1. The wrongness of an offense against a being scales with the dignity of the being. Dignity is a slightly hard thing to pin down, but it refers to something like how worthy the being is of good treatment. It is worse, Ethan claims, to mistreat a human than a non-human animal, even if you cause them comparable suffering. It is similarly worse to mistreat a virtuous human than a vicious one, because the virtuous person merits better treatment. Thus, when we offend against God, we merit infinite punishment.

2. We shouldn’t really trust our moral intuitions about hell, because many smart and ethically reflective people disagree with us about them. Holy saints Padre Pio and Alphonsus Liguori both thought they deserved hell. When you have a moral intuition that conflicts with the moral intuitions of other people, absent some special reason to think you are more reliable, you shouldn’t really trust yours over theirs.

Unsurprisingly, as a person who thinks the doctrine of eternal hell is profoundly wicked, I didn’t buy either of these arguments. I thought it would be worth explaining why.

I don’t actually share the intuitions that Ethan’s first argument is based on. I don’t think that harms to humans are worse than comparable harms to non-human animals. In fact, I think there are decisive arguments against this view (and I should have another piece out soon giving an argument that’s even more decisive). Similarly, I don’t believe in desert, but even if I did, it doesn’t follow from the fact that bad people merit bad things that the badness of an offense against someone scales linearly with their greatness.

Imagine that there was an infinitely good being. Suppose you stole a pack of gum from them. I don’t think that merely because of their infinite moral status that would make you infinitely deserving of punishment. This is especially clear if we imagine your action doesn’t lower their welfare at all, and you know in advance that it won’t.

Ultimately, if you believe that the badness of an offense scales with the virtue of the being offended against, I think that you should think it reaches some kind of limit. Alternatively, if you think it does scale linearly, you should think that maximally virtuous beings are not infinitely virtuous in the way needed for offenses against them to merit infinite punishment. It isn’t at all clear why a being that is perfect would have infinite virtue—if they do the right thing in all cases, why would they have infinite dignity rather than, say, only twice the dignity of someone disposed to act virtuously in half of cases?

The deeper problem with Ethan’s argument is that it has crazy implications. Applied consistently, it conflicts with many of the deepest moral intuitions. I mean, first of all, if I am confident in anything, I’m confident that nice people that I know shouldn’t be made to suffer for all of eternity. Insofar as the view conflicts with that judgment, I can know it to be false.

Second, it conflicts with the judgment that suffering in ordinary cases is bad. Ethan’s view is that most people go to hell. Even the people who don’t go to hell deserve to. But deserved suffering is good. By this standard, it seems that all else equal, we should want there to be more suffering in the world. Whoever invented anesthetic did a terrible thing—reducing suffering. Cluster headaches and even torture inflict 0% of the suffering that their victims deserve. When you prevent this kind of suffering, by this standard, you are doing something very bad. You are preventing deserved suffering.

Imagine a criminal justice system that treated people this way. If people masturbated, they got tortured in the most brutal way imaginable. On Ethan’s view, it’s hard to see what would be wrong about this. If anything, it would be too lenient because it would only inflict upon people 0% of what they deserve.

Now, here’s one possible response: perhaps it’s not up to us to punish people for their crimes against God. While people deserve infinite punishment, we don’t have the right to inflict it. It’s a bit unclear precisely why this would be—the state has the right to punish all sorts of offenses retributively. Why couldn’t it punish infinitely severe offenses? If someone tortures a child, we don’t ordinarily think that it isn’t up to us to punish them. Yet on Ethan’s view, we all actually merit infinitely more suffering than people standardly think child abusers merit—for people generally think child abusers only merit finite punishment.

In addition, even if you cannot rightly punish those who merit infinite punishment, it at least seems wrong to benefit those deserving infinite punishment. It would be wrong to bake a cake for Hitler, because he doesn’t deserve a cake. So it seems if we applied this consistently, we’d get a general prohibition on doing nice things for people.

Lastly, the view seems to imply that suffering is good, even if we’re not justified in carrying it out. If Hitler suffers, on the standard view according to which he deserves suffering, that is good even if private actors aren’t licensed to punish him. So this view seems to imply that suffering is, in general, a good thing. And if suffering is good, then we shouldn’t try to prevent it.

I think we can reject this view as self-evidently wrong. Think about some time you suffered really intensely. Is it at all plausible that this was a great thing? Is it plausible that if the suffering was more intense and lasted forever, that would be very good?

Here’s one last worry about Ethan’s argument: it seems to imply that it would be permissible to punish us in hell to any arbitrary degree. If we merit infinite punishment, then no matter how much we suffer, we’ll never have gotten what we deserve. By this standard, then, God could permissibly make every second in hell contain toolbox killer-level torture (if you find that intuitive, I encourage you to read more about what they did to their victims).

In fact, God could permissibly make our suffering start out as more intense than all the suffering collectively experienced so far in human history, and have its intensity double every millisecond. In fact, even if he did this, it would be too lenient—there would be no single time during which we’d ever have gotten any fraction of what we truly deserve.

I accept that we’re a lot worse morally than we tend to think. But surely this is not plausible. Surely no one could deserve this kind of thing. And if we are this rotten—so that even the best of us have such darkened hearts that we merit infinite torture—it’s unclear why God would go about making us in the first place. It doesn’t seem one should create a class of beings that are each so wretched that they merit infinite punishment.

Even setting aside these moral intuitions, I find there to be something theoretically inelegant about this account. Humans are complex, having both good impulses and bad ones. It would be a bit odd if this patterned behavior—neither fully dark nor fully light—merited simple unending torture. And if it does, if we really are so bad that we should be tortured forever, allowing anyone into heaven seems deeply immoral. If Hitler deserves infinite suffering, you shouldn’t grant him infinite reward. If we deserve infinite suffering, you shouldn’t grant us infinite reward!

Before I turn to Ethan’s second argument, I feel there’s a hubris that lots of non-infernalists have when it comes to infernalism. We feel as though we are sane and enlightened—opposed to suffering, correctly recognizing that torment is a terrible thing. But I want to remind everyone: torturous suffering lives on earth today as a result of ordinary people’s meat consumption, and years of it can be stopped for a dollar.

The callousness that we display toward non-human animals is every bit as extreme as the callousness of the infernalists. Concerning indifference to suffering, we’d do well to focus on our own malfeasance before chiding our infernalist neighbors. Hell lives on earth today. It is in our power to stop it. We all ought to do considerably more.

Sorry for the proselytizing about animals, returning to hell, Ethan tries to debunk our contrary moral intuitions by noting that they’re not shared by various smart and virtuous people. This, Ethan claims, should make us treat our moral intuitions as merely defeasible evidence. This evidence is then outweighed by the amazingly awesome evidence for Catholic miracles—evidence that’s good for a Bayes factor of 10^60, apparently. Ethan thinks that the odds of having evidence as good as the evidence for Catholic miracles, given atheism, are about equal to the odds of throwing an atom-sized dart across earth randomly and then hitting any particular atom—then winning a 10,000,000,000 person lottery.

Now, suffice it to say, I disagree that the evidence for Catholic miracles is that good. I think no evidence for anything is ever that good. As an atheist, on priors, I would be a lot more surprised to win the 10 billion person lottery and hit a random atom on earth than to have reports of some hard-to-explain healings. And I haven’t been much impressed by the one Catholic miracle I’ve seriously investigated (though I am a bit spooked by Padre Pio—some skeptic should really do a deep dive on him). But for what it’s worth, I agree that if you have evidence that’s good for a Bayes factor of 10^60, that is more than enough to overcome pre-existing moral objections.

Here’s a first reason I don’t buy Ethan’s skeptical argument against trusting anti-Catholic moral intuitions: I think it applies just as much in the non-moral domain. If I should distrust my moral intuitions because smart people don’t have them, then surely Ethan should distrust his photogrammetric analysis of the Fatima photographs when smart people disagree with him. Similarly, he should distrust his extreme confidence that Padre Pio did miracles, when other smart people who have examined the evidence disagree.

I am a lot more confident in my ability to do philosophy than to investigate miracles. So if there’s a conflict between philosophical judgments and miracle-investigating, I’ll generally favor the philosophical evidence. In any case, skeptical arguments cut both ways. The judgment “God wouldn’t torture people forever” seems a lot more straightforward than anything about Padre Pio!

Second, the judgments that I’m giving aren’t exactly contentious judgments rejected by large numbers of people. They’re things like “suffering is bad,” “it’s good, all else equal, for people to have less painful surgeries rather than more painful ones,” “if you got tortured the way the toolbox killers tortured their victims for all of eternity, that would be bad,” and “it’s good to help people and bad to hurt them.”

Even the judgment that people deserve to suffer forever in hell is one that, as best I can tell, virtually no non-religious person has ever endorsed. I’m not very impressed if people endorse some crazy moral judgment because they are religiously committed to it. That tells us little about how counterintuitive it is. People can talk themselves into anything.

My final point is the one that moves me the most but I find hardest to defend: sometimes some verdict is sufficiently clear that you can simply see that you are right and those who reject it are wrong. Suppose, for instance, that I was sent back in time to America in the 1830s. Those around me insisted in large numbers that people with black skin simply were intrinsically less morally important. Even if learned people who were otherwise virtuous were split between my judgment and theirs, I would still be justified in believing with great confidence that they were wrong. It’s just so clear!

Similarly, Bryan Caplan thinks that the badness of one’s pain depends on how smart they are. Bryan is a smart guy—but if he and I were the only people on earth to form judgments on this matter, I could still be confident that he was wrong. Sometimes you can just see that you’re right, even if you can’t convince other people. I feel the same about the judgment that it would be wrong for a perfect being to make people suffer intensely for all time. When I think about that judgment, I just know it’s correct.

Here’s one way of motivating this: the idea that you should take seriously ethical disagreement from other smart people depends on philosophical assumptions. And sure enough, these are plausible assumptions. But they’re way less plausible than the assumption “it’s wrong to make people suffer forever—and having most people suffer forever wouldn’t be part of a perfect divine plan.” So if I’m forced to choose between the two, I’d give up almost any conceivable judgment about peer disagreement long before believing in an eternal hell that most people go to at the hands of a perfect God.

There are other moral judgments that Ethan believes in which I think I can know are false in the same way. For example, if I know anything about ethics, I know that it would be right to use contraception in order to prevent everyone on earth from being tortured for a million years. That judgment is sufficiently clear that even if smart and holy people disagreed with it, I wouldn’t give it up. If Ethan can hold that his reading of the miracle evidence is certain to be right, even though peers disagree with him, I don’t see why I can’t say the same about other obvious moral judgments.

Now, this should only be an option of last resort. In general, you should be worried about declaring any contentious moral view you have to be self-evident—and not changing it even in the face of strong contrary argument. But some things really are self-evident, and that God wouldn’t torture people for infinite time is one of them.

