have been willing to perform (3), and since they were willing, they had a duty to choose (3). This conflicts with the idea that, as McGoldrick puts it, “the saint [...] bestow[s] upon others the free gift of their service”[21] where I take the ‘saint’ to be the person who acts optimifically - giving to the soup kitchen or saving both the children. 

The problem posed by the supererogatory intuition is thus that the optionality as given in i) extends over the “acts that involve greater cost to the agent than the bare minimum”[22] (namely (2) or (3)). But insofar as someone is willing to perform one of these options, they are conditionally obliged to select (3). In this way (3) is not truly (ie. independently) optional[23] and therefore not, in the way that I have outlined, supererogatory. Given that the supererogatory intuition underpins the common assumption about charitable giving, the strongest tactic for the proponent of AoN is to show that “even [if] it is supererogatory to give money to charity, [...] one ought to give it in a way that would do more rather than less good.”[24] In this way, they can undermine that the intuition leads to the common assumption. 

b) Redefining the supererogatory 

To do this Horton redefines supererogatory acts as “morally praiseworthy but not obligatory unless we are willing to perform them”.[25] The addition of this clause entails that dutiful and supererogatory action are not mutually exclusive. On the contrary, Horton argues that “it is largely because you were willing to [choose (3) when (1) would have been permissible], that we should regard you as heroic and praiseworthy.”[26] This is supported by considering a case of Collapsing Building in which the moral agent manages to perform the (3) without incurring the cost presumptively entailed by this action. The act is nonetheless “heroic because they were willing to risk life and limb”.[27] However, I am sceptical of the reformulated supererogatory intuition. Whilst I agree with Pybus that “moral praise is apportioned not to the acts [in] themselves, but to the attitudes [and] dispositions [...] that give rise to them”[28], Horton seems to incorrectly reduce moral praise to a person’s willingness. In Collapsing Building it is not (just) that you were merely willing to perform (3) that we herald you as a saint, but that you actually performed (3) and saved the lives of the two children. This is because being willing to act is wholly compatible with inaction (1). Though I might, for example, be willing to donate to the local hospice, I may not actualise this because a different, competing intention outweighs it - for example, paying for a family holiday. This cannot afford moral praise. 

Yet in response, Horton suggests that what makes (3) supererogatory is the fact that the agent “goes beyond what is morally required of them by being willing to make great sacrifices even when it is permissible for them to give priority to their own interests”.[29] Hence insofar as someone selects (1) they could not have truly been willing to perform (2) or (3); their choice expresses that whilst they might be hypothetically willing to give, they are not comparatively willing in light of the other, mutually exclusive, options. Taking the example, I am not willing to donate X to the hospice given that this means not being able to fund holiday plans (a conflicting personal interest). Yet this, I think, brings out another point of disanalogy between Collapsing Building and Optimific Giving: whilst it could be possible, in theory, to perform (3) and miraculously keep both of your arms, the cost in the case of donating X is the means through which (3) is achieved. This, along with the fact that most people would regard the cost to you of both of your arms as more serious than of some donation X, serves to highlight that willingness to perform (2) or (3) may not have the same normative significance in the two cases. 

c) Moral permissibility 

Being charitable, let us assume that the redefinition holds, and a duty could both be compatible with supererogation and theoretically conditional on willingness.[30] As the centre of moral praise for supererogatory actions no longer lies in their optionality, the redefinition given by Horton supports the idea that (2) is not supererogatory and thereby not praiseworthy; “some optional actions may well be morally bad”.[31] In this way the redefinition can challenge the common assumption. And yet, as Pummer himself articulates, “supererogation is not always good in some absolute sense; it is just comparatively better than a permissible alternative”.[32] Therefore, as (2) is still better than (1) - ie. in Optimific Giving donating is better than not doing so at all - (2) may still be classed as supererogatory. One may of course object, as AoN does in P3, that it is not permissible[33] to save one child over two, or donate X to the homeless shelter over the soup kitchen. However, the redefinition of supererogation in line with conditional obligation entails that (2) is “conditionally permissible: [...saving one] it is the next best option given that you won’t save two”.[34] This is what P4.5 suggests. In this way, the challenge in ii) is pervasive. One possible way to prevent this would be to maintain that (2) is morally wrong and therefore blameworthy, despite being permissible. Perhaps it could be classed as a ‘suberogatory’ action: “a subset of the permissible that are morally bad to do, but not forbidden”.[35] The proponent of AoN could support this reading, in highlighting that (2) is indeed not forbidden to the extent that one has adequate ‘agent relative reasons’ to justify selecting it. 

