Rob Wiblin and Russ Roberts on Charity, Science, and Utilitarianism
What this covers
Rob Wiblin, host of 80,000 Hours, interviews Russ Roberts of EconTalk in a conversation spanning charity, evidence, and moral philosophy. Roberts argues that most consequential decisions—where to donate, which career to pursue, which policies to enact—involve irreducible uncertainty that data cannot adequately resolve. He contends we should be skeptical of attempts to optimize good through measurement and favors instead local, nuanced action over grand aggregating schemes like utilitarianism. Wiblin represents the effective altruism worldview, which uses frameworks like the Importance-Tractability-Neglectedness Framework to direct resources toward high-impact causes, while Roberts challenges whether such systematic approaches can work in practice.
The conversation ranges across several interconnected problems with evidence-driven charity and decision-making. Roberts emphasizes how empirical findings, even when presented cautiously, become distorted in public memory—he illustrates this through the deworming case, where a celebrated randomized controlled trial was later undermined by meta-analysis, exposing the fragility of confident recommendations. He argues that certain human goods—parenthood, kindness, true impact—resist reduction to measurable scalars, and that interventions scaling from local to national or international levels founder on information limits and misaligned incentives among institutions. A third strain critiques appeals to expertise and data themselves: common sense is unreliable, elaborate theory can diverge from reality, and empirical research is often low quality, so no single source of knowledge deserves privileged status. Against this, Wiblin argues that specialization enables genuine expertise—decades spent understanding pandemics yield real knowledge—and that reasonable people can act under uncertainty using Expected Value reasoning, accepting incomplete information as inevitable. The two also grapple with whether local identity and tradition, though sometimes parochial, enable moral achievements that cosmopolitan frameworks might undermine.
Roberts argues that because most consequential decisions—charitable, career, and policy—involve irreducible uncertainty that data cannot resolve, we should be deeply skeptical of attempts to optimize good via measurement, and should favor local, nuanced, decentralized action over grand aggregating schemes like utilitarianism.
- Empirical evidence (e.g. deworming RCTs) is often misleadingly confident and fails to replicate
- Many of life's most important goods (parenthood, kindness, impact) cannot be reduced to a measurable scalar
- Interventions that work at the family/local level cannot be reliably scaled to national or international levels due to information limits and bad incentives
Radical uncertainty pervades impact assessment; empirical research, theory, and common sense all have major flaws warranting epistemic humility.
- The public latches onto specific charity recommendations as if they were settled science, but the more you focus on these issues the more you realize how little we know and how clueless we are about the effects of our actions.
“Where I'm more skeptical though, is the idea that we can use science, or statistics, or data to reliably hand out that money effectively.”
- Career advice is an art not a science; the goal is not to measure impact precisely but to provide tools and information that help people make marginally better decisions under massive uncertainty, while avoiding the trap of over-analysis when one should just start and learn by doing.
“all we're trying to do is provide tools and information that might help people make a marginally, an incrementally better decision than they might otherwise.”
- There are three ways to understand the world—prior beliefs/common sense, careful theory and reasoning, and formal empirical research—and all three have major problems: common sense is sometimes completely off base, elaborate theories can fail to resemble reality, and empirical work is often low quality, so one should not single out empirical research for skepticism.
“Common sense is sometimes just completely off base. You can build elaborate theories that just don't resemble reality at all. Randomized controlled trials can be done improperly and there's lots of just incredibly low quality empirical research.”
- The genuine problem is not that growing skepticism will make people reject science and become anti-evolution fanatics, but that people trust science too much and are then let down when it cannot deliver on excessive promises; sciences can deliver knowledge within reason but should not be trusted excessively over common sense and priors.
“I think the real problem is people trusting something too much. And, then they're let down when it can't deliver on these excessive promises.”
- GiveWell recommended deworming charities while explicitly stating there was a high probability the impact would be much smaller than the original study suggested or even nothing, justifying it as good value in expectation because it is so cheap and has no negative side effects—a reasonable expected-value way to make decisions under hazy evidence.
“they said there's a 20% chance that it does work like this and there's another 20% chance that it has a smaller impact--if you do the expected value calculation, given how cheap it is, and given that doesn't really have negative side effects, we think it's good value in expectation”
Pressing problems demand specialized expertise and community division of labor to make marginal progress despite individual limits.
- 80,000 Hours advises people to contribute to pressing global problems judged by scale (how many affected and how much), neglectedness (low-hanging fruit), and tractability, then to seek leverage by addressing the binding bottleneck—ideas, money, or skills—often via government policy or fundamental research, while finding a role of strong personal fit and building career capital early.
