Eliciting donations through causal closeness

A slightly longer version can be found on my blog.

There’s a decent-sized psychological literature investigating factors which influence charitable donations. One of the most prominent and important such results is donors’ apparently robust preference for ‘identifiable victims’ over ‘statistical victims’. It looks as if the more detail about the identifiable victim, the more money is normally elicited. Schelling (1968) surmises: “the more we know, the more we care”.

This makes me think of other sources of motivation for donating to a charitable cause, and other ways in which we might “know more” about the effects of our donations. There’s a bundle of considerations which effective altruists have pointed to as (morally illegitimate) reasons people may feel averse to giving. I wonder if it’s possible to abstract a common thread from these reasons. If it already has a name I haven’t come across it, so I’m calling it ‘causal closeness’. Roughly speaking, the consequences of an act can feel more or less causally close depending on how ‘direct’ versus ‘diffuse’ they are. This might vary at least along the lines of (1) space, (2) time, (3) certainty, and (4) number. At one extreme, saving a drowning child from a shallow pond is direct in every sense: the child is right there in front of you, there is no need or even possibility of waiting to help, the survival of the child hinges on your actions with near certainty, and exactly one identifiable person receives your help. Consider the opposite corner of this range. (1) The intended consequences of your act can take place far away from you; like most donations from high-income countries to charities operating in low-income countries. (2) You might have to wait a long time for the tangible consequences to unfold; as in patient philanthropy and longtermism in general. (3) You might even think those consequences are positively unlikely to transpire even if the outcome is valuable in expectation (think how discussions of existential risks are shrouded in so many unknowns). (4) Finally, the intended consequences of our acts might benefit many people a small amount, rather than a few people a great deal. The benefits of choosing to reduce emissions are spread across millions of people; but the benefit per person is going to be extremely small. What I think all these examples have in common is a sense of ‘indirectness’, ‘diffuseness’, ‘disconnectedness’ or something like ‘causal distance’. I don’t intend to mean anything technical here; my goal is to get across a common psychological flavour or feeling.

You could say that EA thinking is motivated and characterised to a large extent by the recognition that these forms of indirectness don’t morally matter (think Parfit and Singer). But I want to suggest that building in a sense of causal closeness or directness might represent an opportunity to elicit donations and boost motivation.

Basically every effective charity involves some or all of the forms of indirectness listed above. When I donate, my money typically falls into a large pot. Since money is fungible, in such cases it’s not even meaningful to ask which payments out of that pot came from my donation or somebody else’s. As such, there’s no meaningful way to ‘trace’ the consequences of my donation to a particular person, location, purchase, time, or event. To the extent that my donation is no more associated with any particular output of a charity over any other, there is a sense in which the benefits of my donation are ‘spread’ across many people, places, and times. My guess is that this numerical ‘diffuseness’ is the most underexplored source of this feeling of indirectness.

Here’s a hypothesis: most people (myself included) would be more inclined, all else being equal, to donate more to a charity where donations feel more direct or causally close (along all four dimensions I suggest, holding the expected value of a donation constant). In this way, a charity might benefit from providing some (artificial) way of tracing my donation to a particular person, location, purchase, or event. Some charities already do this, to varying degrees. In exchange for a £60 donation, Toilet Twinning sends the donor a certificate with a photo of a newly built toilet with a photo and GPS coordinates. I’ve seen these certificates proudly displayed in office toilets in the UK – they seem like a nice, gratifying way for a company to flout their philanthropic credentials. GiveDirectly offer a newsfeed of unedited stories from recipients of their cash transfers, although individual donations are not tied to recipients. Animal adoption charities basically commodify this donor-recipient closeness by selling glossy images and profiles of the particular animals you’ve chosen to adopt.

This leads to a question of how charitable donations might implement ‘causal closeness’ with a view to eliciting donations. Clearly, an outcome that is necessarily distant in time or space cannot normally be made to happen any sooner or closer. Equally, some outcomes just are downright uncertain or risky. It therefore looks as if the most tractable avenue for engineering causal closeness is in the feeling of numerical ‘spread’ (versus identifiable outcomes). This presumably involves making records of particular events or purchases, e.g. by taking photos or collecting quotes from recipients, and then matching them with donors. I have in mind a setup where, for instance, a £100 donation to AMF comes with an (optional) email containing a picture of the village where ‘my’ money bought some nets.

There’s a worry that this might introduce a kind of white saviour type paternalism – maybe there’s something objectionable about making it look as if the worse-off recipient of a donation depended on you, the well-off donor; or of donors ‘collecting’ or ‘displaying’ those people. The plain costs could also make this a non-starter for extremely lean, ‘gimmick’-free charities. Then there’s the technological challenge of finding some not-entirely-arbitrary way of dividing individual donations between individual outcomes. One way to make that easier might be to offer some larger fixed-sum donation which allows for a one-to-one mapping of donation to recipient or donation to outcome, in the model of Toilet Twinning. Finally, some interventions just rule out any way of mapping donations to particular outputs, because their outputs are not discrete in the right way; such as in flour fortification. Because of these complications, implementing a kind of ‘causal closeness’ is not going to be viable for all or even most effective charities. That said, I think the benefits of artificial ‘causal closeness’ might extend beyond eliciting more donations in general to motivating those people already set on giving significant amounts of their time or money to effective charities. Effective altruists are people too, and an obvious and endemic barrier to effective giving is the absence of gratifying, identifiable feedback of the kind you do get when you’re face-to-face with others in e.g. a volunteering role. My suggestion is that in some cases these direct links can and should be engineered back in, where the costs of doing so are justified by the boost to donations and motivation.

More generally, I think a concept like ‘causal closeness’ (better name desperately needed) may be useful because, at least to my lights, there is something in common between people’s intuitive moral aversion to spatial and temporal distance, uncertainty, and widely ‘spread’ harms and benefits which isn’t captured by a single term (to my knowledge). Note again this is not supposed to be a technical concept, but a way of pointing at some vague, common psychological feature.

I have some questions about this. Firstly, does anyone know of any research that might be relevant for what I’ve called ‘causal closeness’? In particular, do people tend to give more to charities which offer a way of tracing particular effects of their donations rather than paying into a big fungible pot? In general, do people in some sense prefer actions or choices whose effects are narrowly focused rather than ‘spread’ across many people or things? This sounds like a textbook behavioural econ result, but I don’t have the Google skills to find it. Secondly, does this actually make sense and sound reasonable? Have I missed something obvious?

Thanks for reading!