Yet herein lies a final challenge. The most apparent reason why someone may give non-optimifically to a charity is because this better allows them to “promote, secure or maximise some value”[36] they hold. As noted, I assume a similarity in causes between (2) and (3). I therefore compare our paradigm ‘agent-relative’ case - the hospice that cared for my sick relative - with another nursing charity that can use the donation more effectively. The problem here is that it seems that the value that grounds the non-optimific choice (for example, personal connection to the hospice) is necessarily not the morally relevant value the agent used to ground their choice of cause. In this case, the value would be something like caring for the sick. This is because, were they acting to promote the morally relevant value, the agent would act optimifically (ie. choose (3)). Hence, I do not think there are adequate agent-relative reasons to justify favouring (2) over (3)[37]. This makes salient my calling it the ‘All or Nothing’ argument; (2) is permissible only in theory but not in practice. Moreover, since “conditional obligations [...] are the result of moral options - moral permissions to do less than the best act or fail to do what one has a duty to do”[38], if (2) could never be permissibly actualised, I do not think it could ground the duty to choose (3). 

Conclusion

Despite the restricted interpretation of ‘always’ assumed by AoN - that one should always give to the most effective charities, if we are willing to give at all, and we do not have agent-specific reasons to give non-optimifically - I have argued against its weak conclusion. This was through the problem of supererogation; the intuition that underpins the common assumption about the optionality of all charitable giving, and hence directly counters the idea that there is any obligation to give optimifically. Not only does this challenge various premises in AoN, but also the strength of the analogy between Collapsing Building and Optimific Giving. Therefore, even though my analysis raises the possibility that the weak conclusion may dissolve into a stronger one (eg. any obligation to give to (3) could not be a conditional one, unless (2) is an actualisable option), this cannot be arrived at via the argument from analogy AoN presents. Furthermore, given that the weak conclusion can be undermined, it is difficult to think of what could ground a stronger one with much more stringent and counter-intuitive moral demands. Hence, I argue, we should not give to the most effective charities.

• [1] Singer, P. 2015, pg.5 - from Timmerman,T. 2016, pg.661 
• [2] I assume X is some non-insignificant amount relative to the agent 
• [3] Horton, J. 2017 
• [4] Horton, J. 2017, pg.94 
• [5] Pummer, T. 2016, pg.87 
• [6] Ibid. pg.84
• [7] Ibid. pg.93 
• [8] Ie. AGW[strong] would almost always entail an obligation to act as well as where to do so 
• [9] My formalisation of Horton, J. 2017, Pummer, T. 2016
• [10] Horton, J. 2017, pg.102 
• [11] Because, again, otherwise there would be an obligation to act 
• [12] Horton, J. 2017, pg.97 
• [13] Ibid. pg.98 
• [14] Pummer, T. 2016, pg.82 
• [15] Ibid. pg.78 
• [16] Mcgoldrick, P. 1984, pg.523 
• [17] Munoz, D; Pummer, T. 2021, pg.1431 
• [18] McMahan, J. 2018, pg.6 
• [19] Benn, C. 2017, pg.2408 
• [20] And of putatively high moral significance
• [21] McGoldrick, P. 1984, pg.252 - my emphasis 
• [22] Benn, C. 2017, pg.2404 
• [23] Ie. the set of (2) and (3) is jointly optional, but (3) is not optional insofar as one is willing to choose action over inaction 
• [24] McMahan, J. 2018, pg.2 
• [25] Horton, J. 2017, pg.101 - my emphasis 
• [26] Ibid. 
• [27] Benn, C. 2017, pg.2405 - my emphasis 
• [28] Pybus, E. 1982, pg.197 
• [29] Horton, J. 2017, pg.101
• [30] For discussion on conditional obligations: Munoz, D; Pummer, T. 2021, and Rulli, T. 2020 
• [31] Benn, C. 2017, pg.2409 
• [32] Munoz, D; Pummer, T. 2021, pg.1436
• [33] Cf. fn.17 
• [34] Munoz, D; Pummer, T. 2021, pg.1433 
• [35] Driver, J. 1992 - from Dorsey, D. 2016, pg.329 
• [36] Berg, A. 2018, pg.270 
• [37] Ie. where (2) and (3) are alike 
• 38 Rulli, T. 2020, pg.372 - my emphasis