“we have various rules of thumb for trying to figure out what global problems are especially pressing, like how many people are affected by the problem and how much. How many people are already trying to solve the problem--so was it neglected and might there be low hanging fruit still there?”
- Problems like reducing US-China conflict risk are unfathomably large for any individual—understanding the issue would consume a whole career—but a community of hundreds or thousands of well-meaning, careful people can divide the labor, with thinkers producing research that guides practitioners in government, making real marginal progress on a human timescale.
“That's why you need a community of people. It's an unfathomably large problem for one person, but by the time you properly understood the issue... then your career would already be over. But, I think, if you do have hundreds of people, thousands of people, then you can have some of them spend their time thinking”
- If you try to follow every pressing problem in the world you will understand none of them well enough to act, so specialization is a good reason to focus—someone who spends decades studying pandemics can develop genuine expertise and suggest things that help at better than even odds.
“that's a good reason to specialize when you're trying to do good. If you try to follow every pressing problem in the world then you're going to end up properly understanding none of them”
Expanding the moral circle to all sentient beings and reducing catastrophic risks are justified by moral progress and technological risk acceleration.
- One of the most reliably valuable things humanity could do is continue expanding the moral circle so that all conscious beings with welfare get moral consideration, building on the historical trend from caring only about white men to including women, people of color, and eventually all sentient beings.
“You've had kind of an expanding circle of moral consideration. And I think one of the most important things that we could do... is to continue expanding that circle”
- People have become somewhat more moral over time, but technology and our ability to do horrible things—such as destroy the world through nuclear war—have advanced far faster than our wisdom, kindness, and prudence, creating an unprecedented and dangerous gap that justifies focusing on global catastrophic risks.
“we've also become technologically much more advanced and our ability to do horrible things has increased much more than say has our wisdom or our kindness.”
- Effective altruists are ultimate pragmatists: if promoting greater concern for everyone would actually make people more selfish or do less good, that would be a good reason not to pursue it, and even the possibility of such backfiring is a mark against going that far.
“we're kind of the ultimate pragmatists. So, inasmuch as trying to increase concern for everyone is actually going to result in people being more selfish, or not doing more good, that would be a good reason not to pursue that.”
Personal decisions about impact and wellbeing involve poor feedback and cannot be mastered; comparisons across lives are possible only at extremes.
- Global health advocates argue the real tradeoff is helping a known friend versus a stranger who needs a cheap measles vaccine or antibiotics—remaining developing-world problems where fixes can be scaled without deep local knowledge because antibiotics cure bacterial infections in most places, and where the poor cannot afford things that would greatly help them.
“They've tried to find remaining issues in the developing world where it does actually seem like there are fixes that can be scaled without having that deep local knowledge... we kind of think that antibiotics help to cure bacterial infections most of the time in most places. It's not super-contextual.”
- Despite the limits of surveys, a striking correlation between pre-decision answers (do you enjoy being around kids, are you already enjoying life) and later satisfaction could help someone identify which reference class they belong to and thereby make a marginally better decision about having children.
“I think that could help give me some idea of what reference class am I in? Am I in the reference class of people who say that they are super glad that they had kids and they have no regrets?”
- Living a good or impactful life cannot be mastered like golf or chess because the problem is harder, feedback is poor, and you only get one go; a better analogy is being a CEO who can improve at difficult on-the-fly decision-making but never masters it, since even the best make many mistakes.
“you definitely can't achieve mastery in career choice in the way you can at golf or sports... the feedback is very bad; and you only get one go.”
- While aggregating happiness into precise numbers is suspect, clear personal experience—being a cheerful person versus a friend with major depression, or stubbing a toe versus a catastrophic car accident—shows that in principle we can compare levels and differences of wellbeing across people, at least at the extremes.
“I think that I can say with some confidence that I'm happier. My subjective experience is better than someone who has, like, major depression.”
- In policy decisions one should presume in favor of inaction out of respect for autonomy when the consequentialist pros and cons are finely balanced, intervening with taxes or regulation only when the welfare gains are enormous relative to the costs—and one should also avoid the worst downsides like tyranny rather than optimizing for the average outcome.
“I think that we should have a presumption in favor of inaction, because I also care about autonomy. And, if the positives and negatives from a consequentialist or utilitarian point of view are very finely balanced, then it seems like you should just leave people alone”