Hanania’s evidence for the populism to kakistocracy pipeline is both considerable and convincing. He starts by defining populism. Populist leaders are those who oppose elites and blame them for social problems. Elites, as Hanania defines them, are those in contact with long-lasting established institutions with traditional power.

The influence of the elites comes from their connections to those institutions. They gain status by impressing other elites. New York Times columnists are elites because they gain prestige from connection to an established journalistic outlet. Joe Rogan isn’t an elite, though he has a lot of influence, because his influence comes from connection with a mass audience, rather than with a traditional institution.

So what’s not to like about these populists?

Start with the empirical literature. On every measurable outcome, populists do worse. On average, electing a populist leader shrinks GDP by 10%. This holds across a wide range of measures of who counts as populist. The world’s economically worst-performing countries are all either war-torn African or Middle Eastern nations or Latin American countries with long-lasting populist control.

Populist governments expand executive authority and undermine existing institutions. Electing populist leaders is associated with an increase in corruption and democratic backsliding. Populists tend to rewrite constitutions at high rates, erode freedom of the press, and undermine civil liberties. 40% of populist leaders get charged with corruption—a fact particularly remarkable in light of their consistent remolding of the judicial process in their own image. And you shouldn’t attribute this academic opposition to populism as a byproduct of left-wing bias, given that academics find similar—and often worse—effects of left-wing populists. In addition, studies that compare places where populists won to places where populists almost won often show similar negative effects of populist leadership.

Yet one doesn’t need to consult studies to see how populism leads to kakistocracy. Just consider America today. Trump, the quintessential American populist, appointed RFK Junior, who holds a diverse assortment of crankish medical theories. That has not stopped him from being appointed to head HHS, where he has haphazardly dismantled research into mRNA. RFK Junior is a nice encapsulation of kakistocracy, given that he may be among the worst people on the planet to run HHS.

It is as if the Trump administration has been single-mindedly aiming to confirm the kakistocracy thesis. This is an administration that peddled easily debunked election lies and used them as an excuse to try to overturn the results of a fair election. Four years after Trump did this, he was elected president.

While Trumpism is based around the claim that traditional elites are stupid, corrupt, and immoral, Trump routinely manifests these vices to degrees that border on the infinite. As Hanania points out, despite Trump’s repeated promise to “drain the swamp,” he has been personally enriched by the presidency to a degree never before seen in history. Don Jr. founded a DC-based club with a $500,000 membership fee alongside Trump’s envoy to the Middle East and Trump’s crypto czar. When Hunter Biden was rumored to have done similarly, this dominated headlines for years on end; Trump’s son does it out in the open along with multiple high-ranking administration officials, and it barely makes the headlines.

Similarly, a Trump-branded real estate project in Qatar was announced weeks before Trump visited the Gulf States. Shortly after this, an Abu Dhabi-backed fund made a $2 billion business deal using the digital coins of Trump’s firms. Purchasing Trump-backed crypto allows people to easily buy leverage. And Hanania didn’t even mention a number of other instances of obvious Trump corruption—say, the fact that he was gifted a jet by Qatar! It is maddening that when a politician who has run on a platform of stamping out corruption engages in historically unprecedented corruption on a daily basis, nobody seems to care.

Trump has claimed repeatedly that American elites are stupid and incompetent. Yet his administration replaced competent elites with a ridiculous clown show of Trump sycophants who could be found at the local asylum. A priori, it would be extremely surprising if the most qualified people to run various complex federal departments included various unaccomplished graduates of low-ranking institutions without any relevant experience whose primary pre-appointment credential was their seemingly limitless zeal for kissing Trump’s ass.

Does anyone really believe that Pete Hegseth is the most qualified man in America to lead the Department of War? Does anyone believe that Christopher Wray—from Yale Law School no less—is less qualified than Kash Patel? Patel got his JD from Pace University, “which is not ranked among the top hundred law schools in the country and whose incoming students currently have a median LSAT score that is at the fiftieth percentile of all test takers.”

Patel’s primary credential was a willingness to publicly embarrass himself with North Korea-style displays of public loyalty, including selling Trump shirts and children’s books that detail Trump’s greatness. In addition, Patel sold quack supplements rumored to reverse the negative effects of vaccines. It would be extremely surprising if the nation’s most proficient sycophants also happened to be the best people to work various important jobs.

Trump is the most well-known populist leader, but other populist leaders have similar tendencies to induce disaster. Maduro obliterated Venezuela’s living standards. Chavez’s economic policies were similarly disastrous. When Sri Lankan populist leader Gotabaya Rajapaksa banned importation and use of synthetic fertilizers and pesticides, against the advice of all experts, this led to a nearly 50% increase in price. Illegitimate attempts to hold onto power are similarly common among populist leaders: Yoon Suk Yeol of South Korea declared martial law to try to maintain power and Jair Bolsonaro inspired a violent attack on his nation’s capital (sound familiar?).

Nor is it uncommon for populists to promote crankish conspiracy theories. Călin Georgescu of Romania was broadly opposed to the practice of importing things (an opposition that is, unsurprisingly, not shared by economists). One of the funnier sentences in the book notes that Georgescu “has questioned whether the moon landing was a hoax, claimed that carbonated drinks contain nanochips, and once asserted that residents of the Marshall Islands used to live between 150 and 200 years before they suffered the ill effects of US nuclear tests.”

Russian populist elites are also, at fairly high rates, either quacks or plagiarists—with one having made the rather interesting claim to have invented the “nooscope” “‘a device that scans transactions between people, things and money’—a breakthrough on par with the microscope or telescope.” Big if true! Perhaps he was employing the unconventional inference rule “induction from other words that have ‘scope’ attached to the end.”

This raises the question: why are populist leaders so terrible?

One reason is that they gain power by appealing to a generally uninformed public. Insofar as one hates elites, one won’t have any inclination to defer to elites. But if the public doesn’t know much and doesn’t defer to elites, the only people the public favors are those who endorse extremely simple solutions. That is why populist movements tend to espouse simplistic narratives—being more likely to believe high housing prices are the result of nefarious megacorporations than restricted housing supply. This is also why they blame problems on immigrants—being much more likely than economists to believe immigrants are behind economic woes. One doesn’t need to understand the world much to blame other people for their problems.

Another reason populists are so demented—explored by Hanania at length—is the cancerous media environment that births populist leaders. Populist news media promotes sensationalist news stories. The American right doesn’t have much interest in reading traditional news sources—preferring podcasts and other forms of alternative media. Despite the Wall Street Journal being ostensibly right-wing, it is read by more Democrats than Republicans.

Traditional news media, though imperfect, tends not to make outrageously false and easily debunked claims. Yet alternative media selects for sensationalism; this begets a non-stop firehose of conspiracy theories and general craziness. Joe Rogan, America’s most listened-to podcaster, has promoted each of the following stories in the last few years:

There is an ancient city beneath the Giza pyramids; HIV does not cause AIDS; there were advanced human civilizations that predated those accepted by archaeologists and historians; 9/11 may have been a government operation; mind reading is real; Covid vaccines are more dangerous than the disease itself; and humans became more susceptible to polio due to vaccination.

Another darling of the alternative media world, Candace Owens, has over 6 million YouTube subscribers. She has claimed, among other things, that Macron’s wife is a man, that some odd Jewish sect called Frankism has orchestrated numerous major events in world history, and has even suggested that some of her political opponents may be robots. This is one reason why highly shared right-wing news stories are much likelier to be false than highly shared left-wing news stories.

The book has lots of other interesting content. Hanania discusses and criticizes a number of major theories of why populism arises. He discusses some cases of populist successes—usually under South American economic libertarians like Milei and Fujimori. He discusses many more instances of populist leaders being corrupt and crazy. Hanania is offering a year-long paid subscription to his Substack if you pre-order the book. If you want to learn more about the populist descent into madness and badness, it is hard to think of a better book.

It’s been annoying seeing supposedly serious public intellectuals make such basic and obvious errors. It’s really a stark reminder that success as a public intellectual can exist alongside severely impoverished intellect. Worst of all, these basic mistakes have been weaponized in defense of a very bad cause—the child-killing decimation of foreign aid. The recent debacle has been illustrative of how profoundly politics can warp people’s brains and cause them to behave in a way that is both immoral and idiotic.

1 Did anyone die?

It began with Elon doing what he often does—saying something false:

Now, it is obvious that if you “feed into the woodchipper” a bunch of programs that provide medicine to sick children in large numbers, this will cause a non-zero number of deaths. This holds even if you think it was good policy overall. Every major policy change causes a non-zero number of deaths. If Elon is not aware of this fact, he is sufficiently ignorant about how the political world works that he shouldn’t be trying to influence it. Fortunately, Nicholas Kristof came with the receipts. He listed the names of many of the children who died as a result of the cuts.

You’d think that this would be an open-and-shut case. Elon says no one can name names. Kristof, who did the reporting, names some names. Case closed. Sadly, it has not been. The response from Elon’s countless defenders has been simply to change the subject. Here is one of the top replies:

Remember the topic of dispute: Elon said no one died from the cuts. Kristof shows that’s wrong. There’s a separate debate to be had about how blameworthy Elon is—whether what he did is akin to killing vs merely failing to save. But that is separate from Kristof’s simple point that some people died. All talk about Elon’s degree of culpability is an obvious distraction. Konstantin Kisin weighed in to say:

This is why no one in any Western country can cut any welfare spending ever and why we keep going further and further into debt.

Because any time anyone reduces spending on anything, an example can be wheeled out of a person who suffered or even died as a result.

Obviously one should sometimes enact policies that lead to some people dying. But if you have better memory than a goldfish, you will remember that Kristof’s point was in response to Elon disputing that cutting USAID led to any deaths. How culpable that makes Elon is a separate topic. Kisin followed his initial commentary up with, “I am making the point that saying a policy led to the deaths of innocent people is not a sensible way to analyse complex policy debates.”

This is confused. It is obviously relevant whether a policy led to deaths. That doesn’t settle whether it’s good policy, of course. Every policy will lead to some deaths. But it’s one of the things worth considering. If a policy leads to a million deaths, that makes it worse than an otherwise similar policy that does not. Kisin seems unable to distinguish between “we should weigh lives lost when evaluating policy” and “any policy that leads to any deaths is bad.”

2 Morally culpable?

Here is how the moral debate has been playing out on Twitter:

Liberal: Elon Musk killed millions of people.

Conservative: LOL, as if. Not saving people isn’t the same as killing them. By this logic, you have killed lots of people by failing to give to effective charities.

I think both of these positions are a bit too quick. Undeniably, Elon Musk did something that led to a lot of people dying. But whether his actions count as killing is a matter of philosophical debate. Still, I think there’s a strong case to be made that his actions were more like killing than failing to save. Even if they’re more like failing to save, they were still very bad.

First, why might his action count as killing? Well, traditionally Congress has power over spending and shutting down agencies. So there’s a reasonable case to be made that Elon behaved unlawfully. Even if he did not, he behaved in some ambiguous gray zone of legal activity.

