Lesson 4: A tour of cause areas
The ITN framework is a tool, not an oracle. It doesn’t tell you what to prioritise. It structures a conversation that requires you to supply the inputs—and the inputs are where all the action is. Your empirical beliefs, your moral commitments, and your tolerance for different kinds of uncertainty determine where you land.
What follows is a tour of the cause areas that EA community members most commonly prioritise. For each, we’ll sketch the case for the area—the strongest version of the argument, as its proponents would make it—and then identify the key assumptions the case depends on. These aren’t the “right answers.” They’re where careful application of the framework, filtered through different worldviews, tends to lead people.
Before the tour: taking weird ideas seriously
The cause areas you’re about to read about will not all feel equally normal. Children dying of malaria is a problem any reasonable person can grasp without much philosophical scaffolding. The welfare of factory-farmed chickens, the long-term consequences of advanced AI, the wellbeing of people who won’t be born for a thousand years—these pattern-match more easily to “weird” than to “obviously important.” Two ideas are worth holding in mind as you read.
Radical empathy. Your moral instincts are calibrated by your biology, your culture, and the kinds of beings you can easily picture. They reliably underweight the suffering of those who are physically distant, biologically unfamiliar, or temporally remote. The ITN framework asks who counts—and your gut, left to itself, tends to answer with whoever you happened to grow up around. Radical empathy is the practice of widening the moral circle on purpose: asking who would count if you reasoned carefully rather than felt your way to the answer. It’s much of why the case for animal welfare looks the way it does, and why longtermism is more than a thought experiment.
Fringe ideas. Most ideas that ended up reshaping how we live started fringe. The abolition of slavery, votes for women, the welfare of factory-farmed animals—none were intuitive to most reasonable people at the time, and the people who took them seriously early on were, by the standards of their day, weird. The asymmetry is what matters: dismissing an important idea before its time is a much bigger mistake than seriously engaging with an unimportant one. So when something in this section pattern-matches to “weird,” that pattern-match is information about your priors more than about the argument.
As you read, the cause areas will vary in how comfortable they feel. Notice when one feels strange, and try to separate the discomfort of unfamiliarity from the substance of the argument. The framework can’t do that for you.
Global health and development
The case. Roughly 700 million people live below the World Bank’s extreme poverty line of about $2.15 per day (2017 PPP). Children die of diseases we know how to prevent. The interventions are among the most evidence-backed in all of philanthropy: insecticide-treated bed nets, deworming, direct cash transfers, vitamin A supplementation. GiveWell estimates that top charities can save a life for roughly $3,000–$5,000. The suffering is real, measurable, and happening now.
ITN profile. Importance: very high—millions of preventable deaths annually. Tractability: very high for specific interventions—we know what works and the evidence is strong. Neglectedness: moderate—significant funding already flows here, but cost-effective opportunities remain.
Key assumptions. The case is strongest if you weight: (a) certainty of impact highly—preferring proven interventions over speculative ones; (b) currently-living humans as the primary moral patients; (c) measurable outcomes as the right metric. The case weakens if you think the most important problems are speculative-but-enormous, or if you give significant moral weight to non-human animals or future generations.
Strongest objection. The marginal cost-effectiveness may be declining. As more funding flows to the top charities, the next dollar does less. And direct health interventions, while life-saving, don’t address the structural causes of poverty. Some argue the expected value of systemic change—even though it’s less certain—is ultimately higher.
Animal welfare
The case. Factory farming is, by the numbers, the largest source of suffering on Earth. Roughly 80 billion land animals are raised and killed each year, the vast majority in conditions that would be illegal if applied to dogs or cats. Most of these animals are chickens—and their capacity for suffering, while debated, is increasingly supported by neuroscience. Yet barely any philanthropic resources address this. Annual spending on farmed animal welfare is roughly $200 million—compared to $50 billion for global health, and dwarfed by spending on companion animals (the US pet industry alone exceeds $150 billion per year, more than 750 times the entire global farmed animal welfare budget). The neglectedness is extraordinary.
ITN profile. Importance: potentially the highest of any cause area—if animals’ suffering counts for even a fraction of what human suffering counts for, the sheer numbers are staggering. Tractability: moderate—corporate campaigns have a track record (as you saw in Week 3), and the cultivated meat industry is advancing, but ending factory farming entirely remains a long way off. Neglectedness: very high—this is one of the most neglected problems relative to its scale.
Key assumptions. The case depends critically on: (a) the moral weight of animal suffering—how much a chicken’s suffering matters relative to a human’s; (b) whether animals in factory farms are actually suffering (the evidence strongly suggests yes, but the extent is debated); (c) whether tractable interventions exist. If you think animal suffering counts for nothing, the cause area disappears. If you think it counts even modestly—say, a chicken’s suffering is worth 1/1000th of a human’s—then the numbers still make this an enormous problem.
Strongest objection. Moral weight is the crux. Reasonable people disagree about whether and how much non-human animal experiences matter morally. If you have low credence in animal sentience mattering, the case collapses regardless of the numbers. There’s also a tractability concern: the political and cultural barriers to reforming animal agriculture are immense, and it’s unclear whether current interventions will lead to lasting structural change or merely incremental improvements within a fundamentally exploitative system—a tension explored in Part 3’s section on moral uncertainty.
Global catastrophic risk reduction
The case. Some risks threaten not just millions of people but civilisation itself. An engineered pandemic could kill billions. A nuclear exchange could trigger nuclear winter. Misaligned artificial intelligence could, in the worst case, end humanity’s ability to shape its own future. These events are unlikely in any given year—but the consequences are so severe that even small probabilities translate into enormous expected value.