Bibliography

Primary sources: 

• Horton, Joe. “The All or Nothing Problem”, Journal of Philosophy Inc. (2017 Edition, Vol.114), pg. 94-104, DOI: 10.5840/jphil201711427 
• Pummer, Theron. “Whether and Where to Give”, Wiley: Philosophy and Public Affairs (2016 Edition, Vol. 44), pg.77-95, DOI: 10.1111/papa.12065 

Secondary sources: 

• Benn, Claire. “Supererogation, optionality and cost”, Springer: Philosophical Studies (2018 Edition, Vol.175), pg.2399-2417, DOI: https://doi.org/10.1007/s11098-017-0965-7 
• Berg, Amy. “Effective Altruism: How Big Should The Tent Be?”, University of Illinois Press: Public Affairs Quarterly (2018 Edition, Vol.32), pg.269-287, URL: https://www.jstor.org/stable/10.2307/26910000 
•  Dorsey, Dale. “Ethical Theory and Moral Practice”, Springer: Ethical Theory and Moral Practice (2016 Edition, Vol. 19), pg.329-342, URL: https://www.jstor.org/stable/24762630 
• Driver, Julia. “The Suberogatory”, The Australasian Journal of Philosophy (1992 Edition, Vol.70), pg.286-295, DOI: 10.1080/00048409212345181 
• McGoldrick, Patricia.M. “Saints and Heroes: A Plea for the Supererogatory”, Cambridge University Press: Philosophy (1984 Edition, Vol. 59), pg.523-528 URL: https://www.jstor.org/stable/3750369 
• McMahan, Jeff. “Doing Good and Doing the Best”, Oxford University: Manuscript (2018 Proof), DOI: 10.1093/oso/9780190648879.003.0004 
• Muñoz, Daniel; Pummer, Theron. “Supererogation and conditional obligation”, Springer: Philosophical Studies (2021 Edition, Vol.179), pg.1429-1443, DOI: https://doi.org/10.1007/s11098-021-01724-y 
• Pybus, Elizabeth.M. “Saints and Heroes”, Cambridge University Press: Philosophy 57 (1982 Edition, No.220) pg.193-199, DOI: 10.1017/s0031819100050762 
• Rulli, Tina. “Conditional Obligations”, Florida State University Department for Philosophy: Social Theory and Practice (2020 Edition, Vol. 46) pg.365-390, DOI: 10.5840/soctheorpract20204189 
• Singer, Peter. “The Most Good You Can Do: How Effective Altruism Is Changing Ideas About Living Ethically”, Yale University Press (2015 Edition), ISBN: 9780300180275 
• Timmerman, Travis. “Review of ‘The Most Good You Can Do: How Effective Altruism Is Changing Ideas About Living Ethically’ by Peter Singer”, Oxford University Press: The Philosophical Quarterly (2016 Edition, Vol. 66), pg.661-664 URL: https://www.jstor.org/stable/44505379