Notably, Elon wasn’t just choosing not to give his own money to save lives overseas. He was blocking existing lifesaving government programs in a way that was both haphazard and illegal. If someone is driving an ambulance to a hospital and you direct them to stop working. This leads to the death of the person in the ambulance. Presumably, you are on the hook for killing someone.

Alternatively, imagine that there’s a hospital that is saving a bunch of lives. You seize control of the hospital by taking a series of legally dubious actions. You order the hospital to stop treating patients. Arguably, you are on the hook for killing. What you did isn’t the same as simply failing to save people.

One reasonably accurate test for whether a person is responsible for killing: this: would their alleged victims have been alive had they not acted instead of doing as they did? Ted Bundy killed people, because had he not taken the actions that he did, his victims wouldn’t have died when they did. Failing to donate is different because it doesn’t leave you counterfactually responsible for any deaths. There aren’t extra people dead because of how the person who failed to donate acted.1

By this standard, Elon killed people—had he never existed, there are lots of extra people who’d be alive today. If you want to see a more detailed case for this being the correct analysis, see my article here. Remember, there were people left with medical devices inside of them because of Musk’s actions. Surely, taking active steps so that medical devices are left inside people midway through treatment isn’t exactly the same as not giving your money to charity.

But let’s suppose Elon’s actions weren’t equivalent to killing but were more like letting die. Does this mean he’s off the hook? No. It’s bad when people die. The only reason millions of people aren’t going to die (probably) is that Elon wasn’t successful in destroying foreign aid. And still, his actions very likely led to large numbers of predictable deaths.

If the U.S. can prevent very bad things from happening at small cost, it should. It would have been good to stop the Rwandan genocide, for example, if we could have done so easily. Similarly, one of the big upsides of intervening in World War II was stopping the Nazi holocaust. It is good to prevent bad things from happening, even if you don’t stop every bad thing in the world. If a hundred thousand children were drowning in pools, and you eliminated the program that would have pulled them all out, that would be very bad, even if it isn’t the same as killing.

Common-sense morality isn’t consequentialist. It holds that consequences are among the things that matter, but they’re not the only thing. In terms of bad consequences, the USAID termination is just about the worst policy that happened in my lifetime. So if consequences matter at all, this was very bad. And remember: the impact on the federal budget of DOGE’s activities was negligible, and the impact on the deficit was negative. So this was a very large number of deaths caused without any upside.

Now, for what it’s worth, I do think that ordinary people are culpable for failing to donate effectively. Your donations can save lives. When you don’t donate, it’s like you’re walking past a drowning child. You are letting someone die when you could have saved their life.

This doesn’t mean we are obligated to spend every single penny on donations. If there were enormous numbers of drowning children—too many to save them all—you wouldn’t have to forego all pleasure and dedicate your whole life to saving them. But you would be obligated to make saving them a major life project. If giving to life-saving charity is similar, then we’re obligated to do a lot more than we mostly do. I hope you’ll join me in giving away at least 10% of your income to effective charities (though I give a good amount more) and taking the Giving What We Can pledge.

3 How many deaths?

Did Elon’s actions lead to millions of deaths? Some early studies predicted that they would.

But these generally had serious methodological problems and assumed a total dismantling of foreign aid rather than the mere sizeable cuts we’ve actually seen. We’ve also gotten some new evidence against the early extreme projections of USAID deaths. Data from South Africa doesn’t show an uptick in deaths.

I grant that this is some evidence against the higher-end estimates. But in light of earlier estimates predicting many deaths, the amazing effectiveness of foreign aid in general, problems with the data, and specific reporting on deaths caused, we’re still probably looking at high estimates of deaths. Also, South Africa has more functional institutions than other African countries, so they’re better able to respond to cuts.

Ultimately, this leads me to be more skeptical of the highest estimates. But we’re still likely in for a lot of deaths. We’ll have to wait and see when data comes in from other countries. Even if it only leads, on the low end, to 100,000 annual deaths—a lot lower than many of the estimates—this would still be almost an Iraq war’s worth of deaths over the course of four years. That’s a lot of pointless death.

Someone well-versed in statistics should try to come to a comprehensive estimate of lives lost from USAID cuts. Though it is great news that it looks like some of the highest estimates were too high.

4 Other errors

Many other influential people have jumped in to say wrong things. For example, here is a statement from Gad Saad—perhaps the single man in America whose confidence is least accurately apportioned to his competence.

Reply: obviously Kristof does not have the power to prevent all deaths anywhere in the world. That claim is ridiculous. Also, when there are many different actions that save lives, you aren’t culpable for all the ones you don’t pick—you can, after all, only pick one. This isn’t like axing DOGE, because the money didn’t go into saving comparable numbers of lives. In addition, as described before, there is a distinction between not saving lives and actively sabotaging existing life-saving programs.

Gad is evidently trying to make a parody argument—that the same moral calculus that says Musk is blameworthy also says everyone in the world is blameworthy for every death. But this parody is unsuccessful because the factual claim he’s making is obviously false, he ignores the distinction between doing and allowing, and he ignores the obvious fact that there’s a big difference between not spending on some life-saving programs in order to spend on others, and not spending on any life-saving programs.

Reply: that was not Kristof’s claim. Kristof noted that when Elon abruptly shut off the services providing fuel for ambulances, leading to someone dying, that meant his actions caused a death. Apparently noting the effects of actions is “the most grotesque moral blackmail” Rufo has ever seen. Speaking of Rufo:

Yes, it is a “hidden premise” that African countries often have bad governments that aren’t able to fund life-saving services. This is a dark truth that they don’t teach you in lib school—you have to go to UATX to learn it. Chris Rufo is the first to uncover this worldview-destroying bombshell.

Liberals have indeed asked why nations are dysfunctional. In fact, there are a number of books on the subject. Most are by liberals. Surely Rufo knows this. But whatever the cause of nations’ dysfunctionality, it is still good to save lots of people’s lives. Analogy: if a child is drowning because their nation is too dysfunctional to put fences around pools, it is still good to pull them out of a pond so that they don’t drown.

This is one thing that irritates me about many right-wingers. They act like there is some collection of dark truths that they know which liberals won’t admit to. Usually, these “truths” are either banal and widely-known or false. They then act like correctly analyzing every single policy depends on knowing these truths—usually without any explanation of why this would be. Their attempts to explain the connection between these hidden “truths” and policy tend to be full of edginess and innuendo but wholly bereft of substance.

If you actually investigate this claim, what happened was that the Taliban stole a tent provided by USAID. USAID did not buy a tent for the Taliban. I think it is a bit alarming that the richest man on Earth provides a non-stop firehose of easily-debunked lies.

5 Conclusion

The current discourse has been very dumb. Here’s the right view to have:

1. Whether Elon’s actions are like killing is morally ambiguous. It’s neither exactly the same as shooting someone nor as simply failing to give to life-saving charity. Probably, though, it’s more like killing.

2. Even if it wasn’t killing, it was still very bad.

3. Exactly how bad remains to be seen. Probably the highest-end estimates are inflated.

4. Elon’s actions definitely caused some deaths. Punting to discussion of 1 in response to this claim is a distraction.

5. It doesn’t follow from the fact that Elon behaved badly that we’re obligated to give all our money to effective charities. Still, we are, in fact, obligated to give more than we mostly do.

I also feel like there’s a pretty profound missing mood from right-wing reactions. There are children currently dead because of Elon’s cuts. We can debate, as a policy matter, whether providing foreign aid was worth it. But you’ve lost a bit of your soul if the dead kids don’t weigh at all on your conscience—if your first thought, when reading about the kids who died, is how it can be spun into a snippy lib-owning remark. The attitude towards the dead children hasn’t even been feigned sadness; not a scintilla of somberness has been even attempted by those defending the cuts. Such behavior is pathological and inhuman.

1 Donation matching

A funder is temporarily matching donations to Animal Charity Evaluators’ (ACE) movement grants program. For those unaware, ACE is basically the GiveWell of animal welfare. They do high-quality research to find the best charities in the animal sector. They publicize which charities are effective and give out grants to effective animal charities. You can see some of the grants they’ve awarded here—many go to incubate fledgling organizations with the potential to be very impactful. Their movement grants program has been endorsed by Giving What We Can.

The basic pitch for giving to the grant matching challenge is straightforward: animal welfare is a really big deal. By the billions, animals undergo hellish torture in factory farms—their bones are broken, they’re boiled alive, they’re selectively bred so they can barely move, they’re kept in crowded conditions where disease spreads rapidly, and so on. To within a rounding error, all the suffering currently in the world is experienced by animals.

Many organizations stop lots of animal suffering for little money, often sparing animals from multiple years in a cage per dollar. In light of this, it makes sense to fund effective animal charities. But you probably don’t have the time to comprehensively investigate which animal charities are the most effective. So it makes sense to give money to people who carefully research which organizations help animals most effectively and give out grants to those organizations.

The matching program is also conditional. A donor has agreed to match the first $300,000 in donations. In other words, this donor will give up to $300,000—matched with how much other people give. People have already given $260,000. The donation window runs through the 10th of this month. (Edit: an earlier version of this post said that they would only donate at all if the full $300,000 was raised. This was false, sorry! They’ll match however much was given).

In addition, a group called Mobius has agreed to match every new monthly recurring donation, up to $10,000. So if you set up a new monthly recurring donation, your gift will be matched twice!

I’ve chipped in $500. My best guess is that each dollar donated spares many animals from a lifetime of suffering. So please, if you can, donate.

2 Organizer support program

As is known, this blog is very hip and cool. A sizable chunk of its readers are in university; this is because my readers tend to be young, and smart (also, in general, very cool). Oh, and they also tend to be deeply moral. Thus, inquiring minds want to know: how can they make a big impact on the world while they’re still in university?

Here is one good answer to that age-old question, asked since the days of Gilgamesh: you can be an organizer for your university’s effective altruism group. Oftentimes EA events are organized by only a few dedicated people across an entire university, and it may be the case that your university has no existing EA group at all (SAD!). This could be the case even if there’s plenty of latent interest among the student body, just waiting to be tapped. If you get just a few more people involved in EA through your organizing, this will have enormous value. And if the people you get involved are as impactful as you, just one new recruit can double your lifetime impact.

Friend of the blog —who, like most of my readers, is very smart and cool and young and handsome—has a piece on the EA Forum about why people should become university organizers. The whole thing is worth reading, but here’s a particularly important bit:

1. Scope – … a few counterfactual EAs potentially means millions of dollars going to either direct work or effective charities. Getting one more cracked EA involved can potentially double your impact!

   1. According to this post from 2021 by the Uni Groups Team: “Assuming a 20% discount rate, a 40 year career, and $2 million of additional value created per year per highly engaged Campus Centre alumnus, ten highly engaged Campus Centre alumni would produce around $80 million of net present value. The actual number is lower, because of counterfactuals.” It should be noted that campus centre alumni is referring to numbers estimated from these schools.