This category is broad and contains several distinct sub-areas, of which the two most commonly prioritised within the EA community are:
Biosecurity and pandemic preparedness. COVID-19 demonstrated how vulnerable we are to novel pathogens. Advances in biotechnology are making it easier to engineer dangerous organisms. The overlap of rapid technological progress and inadequate governance creates a window of vulnerability. Investments in pathogen surveillance, rapid vaccine platforms, and international governance could dramatically reduce the risk.
AI safety. As AI systems become more capable, the question of whether we can reliably direct them toward human-beneficial outcomes becomes urgent. The concern isn’t science fiction—it’s a coordination and technical problem: how do you ensure that systems more capable than their operators remain aligned with human values? The field is young, the problems are hard, and the timeline is uncertain but potentially short.
ITN profile. Importance: potentially astronomical—if you weight the long-term future heavily, preventing extinction or civilisational collapse has essentially unlimited value. Even if you focus only on currently-living people, a global catastrophe killing billions would be the worst event in human history. Tractability: uncertain and varies by sub-area—pandemic preparedness has relatively clear interventions (better surveillance, better governance), while AI alignment is more conceptually uncertain. Neglectedness: moderate and changing—funding has increased significantly in recent years, especially for AI safety, but remains small relative to the scale of the problems.
Key assumptions. The case rests on: (a) the probability of catastrophe being non-trivial—if you think the chance of an AI-caused catastrophe this century is 0.01%, the expected value calculus looks very different from if you think it’s 10%; (b) whether current interventions can actually reduce the risk—tractability is genuinely uncertain; (c) how much you weight future generations. The case is strongest for those who take a long-term perspective and who believe current decades are unusually important for shaping the trajectory of advanced technologies.
Strongest objection. There are several. First, the probabilities are deeply uncertain—and when you’re multiplying a tiny probability by an astronomical value, small changes in your probability estimate swing the conclusion wildly. This makes the reasoning feel fragile in a way that “distribute bed nets, save lives” doesn’t. Second, there’s a tractability concern: it’s not clear that marginal funding for AI safety research actually reduces the risk, versus simply funding people to think about an intractable problem. Third, there’s a reasonable worry about “Pascal’s mugging”—the idea that sufficiently large stakes can justify almost any expenditure, which feels wrong even if the expected value maths supports it.
Longtermism
The case. Most ethical frameworks accept that future people matter morally—a child born in 2080 has the same moral worth as one born today. But the implications of that view are easy to miss. There could be vastly more future people than have ever lived. If we take their interests seriously, then influencing the long-term trajectory of humanity becomes one of the most important things we can do. This is the core idea of longtermism: positively shaping the long-term future is a key moral priority of our time.
Interaction with global catastrophic risk reduction. This is where longtermism reshapes the picture most dramatically. Without longtermism, the case for reducing existential risk rests on the eight billion people who would die—already enormous, but a fixed number. With longtermism, the same risk reduction also protects every future generation that would otherwise have existed. Max Roser’s “The future is vast” works through one defensible scenario: if humanity survives for a typical mammalian species lifespan of a million years at a stable population of around eleven billion, roughly a hundred trillion future people are at stake. Longtermists don’t invent the case for catastrophic risk reduction—it stands on near-term grounds—but they multiply its expected value by orders of magnitude.
ITN profile. Importance: depends almost entirely on how much weight you give future generations. If you give them serious weight, the numbers swamp everything else. Tractability: the most contested dimension, and worth unpacking. Longtermist work splits roughly into two strategies—surviving (preventing extinction or civilisational collapse) and flourishing (making the long-term future go well even if we get there). Surviving is what catastrophic risk reduction targets, and its interventions are relatively concrete: pathogen surveillance, AI alignment research, nuclear de-escalation. Flourishing is broader—work on the values embedded in transformative technologies, on lock-in risks, on civilisational resilience, on the governance structures that will outlast their designers. Forethought’s “better futures” research argues that even if humanity survives, the gap between a mediocre future and a flourishing one could be vast—and that flourishing-focused work is potentially of comparable expected value to survival-focused work. The two strategies differ in tractability: surviving has more concrete intervention pathways and at least visible progress; flourishing relies on longer, harder-to-defend causal chains from today’s action to outcomes centuries out. Some longtermists argue we’re at an unusually pivotal moment—the development of advanced AI, the deployment of new biotechnologies, the institutions being built around them—which lifts tractability for both kinds of work right now. Neglectedness: very high for both, but particularly extreme for flourishing-focused work; what exists clusters in a handful of subfields, mostly survival-adjacent.
Key assumptions. The case rests on: (a) future people having full or near-full moral weight—a view most ethical traditions endorse in principle, though they disagree about how to compare future people with present ones; (b) some of today’s actions having durable, predictable effects on the long-term future—the “trajectory change” claim, much harder to defend than (a); (c) the present being unusually important, what some call the “hinge of history” hypothesis. Each is contested, and weakening any of them weakens the case considerably.
Strongest objection. The deepest worry is epistemic. We can be reasonably confident that bed nets prevent malaria. We are far less confident that any particular intervention today will reliably improve the world in 2500. Critics argue that under this much uncertainty, longtermist reasoning collapses into speculation dressed in expected-value clothing—the maths only “works” because we’re allowed to imagine arbitrarily large numbers of future people. Related worries include cluelessness (we struggle to predict the second-order consequences of our actions this decade, let alone in 2500), fanaticism (the willingness to act on tiny probabilities of vast payoffs), and the political risks of placing enormous weight on people who can’t push back. Many longtermists take these objections seriously; some respond by adopting a “weak” longtermism that prioritises near-term, robustly good actions whose long-term effects are also positive.