The hummingbird should have pecked the elephant

Epistemic status: thinking out loud about something I had permanently inked on my body at 19, so calibrate accordingly. Also note, written in a personal capacity. My views, not my employer's.

TL;DR: The hummingbird parable is beloved because it validates individual effort. But there is an elephant standing right there with a trunk that holds eight litres. The highest-value action available to the hummingbird is not carrying another drop of water, it is annoying the elephant until the elephant helps. The hard part is working out which actor is the elephant (which I think is a candidate-generation problem, not a maths problem).

On my right ankle I have a tattoo of a hummingbird carrying a drop of water.

It comes from a story Wangari Maathai used to tell, usually described as a Quechua or Japanese folk tale (though the provenance seems a bit murky). The forest is burning. All the animals flee to the edge of the forest and watch horrified. All except the hummingbird, who flies back and forth to the stream, carrying one drop of water at a time. The elephant asks the hummingbird what on earth he thinks he can achieve, and the hummingbird replies: I am doing the best I can.

I got a visual representation of the story because I found the ethos beautiful (and I was 19, and who doesn't do things like that at 19). But I was watching a film about the California wildfires the other day, and realised I have started to disagree with my own tattoo.

Doing what you can is great, but sometimes we need to take a step back and ask whether we are doing the best thing we can. There is an elephant in the story. It has a trunk that holds something like eight litres. It is standing right there, doing nothing, just talking to the hummingbird. The single highest-value action available to that hummingbird is not carrying another drop of water, it's annoying the elephant until the elephant starts carrying water. Whatever the hummingbird achieves with its beak is a rounding error against what happens if the elephant moves.

When this occurred to me, my first thought was "oh look, I've reinvented effective altruism". But it might be a bit more complicated than that.

Is this actually EA?

To be fair to EA, leverage reasoning is squarely inside the tent. People (e.g., 80,000 Hours) have argued for years that persuading two other people to do the high-impact thing beats doing it yourself, and multiplier arguments of that shape are standard. So "don't just carry water" is not news to anyone here.

But I think the elephant is doing something slightly different from a multiplier. The first version is essentially replication: recruit more hummingbirds, get more drops. The elephant is not a large number of hummingbirds. It is a categorically different actor which already possesses the capacity you lack, is already present, and is currently choosing not to use it.

That distinction is more or less the whole content of the pillars of support model. I've argued this at length with co-authors in Humanities and Social Sciences Communications (paper here): power rests on institutions which can be peeled away, and movements should target those pillars rather than the general public. The Birmingham campaign in 1963 worked because it made trading impossible for the white-owned businesses the city government actually listened to. The organisers did not recruit their way to victory. They changed the incentives of an actor who was already in the room.

So the two traditions converge on "don't (necessarily) just do the object-level thing yourself," and then diverge on what to do instead. One says recruit and multiply. The other says find the existing pillar and move it.

Where they come apart

Most of the time these recommend similar actions. The interesting cases are actors where size and decisiveness are wildly asymmetric.

Take a national authority that has to approve novel food applications, like cultivated meat or precision-fermented dairy. It's small: a few dozen staff, a modest budget, no ability to shift consumer behaviour, no campaigning capacity, no money to give you. Score it on scale and it looks like a poor target, and score it on tractability it looks worse, because it's a technical body with formal procedures rather than a public-facing one you can pressure.

But it holds a veto. Nothing reaches the market without it. Its throughput determines whether an entire product category exists in that jurisdiction for the next five years.

A multiplier framing struggles here, because there is nothing to multiply. You cannot recruit your way into a regulator, and influencing forty people is not a scale story. A pillars framing sees it immediately: this is an institution the current equilibrium depends on, so the question is not how big it is but what it is load-bearing for.

The obvious objection is that a competent expected-value calculation simply prices the veto in. That's right, but my claim is narrower: the failure happens at the candidate-generation stage, not the arithmetic. Pillars analysis asks a structural question (what does the status quo depend on?), and that question surfaces targets which a scale-first search may never generate as candidates to evaluate in the first place. You can't price in a veto-holder you never listed.