   2. They also included an anecdote of a potential near-best-case scenario that I think is worth paraphrasing: The 2015 Stanford EA group included: Redwood CEO Buck Shlegeris, OpenPhil Program Director Claire Zabel, Full-Time EA Journalist Kelsey Piper, and former CEO of Redwood and Constellation Nate Thomas. However, the Stanford group went dark in 2016 – for years, there was only one active member and few events were run. “As a result, we probably lost a few Bucks, Claires… Kelseys and Nates. That’s a lot of impact missing.”

And there are a bunch of other benefits. Organizing gives you valuable experience and looks good on your resume. You’re likely to meet incredibly cool people while doing it. The kinds of people who get involved in EA are disproportionately moral and smart; they make very good friends. These connections will also make it easier to remain an active EA by getting you more deeply embedded into the community. Some of my favorite people that I met at university were people I met through the EA club.

Running your university’s EA club, as I found out the hard way, is pretty difficult. I was very bad at it. When I was the president of the EA club at my university, it basically died. This was bad! Fortunately, the Centre for Effective Altruism has a program called the OSP, short for Organizer Support Program. It’s called that because it’s a program that supports organizers.

The program allows you to have meetings with mentors who advise you on how to run your local EA group. This is useful. The mentors have lots of experience with organizing, so they give useful advice. They also provide resources, like lists of readings for fellowships. To get their help, you have to apply. Applications close Sunday. So if you’re going to apply, you should do it soon, through this link.

If you’re not in university, you can still consider starting a local EA group, where you can still get assistance through OSP. There are people in every place that have the potential to become passionate EAs—but somebody has to take the first step in reaching out to them. That person very well could be you. Paraphrasing externality-illiterate tree gremlin The Lorax unless someone like you cares a whole awful lot, the EA group at your university isn’t going to get better, it’s not.

Irrational

If you live in an extremely close swing state, the odds you’ll flip an election are below one in a million. We’ll be generous and assume the odds are one in a million.

Let’s assume the candidate who is better for you is way better for you. In fact, he’ll benefit you by $100,000—an unusually large amount. In this case, voting for him would benefit you by the equivalent of 10 cents. So voting in this scenario would be like walking all the way to the polling station and filling out your ballot just to collect a nickel. That would hardly pass any reasonable cost-benefit analysis.

What if you’re so rich that the candidate would cut your taxes by ten million dollars? Even in this scenario, voting would only be equivalent to making ten dollars. Surely if you’re so rich that a candidate might cut your taxes by ten million dollars, you shouldn’t waste your time haggling over ten bucks!

What if we assume the worst-case scenario? One of the candidates is literally going to kill you if he gets elected, it’s a razor-close election (pun intended, because he’ll kill you with razors), and you’re in a swing state. You have maybe forty years left of life—about twenty million remaining minutes. Let’s say you spend a quarter of that time sleeping. Then you should vote if doing so takes less than fifteen minutes (for voting has a one in a million chance of adding fifteen million minutes to waking life). So even in this absurd homicidal scenario, voting probably wouldn’t pay off, if we factor in time spent registering and so on.

In a local election, prospects might be a bit better. But generally speaking, the smaller an election is, the less benefit you get from the better candidate. If your vote has a one in ten-thousand chance of being decisive, even if the candidate would benefit you by $10,000, voting is equivalent to gaining $1 in expectation. Not exactly a great deal!

So if you’re merely self-interested, it basically never makes sense to vote. However, if you’re concerned with the welfare of others, the equation changes. If the better candidate would benefit millions of people, their victory would be hugely important and make the time you lose by registering and filling out a ballot worthwhile. But if you only care about yourself, then voting isn’t worth it, because you don’t count the benefits to others.

Immoral

Voting for the candidate that is better for you and worse for others is also transparently immoral.

America is a large country (SOURCE???). It has about 340 million people. Morally, you’re arguably permitted to put your interests above others. But you aren’t permitted to weigh your interests 340 million times more than others’ interests. The average human life lasts about 2 billion seconds. So the scale at which you’d need to prioritize your interests over others to vote selfishly would also imply that you would prefer to extend your own life by 7 seconds rather than save the life of a baby. Note that this is conservative because it totally neglects the interests of all non-voters.

Here is an analogy: imagine that there were 340 million people who were each on track to get $1,000 worth of goods, and you were one such person.¹ However, you could also choose to get $2,000 while leaving no one else with anything. Surely that would be immoral. Similarly, it is immoral to vote self-interestedly, if this conflicts with voting for the better candidate. If your vote has any effect, it will make millions of people worse-off on net, even if it makes you better-off. You shouldn’t do that.

This point is clearest in national elections, but it still basically holds in local elections. They’re still large enough that even if you’re pretty extreme in your prioritization of your own interests, you should still vote for the better candidate.

Objections

There are two plausible-sounding objections to my argument, but I don’t think either of them work.

First, suppose that when making decisions, you believe it’s important to consider what would happen if everyone did the same thing you did. Perhaps you’re a rule utilitarian or Kantian of some sort. Now, here’s an empirical claim you might make based on this sort of belief: maybe things would be best if everyone voted in their own interest. That way, politicians would be incentivized to make things better for people in general.

If you think both of these things then you should vote selfishly. That’s because you’d think both:

1. You should act in the way it would be best if everyone acted.

2. It would be best if everyone voted self-interestedly.

I have four main objections to this:

1. This isn’t an argument for voting in the way that’s best for you, rather than the way that’s morally best. Rather, it’s an argument for why self-interested voting is morally best. So it’s not an objection to my thesis that when voting, you should only consider morality.

2. It’s very implausible that the right action is the one that would be best for everyone to do. Suppose that if everyone in the universe kicked nearby cats whenever they saw them, that would bring about utopia on earth. However, people aren’t going to do that. Would this fact give you reason to kick nearby cats? Would kicking nearby cats be the right thing to do, even though you know that it would have precisely zero positive effects and would hurt the cat? No, clearly not.

3. It doesn’t seem like voting self-interestedly really is the best thing for everyone to do. Among other things, it will lead to complete neglect of the interests of every non-voter. So really it seems the rule you’d end up with is more along the lines of “vote self-interestedly unless one candidate is better for non-voters.” Your main consideration, then, should be the impacts of the candidates on non-voters.

4. Voting self-interestedly wouldn’t be good if everyone did it. It would lead to 51% of people screwing over the other 49%. So then we must narrow the rule. But then we’ll probably end up with some rule a bit like “you should vote for the candidate you think is best, but you should take the candidate that’s better for you to give some evidence about who is best.”

The second big objection to this proposal: suppose you have some view of decision-theory on which what matters is more than the direct causal effects of your actions. For instance, you might be an evidential decision-theorist. Such people think that you should take the action which is such that after taking it, you expect to have the most utility. This is so even if the action doesn’t actually cause you to have the most utility. If there’s some action which correlates with being rewarded, EDT says that you should take it, even if it doesn’t cause you to be rich.

Now, perhaps you voting for one candidate correlates with them winning. If you’re a standard voter, then you voting for a candidate is evidence that they’re going to win. So arguably EDT licenses selfish voting by making the odds that you’ll influence the election non-trivial.

You might also be a functional decision theorist. You shouldn’t be, but you might be. FDTists say that when making decisions, you should think of yourself as deciding what your “cognitive algorithm” does and pick the action that timelessly gets you the most utility. So if you having the disposition to take a given action leads to you having more expected utility across time, then that’s the action you should take. Arguably, your algorithm being disposed to vote for one candidate influences other algorithms counterfactually, so if you vote for a candidate, their odds of winning are a lot higher.

Still, I don’t think either of these are quite right.

Start with EDT, because it has the nice feature that it gives clear verdicts on account of being a theory rather than a tissue of specific verdicts. Does your voting for a candidate on self-interested grounds give you much evidence that they’ll win? No, not really. It might if you were a totally random voter and had no extra information. But if you’re deciding based on the recommendations of EDT, you’re decidedly not that way.

Imagine that you couldn’t remember if you’d voted for one of the candidates. Would learning that you have—in light of other evidence, like polling results and so on—give you much evidence that they will win? No, not really—certainly not enough to make voting prudentially rational. It might boost the value of voting conditional on you being impartial, but it won’t be enough to make voting selfishly rational.

An intuition pump for this: if your vote has enormous evidential impact, then you voting becomes just about the most important thing in the world. Plausibly, you specifically voting then saves tens of thousands of lives in expectation. I agree that voting is good, but surely it isn’t this good.

It seems about as plausible that your voting impartially correlates with other people being more impartial. So then voting impartially is still good for you, because it makes them more impartial.

Turn next to functional decision-theory. Now, we have a problem here, because no one has any clue what the view says. There’s no robust theory about how to analyze what impact one cognitive algorithm being different would have counterpossibly on other algorithms. But let’s set aside that little problem.

The same basic points apply. Your algorithm endorsing one candidate isn’t really correlated with other algorithms voting for them. You’re different enough that it doesn’t seem that switching your vote to be selfless will affect their votes. And if it is, then it probably also correlates with other people behaving more altruistically. Other people being more altruistic is good for you.

In short, the same reasons you shouldn’t vote selfishly even if you adopt EDT also probably apply to FDT. Though it’s a bit hard to say, because nobody has ever successfully derived a single verdict in the actual world from FDT ever in world history. FDT lives right now as a few judgments about specific idealized cases, a promise, and a dream; it’s not a complete theory.

Conclusion

The basic reason you shouldn’t vote selfishly is that there are a lot of voters. This means your vote has a low probability of having an effect. It’s not worth it prudentially. And it’s immoral to prioritize your interests over those of millions of other people. That’s the straightforward intuition, and neither eccentric decision-theory nor universalization undercuts this core reasoning.

(This article was edited. I’m thinking of hiring the person who edited it as an editor long-term, so let me know if you found the writing noticeably cleaner than usual).

I’ve convinced a number of people that eating meat is extremely horrible—the worst thing they’re doing by far (unless you count failing to give to effective charity). Often, they’re convinced that it’s about as bad as torturing and killing thousands of dogs over the course of their life. This isn’t a crazy comparison because eating meat involves torturing and killing thousands of animals.

Still, most people keep on eating meat. This is a bit surprising. If someone is a non-sociopath, you’d think learning that something they do is profoundly evil—much worse than the next worst thing they’re doing—would change their behavior. If you learned that every time you took some route to work, you drove over an old lady, only a monster would keep taking that route. So why do decent people keep knowingly doing horrendously evil things?

Similarly, people are often convinced by the drowning child argument that when they don’t give their money to effective charities, it’s equivalent to walking past a drowning child. Most of these people still don’t give their money to effective charities. In other words, you can convince a person that some act A is like a different act B which is so immoral that they’d never perform it. Even when convinced, they mostly keep doing A.

Why?

