34 | Paul Bloom on Empathy, Rationality, Morality, and Cruelty
What this covers
Paul Bloom, a psychologist at Yale who wrote Against Empathy, joins Sean Carroll to argue that empathy—the act of feeling what others feel—is a fundamentally flawed moral compass, and that we would do better guided by reason and what Bloom calls rational compassion. The conversation ranges across moral psychology, the nature of cruelty, and the relationship between emotion and rational deliberation. Bloom contends that empathy is intrinsically biased and innumerate: it draws us toward those similar to ourselves and focuses our moral attention narrowly on single individuals while leaving us indifferent to larger numbers. He argues instead for moral decisions grounded in principles and reasoning, supported by social institutions designed to counteract individual cognitive biases. Carroll pushes back, defending empathy's role in preventing rational people from ignoring inconvenient facts about those unlike themselves.
A second current running through the exchange challenges a widespread assumption about cruelty: that it stems from dehumanization. Bloom argues the opposite—that much human cruelty arises from a full recognition of others' humanity, turning on our capacity to humiliate and moralize. He draws on experimental work by Tage Rai showing that when harm is framed as moral punishment rather than instrumental, we exaggerate the target's human qualities to justify hurting them. The pair also discuss whether humans are as irrational as psychology typically portrays them, whether moral realism is true, how partisan politics corrupts reasoning, and what moral progress over centuries suggests about the power of deliberation. Carroll raises the specter of System 2 skepticism—the worry that conscious reasoning is largely window-dressing—while Bloom argues that reason, properly institutionalized, has demonstrably shaped human behavior for the better.
Bloom argues that empathy—feeling what others feel—is a biased, innumerate moral guide that we are better off replacing with rational compassion, and that much human cruelty stems not from dehumanization but from a full recognition of others' humanity.
- Empathy zooms in on single, similar individuals and is innumerate, making it a poor and exploitable basis for moral and policy decisions
- Reason, embodied in social institutions and 'System 2' deliberation, is more powerful than many psychologists credit
- Cruelty and punishment often arise from moralizing humanization rather than dehumanization
Empathy for people unlike oneself is essential to rational moral reasoning and theory-building.
- Empathy for people unlike oneself is a crucial tool of rationality: self-described rational people often attend to some facts and ignore others precisely because they differ from those affected, so to build a sound theory of morality and justice one must work hard to see things from the perspective of people very different from oneself—for example men struggling to grasp the fear women experience walking alone at night.
“if you're really going to be rational, it is absolutely crucial that you are empathetic, that you try to understand what other people who have very different experiences than you have been going through.”
Moral argument works pragmatically without requiring belief that morals exist independently of human construction.
- Carroll argues that even granting one can reason with very different people from common moral intuitions, it is a mistake to leap from that to moral realism; one can accept that morals are constructed by human beings, find the common ground, and reason from there—so the practical practice of moral argument does not require a realist ontology.
“I think it's a huge mistake to then leap from that to say, we need to be more realists. We don't need to be more realists, we just have to find that common ground and reason from there.”
Realistic machines and sex robots will pose serious ethical problems within a decade, though true Turing-test passage remains distant.
- We are not far from machines that could fool people not trying hard, and increasingly realistic machines including sex robots will arrive sooner or later, raising serious ethical problems within roughly a decade; but passing a rigorous Turing test remains far off, with the final 5% of realism—as with self-driving cars—being the hardest part.
“we're not that far from machines that you could be tricked into thinking are human if you didn't try that hard”
Racism is bad.
- Racism is bad.
“Racism is bad. Yeah. That's an official Mindspace platform.”