TL;DR
In a controlled experiment, people asked to empathise with a specific sick child were significantly more willing to move her up a medical waiting list ahead of children who needed it more — and they agreed, when asked, that doing so was unfair.1 That’s the whole problem in one finding: empathy reliably produces action, and just as reliably produces partial action. It operates like a spotlight, which means it needs a particular face and goes dark on everyone outside the beam.2 The fix isn’t to feel less. It’s to notice that compassion and empathy are different faculties, and only one of them scales.2
Sorting donations is not the same faculty as being moved by a photograph, and only one of them scales. Photo: Joel Muniz on Unsplash.3
The Assumption Nobody States Out Loud
Almost every modern appeal to be better rests on a hidden premise: that moral failure is a failure of feeling. If you’d only see the refugee, hear the victim, sit with the suffering, you would act rightly. Cruelty is diagnosed as a deficit of empathy, and the prescription is always more of it.
The premise is intuitive and mostly wrong — or at least far narrower than it sounds. Empathy in the strict sense means coming to feel something like what another person feels. That is a real, powerful, measurable capacity. What it isn’t is a mechanism for figuring out the right thing to do, and the gap between those two things is where a lot of well-intentioned harm lives.
The Waiting List
The cleanest demonstration comes from C. Daniel Batson’s research on empathy-induced altruism. Participants learned about a ten-year-old girl with a fatal illness who was on a waiting list for treatment that would relieve her pain, and they were given the power to move her up the queue. Some were first prompted to empathise with her; others weren’t.
Participants who had been induced to feel empathy were markedly more likely to bump her up the list — ahead of other children who, by the allocation criteria, needed the treatment more.1 The detail that makes this damning rather than merely interesting: those same participants, asked afterwards, judged that acting on partiality was less fair and less moral. They knew. Empathy didn’t corrupt their moral reasoning; it simply outvoted it.
Batson’s conclusion is deliberately unromantic: empathy-induced altruism and the motive to uphold justice are separate prosocial drives. Sometimes they align. Sometimes they pull in opposite directions, and when they do, empathy frequently wins.
The Spotlight Problem
Paul Bloom’s case in Against Empathy generalises this, and his central image is the right one: empathy works like a spotlight.2 A spotlight is genuinely useful — it directs attention and help to where it’s needed. But it has properties that are disastrous in a moral instrument.
It requires a specific target. You cannot empathise with a statistic; you need a face, a name, an imagined individual. It is therefore innumerate by construction — it cannot weigh a hundred sufferers against one, because it can only illuminate one at a time. It follows existing affiliation, so it runs hotter for people who look like us, share our side, or are already cooperating with us, which is a polite way of saying it is biased toward the parochial.2 And it is short-sighted, favouring the vivid present harm over the diffuse future one.
Bloom’s proposed alternative is not detachment but rational compassion — caring about people’s wellbeing and wanting them to thrive, which unlike empathy does not require a single identifiable individual to lock onto, and can therefore be extended deliberately to people you’ll never meet and cannot picture.2
Where the Popular Version of This Argument Overreaches
Here’s where honesty requires cutting against my own thesis, because the single most-cited piece of evidence for empathy’s innumeracy has badly weakened.
The famous demonstration was the “identifiable victim effect”: people donate more to one named, pictured child than to statistics describing millions, and — in the version that made the point unforgettable — donate less when a second child is added. It’s a beautiful finding and it has not held up. A 2023 reanalysis using updated methods detected moderate evidence of publication bias in this literature and strong evidence for the absence of an identified-victim effect, alongside failed replications of the seminal studies. A 2024 preregistered replication of another foundational experiment produced results “largely inconsistent with the original,” finding no significant effect of identifiability or singularity on willingness to give.4 The earlier meta-analytic support was weak to begin with — a correlation of about r = .05.4
So the specific claim “people give less as victims multiply” should be treated as unproven, not established. That matters, because it’s exactly the sort of vivid finding that gets recycled indefinitely after the evidence has moved. The structural argument about empathy needing a particular target survives — Batson’s waiting-list result is a different paradigm and doesn’t depend on it — but the tidiest version of the story does not.
Empathy Is Also Not One Thing
The other honest complication is definitional, and Bloom’s critics have a real point. Bloom argues against empathy narrowly construed as vicarious feeling. But there’s a second capacity usually filed under the same word: cognitive empathy — accurately modelling what another person thinks and wants without necessarily catching their emotional state. That one looks much more like a genuine moral tool, and one review of Bloom’s argument criticises exactly this narrowness, holding that his conclusions are at best overstated and that cognitive empathy — trained, cultivated, less prone to burnout — produces more useful understanding than emotional contagion does.1
Which points at the actual distinction worth keeping. Feeling-with is a signal: it tells you something matters. It doesn’t tell you what to do, and it systematically misreports scale, priority, and who counts. Understanding-that plus caring-about — Bloom’s compassion, or “intelligent kindness” as one critic frames it — is the part that can survive contact with numbers.1
The Practical Upshot
None of this argues for feeling less, and it certainly doesn’t license the callousness that sometimes borrows this argument’s clothes. Someone who feels nothing for a suffering person in front of them has a different and worse problem.
The claim is narrower: when empathy and a fairness principle disagree, that disagreement is data, not noise, and the fact that empathy feels more moral in the moment is precisely why it deserves the extra scrutiny. The person who moves the sick child up the list is not being wicked. They’re being moved. And the test of a moral framework isn’t whether it makes you feel the right things about the person in the beam — it’s whether it still holds when you remember the people standing in the dark.