The answer is that most people are conformists. They don’t have that much motivation to do what’s morally right in the abstract. If they learn that the right thing to do diverges from what they want to do, they’ll mostly just keep doing what they want to do. However, people feel bad when they violate social norms or their conscience.

I am convinced that eating hamburgers is a lot worse than being rude to a child in public. Nonetheless, I would feel a lot worse being rude to a child in public than eating a hamburger. How ashamed we feel when we act doesn’t really track our assessment of the act’s wrongness.

Similarly, we sometimes feel sad when we do bad things. If you’re mean to your loved ones, even in private, you often feel guilty. Guilt doesn’t much correlate with assessments of objective wrongness. Learning that failing to give to effective charities is 10x as wrong doesn’t lead to one feeling 10x more guilty.

So the primary drivers of moral behavior are guilt and conformism. There’s no limit to how horribly wrong of a thing people will do, so long as they’re acting in a way that’s socially accepted and they don’t feel guilty about it. Remember, most people will torture someone because they’re instructed to by a man in a white lab coat. Throughout much of history, people have gone to neighboring cities, beheaded the men, then raped and murdered the women.

Conformism and guilt will not save you. Most people are guided by these things. That hasn’t stopped ordinary people historically from raping, enslaving, murdering, and torturing others. These are flawed and low-resolution snapshots of morality. You can’t rely on them to avoid being a bad person or to avoid doing horrendous things.

If you want to avoid doing horrible things, you will have to think hard about morality. This will require actually scrutinizing your society’s practices, rather than telling a just-so story about why whatever your society does is okay. Remember: it was ordinary Germans who led Jews into gas chambers. It was ordinary Rwandans who macheted their neighbors. Being ordinary is not enough. Being able to justify your actions is not enough. You have to be right every time or you risk doing something horrible.

Similarly, it is not enough to note some respects in which the atrocities of the past differ from what our society does. Every atrocity differs from every other atrocity. You can be engaged in grotesque evil without being exactly the same as genocidaires.

There’s another reason you should be moral instead of conformist: otherwise your decision-making hinges on completely ridiculous things. Suppose you’re considering performing some action. Surely what you do shouldn’t depend on what the people around you think about it? If the people around you are wrong, who cares what they think?

We routinely make big sacrifices for the sake of morality. Most people wouldn’t cheat on their spouse even if they could get away with it and would enjoy it. But it’s pretty ridiculous if whether you’ll make a sacrifice for the sake of doing what’s right depends on what other people think about it. I like how put it in Dialogues on Ethical Vegetarianism:

It’s about social conformity. Jefferson “couldn’t give up” his slaves, not because he had some powerful urge to be a slave-master, and not even just because it would be so much against his interests (though it would have been), but because other people in his society had slaves and accepted the practice – that undermined his moral motivation. If he lived today, he wouldn’t dream of owning other people, because it’s so uniformly disapproved.

…

If you think that you can’t do it because you have these overpowering carnivorous urges, or even that you’re just utterly selfish, then it’s unlikely that you’ll make the effort. But once you realize that you make comparable sacrifices to your interests all the time, and it’s not that difficult, then you’re more likely to do it. The reason you make other sacrifices but you’re not making this sacrifice is a really bad reason: not enough other people are pressuring you.

Now, what other people think about your action might affect the cost of the action. But that’s consistent with only taking into account costs and benefits when deciding what to do. Even if torturing dogs would benefit you by more than giving up meat would cost you, you wouldn’t do it. So your behavior really is like that of a puppet on a string—jerked around by whatever your neighbors happen to think. The conformist will often take one act over another, even if the second act is better both for them and the world.

Suppose you’re convinced that you really should think hard about doing what’s right. Where should you begin? I can’t exhaustively describe every potentially wrong societal practice and what I think about it. But I’ll note briefly that some of the things we do that might be wrong include paying for torture multiple times per day, allowing children to die, and entirely ignoring the interests of 99.999999% of sentient beings.

A lot of analytic philosophers are sympathetic to Rationalism (the social movement, not the alternative to empiricism). I don’t know if I’m senior enough yet to count as a philosopher, but I certainly count myself as among those sympathetic. Yet virtually all of them have the same complaint: Rationalists very often make philosophical errors, especially when it comes to decision theory.

The Rationalist community, for those unaware, is a group devoted to forming beliefs rationally. They disproportionately live in the Bay Area, post on LessWrong, think AI is going to be a big deal, adopt various reductionist philosophical views, etc. I’ve written about my thoughts about the Rationalists here—they’re very smart and interesting; they get a lot right but are sometimes overconfident and wrong about philosophy.

The Rationalist decision theory du jour is called functional decision theory (FDT). Academic decision theorists don’t like the theory. The number of academic decision theorists who adopt it could be counted on one hand by someone missing all of their fingers. (Edit: this used to say by someone missing four of their fingers, as I thought that Ben Levinstein adopted FDT. He doesn’t. So I think the number is literally zero!) My position on the view is as simple as can be: I think the view is definitely wrong. It both is sufficiently underspecified so as to give no real recommendations, and also the recommendations that it supposedly gives are extremely implausible on their face.

I have had debates with about 5 million Rationalists on this subject. Half my time in the bay area was spent arguing with people about decision theory. When I sleep, I am haunted by the ghosts of FDTers. If you keep saying some point over and over again, it sometimes makes sense to write it up. I thought I’d do that. But if you want to read more from other people who are better at decision theory than me, and also more sensible and measured, read ’s great piece and also Wolfgang Schwarz’s piece.

2 What the heck is FDT?

(Skip this section if you know what each of the main decision theories are).

Decision theories tell you how to get what you want. Specifically, they tell you how to reason about cases where different options get you different amounts of what you want (the amounts of what you want are measured in units of utility. This doesn’t have anything to do with utilitarianism the moral theory—it just denotes the amounts of whatever it is that you’re optimizing for).

There are two major decision theories that academic philosophers like. One is called causal decision theory (CDT). I’m trying to be impartial, so I won’t tell you that it’s (probably) the correct view. It says that you should take the action that causes you to have the most utility. Specifically, it says that when taking an action, you should ignore non-causal influences that your actions might have on the state of the world and only do what causes the best thing.

There’s a second view called evidential decision theory (EDT). It says that you should take the action which leaves you with the expectation of having the most utility. So when deciding between acts A and B, ask: how much utility would I expect to have if I take A? What about if I take B? If you’d expect to have more if you took A than B, then you should take A. If you’d expect to have more if you took B, then you should take B.

Functional decision theory is different from either. It says you should think of your action as determining the outcome of your decision algorithm. You should take the act which is such that across time, you expect to get the most utility if your algorithm outputs that act.

So EDT asks: what action leaves me with the expectation that I’ll be richest? CDT asks: what action causes me to be the richest? And FDT asks: what action would my algorithm outputting make me expect to be the richest if it was settled at the start of time?

Here’s a famous case to distinguish the theories. It’s called Newcomb’s problem. It’s the most famous dilemma in decision theory.

There are two boxes, A and B. You have the option of either taking just A or both A and B. B has $1,000. One hour ago, a very accurate predictor guessed whether you would take both boxes or just box A. If he predicted you would just take box A, he put $2,000 in box A. If he predicted you’d take both boxes, he put nothing in box A.

Question: should you take both boxes or just box A?

CDTers say: both boxes. Taking the second box causes you to get an extra $1,000. The fact that it correlates with there being less money in the box is irrelevant. By taking one box, CDTers claim, you’re just passing up an extra $1,000.

EDTers say: just one box. If you take just the first box, you’ll generally end up with $2,000 instead of $1,000. EDTers say: you expect to end up with more money if you take one box, so you should take one box!

FDTers say: it depends on how the predictor predicts what you’ll do. Suppose they run your algorithm or an algorithm very much like yours to predict what you’ll do. Well then, by changing the results of your algorithm, you change their prediction. So then you should one-box. The output of your algorithm, then, determines how much money is in the box—FDT thinks of your decisions as determining the results of your algorithm.

But suppose instead that they make predictions by looking at some other characteristic that merely correlates with one-boxing. E.g. maybe they look at whether you had a professor that two-boxed. In this case, FDT says you should two box. The predictor isn’t running your algorithm, so changing the outcome of the algorithm doesn’t change what is in the box.

So what’s wrong with FDT? I have two main gripes: what FDT says is wildly underspecified—there’s no remotely plausible way to fill in the details. Also, the few judgments that FDT supposedly gives are often wildly implausible!

3 FDT doesn’t say anything

The biggest problem with FDT is that it is devoid of genuine content.

3.1 Is there a fact about how other functions would be different in the impossible world where mine was?

FDT says that when taking an action, you should consider how the world would be if your decision procedure gave some recommendation. But what does that mean? Specifically, suppose that you are kind of like me but different in a bunch of respects. Maybe you’re my brother. Maybe you’re Claude Opus 4.7 and I’m Claude Opus 4.6. Maybe you’re an almost exact copy of me. Does changing my algorithm change your algorithm? How could we possibly answer this question?

Remember, my decision algorithm is some mathematical function. So we’re asked to imagine in the mathematically impossible world where some math function outputted something different from what it mathematically has to output, whether other mathematical functions would be different. What could this mean? How could there possibly be an answer to this question? How can you have a theory that depends on there being determinate answers to the question: in the logically impossible world where some necessary mathematical fact was different, how would other necessary mathematical facts be different? What?

FDTers often claim that CDT requires considering counterpossibles too, because it instructs you to hold fixed what the world is independent of your choice and then make the decision that maximizes utility with respect to that. Now, even if this is right, it’s a lot sketchier to consider how other algorithms would be different in counterpossible worlds than just considering irrelevant features of generic counterpossibles. But CDT holds fixed only which things causally depend on your act, not the initial conditions. So it never has to consider a situation where, say, the initial conditions determine that you’ll take some act A, yet you take act B. As Wolfgang Schwarz put it:

For another example, Yudkowsky and Soares claim that CDT (like FDT) involves evaluating logically impossible scenarios. For example, “[CDTers] are asking us to imagine the agent’s physical action changing while holding fixed the behavior of the agent’s decision function”. Who says that? I would have thought that when we consider what would happen if you took one box in Newcomb’s Problem, the scenario we’re considering is one in which your decision function outputs one-boxing. We’re not considering an impossible scenario in which your decision function outputs two-boxing, you have complete control over your behaviour, and yet you choose to one-box. There are many detailed formulations of CDT. Yudkowsky and Soares ignore almost all of them and only mention the comparatively sketchy theory of Pearl. But even Pearl’s theory plausibly doesn’t appeal to impossible propositions to evaluate ordinary options. Lewis’s or Joyce’s or Skyrms’s certainly doesn’t.

And note: this isn’t just some minor quibble with what FDT says in a few cases. This is the core mechanic of FDT. This is what FDT needs to generate a single result in a single case! Every case where FDT gives a recommendation, it does so by analyzing the counterfactual where the output of a mathematical function was different. Insofar as there’s no fact of the matter about that, FDT doesn’t give any recommendations in any cases.