Does that framing survive scrutiny? Or does the disagreement dissolve once you accept that any serious prioritisation exercise has a search step and the search step was always the hard bit…

Two things the parable gets right for the wrong reasons

There are two readings on which the hummingbird comes out looking better than my argument so far allows. Both of them work by saying the parable recommends the right action and gives the wrong justification for it.

1. It might be a claim about morale rather than impact

Maathai told this story to people running very long, very unpromising campaigns. Read as decision theory it's poor. Read as a device for stopping capable people from giving up, it might be excellent. Those are different jobs, and a story can be bad at one and good at the other.

Which does a movement need more of? (I suspect the answer varies with how close the movement is to burnout!)

2. The visible effort might itself be the lever

The hummingbird is carrying water in front of an audience of animals who have all decided nothing can be done. Sparkman and colleagues found in 2022 that Americans underestimate support for climate policy by nearly half: actual support runs at 66 to 80 per cent, and people estimate it at 37 to 43 (paper here). We are collectively wrong about each other in a way that suppresses action.

If that's the situation in the story, the hummingbird's drops are not doing hydrological work. They're doing norm-signalling work, and the parable's own accounting misses it entirely. The hummingbird justifies itself with "I am doing the best I can," which is a claim about effort. The actual case for the behaviour is that it is the cheapest available way to demonstrate, to an audience of paralysed bystanders, that the thing can be attempted at all.

Which means this section partly rescues my tattoo, but not on its own terms. Carrying water is right because it is how you annoy the elephant, not because doing your best is inherently valuable. That distinction matters, because it tells you when to stop: if the drops are buying you an audience, keep carrying. If nobody is watching, put the water down and go find something with a trunk.

I have a small piece of anecdotal evidence for this. For a while I wore an Extinction Rebellion badge, and I chose it deliberately over something like "ask me why I'm vegan," because the XR hourglass is obscure enough that people ask what it is. As a direct intervention, a badge is a drop of water. As a device for getting strangers to start the conversation themselves, it worked better than anything else I tried.

Naming the elephant is the hard part

"Find the high-leverage actor" is easy to say. In practice, most of the disagreement inside movements is about who the elephant actually is, and people are extremely confident about answers they cannot evidence. Governments, retailers, caterers, investors, the media.

Here is the best I currently have, which is two steps rather than one.

Step one: systems mapping, to generate candidates. Causal loop diagrams and stakeholder maps are how you find out what the equilibrium depends on. Who supplies what to whom, where the reinforcing loops sit, which nodes appear in more than one loop. However, mapping tells you where the structure is, not where the slack is. It will show you that supermarkets are central, but it will not tell you whether any given supermarket is movable this year.

Step two: scale, neglectedness and solvability, but applied to stakeholders instead of causes. The framework transfers, but the definitions need translating:

• Scale becomes load-bearing weight: how much of the current equilibrium collapses if this actor defects? Not how big the actor is. This is what the regulator example is about.
• Neglectedness becomes unattended: is anyone already working this actor? Movements pile onto the same three visible targets and leave procurement managers, insurers, standards bodies and trade associations essentially uncontacted. This is where I'd expect the real returns to be, and unlike most of this it's cheaply measurable - you can just ask around.
• Solvability becomes movability: does this actor have a decision-maker who can say yes, an incentive that can be changed, and no structural commitment to the status quo? A regulator legally obliged to consider your evidence is more movable than a much friendlier company whose margins depend on the thing you want to end.

Those three questions at least discriminate between actors, which is more than vibes plus a stakeholder map.

The gap I can't close is that movability is mostly only knowable by trying. Which, if true, argues for running cheap probes across many actors before committing resources to any one of them (rather than picking your elephant from the map and going all in). That's a testable strategic claim and I don't know of anyone who has tested it. If someone has, I'd like to read it.

The awkward bit

You cannot (cheaply and painlessly) edit a tattoo.

I have a permanent, visible commitment to a position I've partly talked myself out of, which is a fairly good metaphor for what happens when you hold public positions for a living. So tell me where I've got this wrong, and I'll add it to the list of things I can't remove.

(Also, the flames around the hummingbird look a bit like the Hot Wheels logo, which is a whole other issue.)