Let’s apply this to Newcomb’s problem. Suppose the predictor predicts what I’ll do by running an algorithm. Presumably it won’t be exactly the same algorithm as the one I’m employing. He’s not running an exact mental simulation of me even if his simulation reliably correlates with what I’ll do. Suppose my algorithm will in fact output one-boxing. FDT requires we answer: in the logically impossible world where my algorithm outputted two-boxing, would the predictor’s algorithm output two-boxing? Clearly there’s no fact of the matter about that! So FDT doesn’t even get clear results in Newcomb’s problem! As long as the predictor isn’t running an exact simulation of you, FDT falls silent on the question of what you should do.

3.2 Statistical correlations aren’t enough

Now, there’s an obvious-sounding solution to this problem. Just consider the nearest epistemically possible world where your decision theory outputs some recommendation, and then tabulate the amount of utility you expect to get. So suppose that you learned that your algorithm was disposed to two-box. Then ask: how much money would you expect to get. Compare that to how much you’d expect to get if you learned your algorithm was disposed to one-box. If you’d expect to have more after learning your algorithm one-boxes than two-boxes, then you should one-box.

But this obvious-sounding solution doesn’t work. It makes the theory into updateless EDT.1 To see this, imagine that some people are born with a gene that correlates heavily with two-boxing. The predictor predicts what I’ll do by looking at whether I have the gene. Two-boxing doesn’t cause the gene or affect whether you have the gene in any way. This solution would recommend one-boxing in this case. If I knew that my algorithm was disposed to one-box, I’d have a high credence in my having the gene, and in my getting rich. But FDT isn’t supposed to say that!

In fact, this leaves FDT vulnerable to the very smoker’s lesion result that FDTers take to be decisive against EDT. Imagine that smoking doesn’t cause your health to be worse. Instead, smoking correlates with having a lesion on your lung that both makes you likelier to smoke and makes your health worse. It seems rational to smoke, because smoking has no effect on whether you have the lesion on your lung. Yet if your algorithm outputs smoking, that makes you expect that you have the lesion, and so it lowers the expected utility that you get according to this solution.

Now, you could modify the view once again so that you only analyze your expectations concerning other algorithms. This way, you wouldn’t look at how much utility you’d expect to get if your algorithm outputted some action. Or, at the very least, you wouldn’t take the action which, if your algorithm outputs, leaves you with the highest amount of expected utility. Instead, when deciding between two actions A and B, you’d imagine:

1. Your algorithm outputting A vs B.

2. What you expect other algorithms to output if yours outputs A vs if yours outputs B.

3. Then you count up the utility from you and other algorithms outputting A vs B. Whichever one leaves you with more utility timelessly (we’ll come back to the timeless thing later) is the one you take.

That way, you only analyze your algorithm’s probabilistic impact on other algorithms. Whether you have lung cancer is not an algorithm. So you don’t treat your algorithm being different as affecting it in the way relevant to decision making.

But this is of no help. Imagine a modified case where the lesion doesn’t make your health worse. Instead, there’s an algorithm that checks to see if you have the lesion. If you do, then it makes your health worse and also makes you likelier to smoke. Now there’s an algorithm in the mix, so this view is back to thinking (wrongly, and contrary to the spirit of FDT) that you shouldn’t smoke. After all, your algorithm outputting “don’t smoke” makes you expect that the other algorithm output “is less likely to smoke and has better health.”

So now the FDTers are in pretty rough shape. They need to have some account of how your algorithm outputting A would affect other different algorithms. But this can’t just be about your credence in the other algorithm having some outcome, conditional on yours outputting A. FDT depends on analyzing how your action being different (counterpossibly) would make other algorithms different (counterpossibly) without looking at how likely other algorithms would be different in the nearest epistemically possible world where yours is different. How could there possibly be a satisfying solution to this problem?

What it needs is some precise specification of how similar two algorithms are that doesn’t depend on:

1. Extraneous factors (e.g. you won’t want to say that how much my algorithm being different affects other agents running algorithms depends on whether they and I make similar jokes).

2. The degree of correlation between ultimate decisions.

But what could it possibly depend on? Isn’t it obvious that there’s no single privileged joint-carving way to decide the similarity of algorithms that doesn’t just look at the statistical correlation between their outputs? Certainly FDTers owe us some account of how this works. It doesn’t do to call it an unsolved problem, when this is the entire engine of the theory—when there’s no plausible story of what a solution would even look like, strong active reason to think there is no such solution, and a solution is needed for the theory to give any result in any case.2

Let’s be a bit more concrete. Imagine that I’m in a prisoner’s dilemma against my twin (note that my twin isn’t exactly like me but is similar). I understand having a credence in my twin cooperating conditional on my cooperating. But if we’re not talking about conditional credences, how could there be a uniquely privileged sharp fact about the non-statistical algorithmic correlation between us two?

3.3 No fact about whether two algorithms are the same

Things get even worse. How do we determine if two functions are running the same algorithm? I’m told this is an “unsolved problem” for FDT. There seem to be a lot of those. And remember, you can’t just look at whether they always output the same thing, because FDT distinguishes between mere correlations and paired algorithms. As put it in his piece:

Even putting the previous issues aside, there’s a fundamental way in which FDT is indeterminate, which is that there’s no objective fact of the matter about whether two physical processes A and B are running the same algorithm or not, and therefore no objective fact of the matter of which correlations represent implementations of the same algorithm or are ‘mere correlations’ of the form that FDT wants to ignore.

…

To see this, consider two calculators. The first calculator is like calculators we are used to. The second calculator is from a foreign land: it’s identical except that the numbers it outputs always come with a negative sign (‘–’) in front of them when you’d expect there to be none, and no negative sign when you expect there to be one. Are these calculators running the same algorithm or not? Well, perhaps on this foreign calculator the ‘–’ symbol means what we usually take it to mean — namely, that the ensuing number is negative — and therefore every time we hit the ‘=’ button on the second calculator we are asking it to run the algorithm ‘compute the sum entered, then output the negative of the answer’. If so, then the calculators are systematically running different algorithms.

But perhaps, in this foreign land, the ‘–’ symbol, in this context, means that the ensuing number is positive and the lack of a ‘–’ symbol means that the number is negative. If so, then the calculators are running exactly the same algorithms; their differences are merely notational.

Ultimately, in my view, all we have, in these two calculators, are just two physical processes. The further question of whether they are running the same algorithm or not depends on how we interpret the physical outputs of the calculator.

Now, as Will notes, standard attempts to measure whether two algorithms are the same generally imply that one system may run many different algorithms simultaneously. If the ultimate account has to do with the mapping between inputs and outputs, then changing the output of your algorithm may have bizarre effects on other features of the world. As Will writes:

For example, if the physical process underlying some aspect of the US economy just happened to be isomorphic with FDT’s algorithm, then in the logically impossible world where FDT outputs a different algorithm, not only does the predictor act differently, but so does the US economy. And that will probably change the value of the world under consideration, in a way that’s clearly irrelevant to the choice at hand.

There’s a related problem. Suppose that there is someone who is psychologically identical to me at all times before Newcomb’s problem. In Newcomb’s problem, they one-box. Should we think of changing the results of my “algorithm” as changing the results of theirs? What could possibly determine this?

There’s a somewhat strange paradox here. Imagine that there’s someone who is psychologically identical to me at all times before the prisoner’s dilemma. I’m in a prisoner’s dilemma against them. They defect. On FDT, I should defect too. But then we’re running the same algorithm. So then I should cooperate. But then we’re running different algorithms, so I should defect.

Now, you might object that the scenario, as I’ve described, is impossible. If I’m basing my decision on theirs, then we can’t be running the exact same algorithm. Here we should imagine that my decision is not based on theirs. We should then consider the question: what action do I have most objective reason to do (instead of which one is best for me to do given what I know).

3.4 Conclusion

So let’s recap. FDT needs a solution to each of the following to give almost any judgment in almost any case:

1. Determining how one algorithm being different would affect another algorithm being different, without depending on the epistemic probability of the second being different if the first was. This also can’t depend on extraneous factors. Remember additionally that we are imagining how other things would be different in the metaphysically impossible world where some mathematical fact is different, and we can’t just rely on epistemic probabilities! This problem seems obviously fatal.

2. Determining whether two algorithms are the same. There is no standard way of doing this, and there are deep reasons to think that any solution to this would have bizarre implications—e.g. on unrelated algorithms that happen to be isomorphic.

Then, even if we had a solution to both of those, FDT would have the problem:

3. It implies that even if two predictive processes are 100% correlated, it would matter which one was used in Newcomb-type problems.

4. It generates a paradox in cases where an algorithm being the same as yours depends on what you do in some situation.

Absent a solution to the first two, FDT isn’t a theory. It’s a collection of suggestions. In every case that has ever arisen in the history of the species and all the standard thought experiments, it is wildly unclear what FDT says. There are deep reasons to think it doesn’t say anything.

4 Should you light yourself on fire for no benefit?

My answer is “no.” FDT’s answer is “yes.” Here’s the case (from , though similar examples abound):

Bomb.

You face two open boxes, Left and Right, and you must take one of them. In the Left box, there is a live bomb; taking this box will set off the bomb, setting you ablaze, and you certainly will burn slowly to death. The Right box is empty, but you have to pay $100 in order to be able to take it.

A long-dead predictor predicted whether you would choose Left or Right, by running a simulation of you and seeing what that simulation did. If the predictor predicted that you would choose Right, then she put a bomb in Left. If the predictor predicted that you would choose Left, then she did not put a bomb in Left, and the box is empty.

The predictor has a failure rate of only 1 in a trillion trillion. Helpfully, she left a note, explaining that she predicted that you would take Right, and therefore she put the bomb in Left.

You are the only person left in the universe. You have a happy life, but you know that you will never meet another agent again, nor face another situation where any of your actions will have been predicted by another agent. What box should you choose?

FDT says that you should slowly and painfully burn yourself to death. After all, having the disposition to do that makes you better off in expectation timelessly. It makes it so that probably there won’t be the bomb in the box in the first place, and you won’t have to pay $100.

But this just seems irrelevant. The bomb is in the box. I have no uncertainty about what will happen if I choose Left. In cases where you have no uncertainty about how the world is, where one action simply leaves you with less utility, you shouldn’t take that action. The fact that this case is rare doesn’t matter! It’s a crazy recommendation of FDT that it tells you to light yourself on fire when you know that if you do so, you will not benefit at all.

FDTers I’ve talked to sometimes have said this is unfairly rhetorically loaded. “It’s not for no benefit,” they claim. “The benefit comes from you being better off if your decision algorithm is disposed to make it.” But at the time you’re taking the act, you have no uncertainty about how the world is. You know what benefits will come about if you take the act: none. So this phrasing is accurate.

And there are an infinite number of other similar examples. Imagine that everyone in the world is put into a deep slumber. Then, the predictor simulates you and guesses if you’ll, thirty years after waking up, painfully cut off your leg for no benefit. The simulation is highly correlated with you, so his guesses about whether you’ll cut off your leg for no benefit are 99.9% accurate. If he predicts that you’ll slice off your leg, he wakes you up. If he predicts that you won’t, then he doesn’t. Assume that waking up is very good for you.

FDT implies that because being disposed to slice off your leg for no benefit makes you likelier to wake up, you should slice off your leg thirty years later. But that just seems crazy. At the time you’re making decisions, you’re already awake. If you’re already awake, it makes no sense to slice off your leg on grounds that it makes you likelier to be awake. The odds that you’re awake are already 100%.

Note: decision theories are theories about rationality. They tell you what decisions are wise and sensible to make. They are not theories about the desirable dispositions to have or about how you should program an AI. There are interesting questions about those sorts of things, but they aren’t what decision theory is about. So don’t think to yourself “would I be better off timelessly having the disposition to slice off my leg?” Think “is it rational, at the time I’m making the decision, to cut off my leg for no benefit.” I think the answer is clear: no! I’ll talk more about this distinction in the next section.

There’s a response to this that I’ve heard from a lot of functional decision theorists. Here’s the idea: you don’t really know if you are the algorithm being simulated in Newcomb’s problem or the actual person. For all you know, you might be the simulation, in which case you outputting “one-box” leads to more utility. I find this response very bizarre:

1. I know I’m not a simulated algorithm. The simulated algorithm isn’t conscious (we can stipulate). I am.

2. Decision theory generally assumes that you’re self-interested. But if I’m the algorithm, then I care about algorithm me—not the version in the real world. So then I wouldn’t care about what the output of the algorithm was.

Thus, I think FDT gives incorrect recommendations.

5 Does FDT get more utility?

A claim that FDTers are fond of making is that following FDT gets you the most utility. Take the version of Newcomb’s problem where the boxes are transparent, for example. So in this case, you can peer into both boxes and see how much money is in each. In this case, both CDT and EDT recommend taking two boxes. After all, at this point you have no uncertainty about how the world is—taking two boxes leaves you with an extra $1,000. FDTers recommend you take one box, because that timelessly leaves you with more utility.

Thus, FDTers generally leave transparent Newcomb’s problem richer than either EDTers or CDTers. FDT proponents claim that FDT “gets you more utility,” and is thus the right criterion of action. I have four problems with this argument.

First, I don’t think FDT says anything in any case because it’s not a complete theory (see section 2). If FDT says nothing, it can’t get you the most utility.

Second, FDT doesn’t always get you the most utility. For example, consider the following exotic possible world: the actual world. In this one, if you hang around academic philosophers, they will think you’re silly if you adopt FDT. This will make you sad. So adopting FDT gets you less utility. Additionally, in the actual world, I would get less utility if I were an FDTer, because I find it fun to argue with FDTers about decision-theory. Or imagine that the government passed a law where they tortured everyone who thought FDT was the right view. FDTers wouldn’t be better off.

Or imagine the following setup. You’re offered a box full of cash. A predictor predicts if you’d one-box or two-box in Newcomb’s problem. If you one-box, they put nothing in. If you two-box, they put a million dollars in. Now, suddenly, it’s the two-boxers who are rich.

These examples may seem unfair. You directly get rewarded based on things that are downstream of your decision theory. But Newcomb’s problem is also unfair in precisely this way. It ties how much money you get in a box to your judgments in a decision problem.

Now, you can get around this by narrowing the claim. You can say something like “FDT gets you most utility with respect to the utility that’s downstream of your decision algorithm.” But similarly, CDTers can claim “CDT gets you the most utility causally,” and EDTs can claim “EDT gets you the most utility evidentially.” The different theories disagree about what kind of utility is decision-theoretically relevant. So just pounding the table and saying “my theory is best by the lights of the criterion that my theory says is decision-theoretically relevant,” is obviously question-begging.

Third, FDT isn’t actually the theory that leaves you with most expected utility on average. In fact, in many cases, it’s EDT (perhaps updateless) that leaves you with the most expected utility. For example, in the smoker’s lesion case, EDTers tend to finish better-off than other people. In smoker’s lesion, smoking correlates with worse health, but it doesn’t cause it. But EDTers are less likely to smoke, so on average they’ll have better health.

Now, FDTers’ reply will presumably be that what matters isn’t just leaving with the most utility on average. Fair enough. But then they can’t appeal to this criterion. They don’t do best by it. Which kind of utility you get the most of can’t straightforwardly tell you which decision theory is right, because the decision theories disagree about which kind of utility matters.

Fourth, this argument begs the question in a different way. Other theories make a distinction between the disposition that are beneficial to have and the ones that are rational. For example, imagine that a highly reliable predictor checks to see if you’ll give into blackmail for $100. If so, then he blackmails you. If not, he doesn’t. In this case, non-FDT views grant that it’s timelessly better to not give into blackmail. They simply think that once you’re being blackmailed, the rational thing to do is to give in. At that time, you’re simply paying $100 to avoid having your life ruined.

Now, FDTers reject such a distinction. But we’ll need some argument against this distinction. Otherwise, this objection simply assumes that there’s no distinction between dispositions that are rational and ones that are beneficial. Non-FDTers have a perfectly sensible reply to this objection: in situations where you are directly rewarded for being irrational—for making some unwise decision—then of course the irrational people will be better off!

And non-FDTers have their own claim that their theory gets you the most utility. In, say, MacAskill’s bomb case, FDTers blow themselves up while CDTers and EDTers don’t. CDTers and EDTers thus leave with more utility when they’re in this situation.

Non-FDTers can grant: something FDTish might describe the kinds of dispositions you timelessly want to have, depending on how the world is. But that’s different from it being the right account of rationality. The dispositions that are beneficial aren’t necessarily the ones that are rational. Decision theories are theories of rationality, not of how to program an AI. If you are only interested in the question of how to program an AI, don’t purport to be giving a decision theory that is superior to the ones philosophers endorse.

6 Conclusion

FDT is both implausible and underbaked. It sometimes licenses setting yourself on fire for no benefit. It depends on analyzing how other algorithms would be different in the logically impossible world where your algorithm was different, but has no account of how to analyze logically impossible worlds, how to analyze what it means for your algorithm to be different, and how to analyze the impact that your algorithm being different has on other algorithms. This isn’t a minor technicality—it means that there is literally no situation where we can derive the correct answer from the theory.

Permit me to go slightly meta for a moment. Ideas like FDT are not unknown to academic philosophers. Various ideas in the vicinity have been proposed. Indeed, a view like FDT—where you one-box in Newcomb’s problem even if the boxes are transparent—is intuitive to a lot of people. But the view is pretty widely rejected because it doesn’t really hold up when you scrutinize it and filling in the details is very difficult. There’s a line by Scott Alexander that I sometimes think of:

My heuristic is that when the mainstream consensus refuses to engage with a critique and hem and haw about it being “problematic”, they are usually wrong. But when they explicitly declare “This is incorrect” and write papers explaining their reasoning, they are usually right.

The response from academic philosophers has been more in the direction of “write papers explaining their reasoning.” FDTers who think their theory is unfairly neglected by the experts need some explanation of why the academic philosophers who hear of FDT nearly always think it’s wrong.

Among laypeople who hear about decision theory, lots of them adopt something FDTish. So you need some explanation of why it is that nearly all the decision-theory experts—who write monstrously complicated papers with math that would go over your head—think FDT is wrong, but intro philosophy students who know almost nothing about decision theory think that it’s right. In general, you should be skeptical of views that are rejected by ~100% of relevant experts, even after considering them at length.

There are two answers that people often give: one and two. Two is the right answer. We’ve known this for decades. There are decisive arguments for it. Parfit settled the dispute in 1984.

And yet people are still making this mistake. For example, here’s Leif “befuddled by stipulative definitions” Wenar:

But let’s picture that person you’ve supposedly rescued from death in MacAskill’s account—say it’s a young Malawian boy. Do you really deserve all the credit for “saving his life”? Didn’t the people who first developed the bed nets also “make a difference” in preventing his malaria? More importantly, what about his mother? She chose to trust the aid workers and use the net they handed to her, which not all parents do.

Wenar goes on to chide MacAskill for his supposed philosophical incompetence, before making a basic error that has been known about by philosophers since before I was born. Not a great look. Specifically, the view Wenar endorses is called the “share of the total view.”1 On this view, you should analyze the impacts of your actions by dividing the total impact by the number of people who acted to produce that impact. If ten people collectively save ten lives, each only gets credit for saving one life. There might be more nuanced ways of doing this, but the basic idea is that you’re supposed to divide up credit for the impact across those who acted to produce the impact.

Contrast that with the correct view called the counterfactual view. On this view, when deciding between two actions, you should look at what happens if you take each one. When seeing how good it is for you to take an action, you should look at which morally important things in the world will be different if you take the action vs if you don’t. In the earlier case, each action is counterfactually responsible for two lives saved. Each person gets credit for saving two lives.

Here’s an easy way to see that the share of the total view is wrong. Imagine that there are two actions. You can only take one:

1. You get someone to tell his friend to pull three children out of ponds. His friend does it. Three lives are saved.

2. You buy a rope (from a vending machine, not a person) which you then use to pull two children out of ponds.

Which action is better? The obvious answer is the first one. Yet using Wenar’s logic, you should take the second one. After all, if you do the second, you get full credit for saving two lives. With the first, you only get credit for saving one third of three lives—so only one life total.

You’ll note an interesting result of the counterfactual view: the total credit awarded by multiple people taking an action can be greater than the benefit of the action. Three people each get full credit for saving three lives, even though only three total lives were saved, not nine. That is because they all took an action that was counterfactually responsible for saving three lives.

Or, to put it another way, they each took an action that if they hadn’t taken it would have led to three additional deaths. This sounds a bit weird at first, but it makes sense when you think about it. There’s no reason that the counterfactual credit coming from a sequence of acts has to equal the total benefit coming from the sequence of acts.

If multiple people are each counterfactually responsible for some effect being brought about, then total counterfactual impacts from each of their acts exceeds the good brought about. If there’s a three-judge panel in a debate round, and two of the judges vote for me, there are three people each counterfactually responsible for my victory: me and both of the judges who voted for me.

Applying this to MacAskill’s case, if you make a charitable donation, you are counterfactually responsible for saving a whole life. So are various other people. That is because all of you took an action which, had it not been taken, would have left one extra person dead.

Counterexamples to the share of the total view abound. Here’s a famous one from Parfit:

The Second Rescue Mission. As before, the lives of a hundred people are in danger. These people can be saved if I and three other people join in a rescue mission. We four are the only people who could join this mission. If any of us fails to join, all of the hundred people will die. If I fail to join, I could go elsewhere and save, single-handedly, fifty other lives.

If you divide the credit by the number of people acting, then it’s better not to join. But clearly, this is wrong. So the share of the total view must be wrong too!

By the way, the things I’m saying here are standard. I don’t know of any papers defending the share of the total view after Parfit argued against it. This is a nice example to illustrate that we do sometimes learn things in philosophy. We don’t just rehash the same debates over and over again. Sometimes, Parfit decisively shows what’s what.

Here’s an objection you might have. Suppose that there are four people. They each have an option like in the second rescue mission. So they can all go out and join the rescue mission, saving 100 lives, or they can save fifty lives on their own. If they each get credit for saving 100 lives by joining the rescue mission, then wouldn’t it be better for them to join it? And if they do that, then they save only 100 lives, rather than 200 lives. So doesn’t the counterfactual view get the wrong answer?

No!

You have to actually consider the counterfactuals. When you do this carefully, every single apparent counterexample dissolves. If any don’t join the rescue mission, then presumably the others wouldn’t either. So then the first person by not joining the rescue mission is counterfactually responsible for saving 200 lives, rather than 100 if they join. So not joining counterfactually saves an extra 100 lives.

That’s assuming that the other people would have joined the rescue mission if the first person hadn’t. What if that assumption is wrong? What if the other people wouldn’t have joined even if the first person had? Well then, even if they join, the mission still fails. So then 150 lives get saved rather than 200. Still, the view recommends not joining.

The only way the view recommends joining the rescue mission is if everyone else will join whatever you do. That way, you joining leads to 200 people being saved instead of 100. But that verdict is correct! So when you think clearly about what the counterfactual is, any apparent counterexample dissolves.

When thought about this way, the view is pretty obviously correct. When deciding between two actions, you should consider which action it would be better for you to take. You should just look at which morally relevant things will happen if you take each action, rather than arbitrarily splitting credit. You simply analyze the actual effects of your taking one action over another.

The share of the total view has other bizarre implications. It implies that credit is shared with other people, but not with non-people. So imagine two actions:

1. You pay $100 to a person who delivers medicine that saves someone’s life.

2. You pay $100 to a robot who delivers medicine that saves someone’s life.

The share of the total view treats these very differently. It implies that the second one is twice as good, because credit isn’t shared with another person. But shouldn’t these be treated the same? When taking an action that benefits someone, why the hell does it matter if the other entity in the loop is a person vs a robot? This just seems obviously irrelevant. And there are weird intermediate cases—what if it was an octopus or a baby? Any line drawn seems ridiculously arbitrary and not like the kind of thing that could feature in the fundamental rules of normativity.

So the share of the total view is simply wrong. The counterfactual view is correct. This is settled. Criticizing EAs for philosophical errors is a perfectly respectable endeavor—I’ve been known to do it myself. But you shouldn’t do it while making basic errors. And if you’re a philosophy professor who claims that “no competent philosopher” could have written a sentence providing such exotic oddities as stipulative definitions—even claiming “their flesh would have melted off and the bones dissolved before their fingers hit the keyboard”—you’re without excuse. Don’t conjoin snark and incompetence. Or, in the words of Leif Wenar:

The crucial-but-absent Socratic meta-question is, ‘Do I know enough about what I’m talking about to make recommendations that will be high stakes for other people’s lives?’

It began with a back-of-the-envelope calculation, or a BOTEC, as the cool kids say.

I had flown to the Bay Area to attend the Manifest prediction markets conference. I gave around seven talks, trying to let people know everything important that I thought was true as quickly as I could. Only like half of what I said was about invertebrates! While I was there, I was accosted by a very persuasive fellow making the case that flying across the country to canvass for Alex Bores was the best use of my time and money. The most infuriating thing about this—he was right! Damn it!

Some nerds crunched the numbers. Their guess was that every three hours spent canvassing gets you about one extra vote. That estimate seems pretty plausible to me. Historically, about 2% of open seat primaries have come down to 100 votes or less. This election was unusually close. It seemed like my odds of flipping the election could be as high as half a percent.

I didn’t want to canvass for Bores. Doing so meant that by the time I’d return to Oxford, my friends would have mostly left. My Oxford friends mostly attend the university, so now that I’ve returned, they’re mostly gone (RIP , Oxford isn’t the same without you). I did not get to say goodbye to some of my favorite people before I left. I also was not thrilled about the prospect of standing out on street corners for eight hours talking to strangers. But sometimes the right thing to do isn’t the thing you most want to do.

I won’t bore you with the details about why this seemed worth it—why a Bores victory mattered so much. You’ve heard me lay out this case in other articles. But the extremely condensed version is that Bores is good on AI safety, having passed a comprehensive AI safety bill called The RAISE Act. The AI companies spent millions of dollars trying to make an example out of Bores, so that the most pro-AI-safety candidate would be crushed and future politicians would be deterred. The race was also looking like it would be close, so it seemed important to help drive votes for Bores.

relays a speech that Paul Shapiro often gives at animal events:

How far would you go to achieve animal liberation?

[The crowd cheers.]

Would you do whatever it takes?

Yes! [comes the reply.]

Would you break the law?

[They respond even louder] yes!

Would you go to prison for animals?

Yes!! [It’s fewer voices, but more emphatic.]

Banging his fist on the table: Would you give your life for animals if that’s what it took to win?

Yes! [They scream out in religious fervor!]

[Grabbing the podium like a Baptist preacher, voice thundering out across the crowd]: Alright then. Would you get a hair cut, put on a suit, and go schedule meetings with politicians and corporate boards of directors?

[Quiet murmurs spread across the room as the activists exchange confused looks with one another.]

I imagine a similar speech being given to a room full of AI safety supporters:

How far would you go to achieve AI safety?

[Crowd cheers].

Would you write on LessWrong?

Yes!

Would you spend 14 hour days typing in Google Docs exhaustively responding to every conceivable objection?

Yes!

Would you do technical research?

Yes!

Would you present quick non-substantive, non-obviously-correct talking points to mostly uninformed likely voters that don’t primarily involve discussing what you think the biggest issues are?

The answer, from a number of people, was “yes.” And so it was that a bunch of AI safety supporters turned up on street corners to talk to voters about Alex Bores. When the stakes are high, you put down the Google Doc, and you fight.

2 There’s so many times I’ve let you down

Winston Churchill allegedly said “the best argument against democracy is a five-minute conversation with the average voter.” Like most of the things Churchill supposedly said, he didn’t actually say it.

I had a lot of five-minute conversations with a lot of average voters.

The best case for democracy doesn’t depend on voters being informed. Of course voters aren’t informed. I’ve read the things written by the libertarian economists about how breathtakingly ignorant voters tend to be—where, for example, they consistently haven’t the foggiest clue how government works or who is in power. The better argument for democracy is that ignorant as they are, voters generally vote against you when you make their life worse—so democracy gives politicians incentives not to make voters’ lives worse.

Now, this isn’t a perfect system. It gives little weight to the interests of non-voters, so sometimes politicians do things like light random African pharmaceutical manufacturers on fire to look tough, try to pass laws locking pigs in crates for months on end, and temporarily eliminate industrial civilization across an entire country. But dictatorship tends to go even worse, so we must settle.

I had a lot of conversations with voters about how they were voting. It confirmed everything the libertarian economists suspected. Multiple people told me that they weren’t voting for Bores because they’d gotten too many flyers in the mail. One woman told me she was voting for Lasher because he was Jewish.

In many cases, people explained that they supported one candidate based on a policy position shared by the other candidate. Often voters showed up at the polls without any idea who the candidates were, having not yet made up their mind between them. In virtually all cases, when voters raised an objection to a candidate, they were wholly unaware of the standard responses—having done at most fifteen minutes of googling. Very rarely, a voter had given the candidates serious thought, and was persuadable with reason.

The best parts of canvassing were the rare occasions when one could have a substantive conversation with a voter. It was also fun to try to persuade people, even non-substantively. The worst was standing around for many hours, trying to look smiley and engage random passersby in conversation about how Alex Bores sponsored the RAISE act, the most comprehensive piece of AI safety legislation in the country and is fighting for you!

There’s a serious debate that rages in society about who the most oppressed groups are. Gamers? The global poor? I think I have an answer: it is canvassers!1

When you are canvassing, many of the people hate you. They would like you to stop bothering them and go die in a hole. They just want to walk past and not hear about the environment, or the election, or the kids in Africa, or whatever the hell is on the flyer you have. And yet we still canvass, because every once in a while someone who had no idea that there was an election going on wants to hear more about it.

When you’re canvassing, people will often tell you their broad theory of politics and the world. You just have to nod along as people explain their unified theory of the world, containing gems like “it’s all about money.” Whoa, deep.

Sometimes, when you are canvassing, the rude things people say to you are pretty funny. I had one lady hand me a copy of an election flyer that had gone up on her car. The flyer said mean things about Micah Lasher. She explained that she was “returning” it and was a personal friend of Micah. As you may be surprised to learn, I am not responsible for the content of every flier put on anyone’s car anywhere in New York. The flyer did seem pretty scummy though, and Micah seems like a decent guy.

Someone else explained that he could tell that I was a communist just by looking at me. He boldly predicted an incoming communist revolution and proclaimed “when the communists line up and kill their own, I hope they kill you first.” Okay.

The funniest thing that happened to me was that I chatted with three kids from the inner city with saggy pants and the smell of marijuana about them. I talked the one who was in the district into voting for Bores. There was a canvasser for Micah who was also standing on the same street corner. They (the youths) asked me if they should “run him off the block,” comparing him to someone “selling drugs on your corner.” I’m pretty sure they were joking? I clarified that they should definitely not do this! Confirmation of what has long been suspected: I am extremely hip and in tune with the youths.

I met Eddie Gibbs who served in the state assembly alongside Bores. He and I were canvassing next to each other, me for Bores, him for himself. Gibbs was amazingly charismatic and much better at striking up conversations with voters than I was. When he saw a voter who had a bunch of flyers, he called out “they should give you a shopping cart for all those flyers.” Gibbs had endorsed Bores.

Gibbs also seemed to know a huge chunk of the passersby. I encouraged him to tell the people he knew to vote for Alex. He sort of did sometimes. He also apparently called Alex to tell him that I was doing a good job after I flipped two would-be Micah voters, which was very generous. Sadly, Alex didn’t pick up (inexcusably in my view—it’s not like he was busy or anything).

As the days passed, I began to like canvassing more and more. I got better at it with each passing day, and things tend to be more fun as you get better at them. I enjoyed convincing voters and outclassing the Micah Lasher canvassers on the same street corners. It still wasn’t fun, but maybe ended up about as enjoyable as a standard job.

3 Every place I go, I’ll think of you

To the people I canvassed with, I love you all. You are amazing.

Maybe this isn’t much of a surprise. People who drop everything they’re doing and start spending hours every day doing something they mostly don’t like because they’re convinced that it matters morally are likely to be unusually thoughtful and virtuous. But even with high expectations, I was amazed by the other people.

Some of the best people I’ve ever met I met during this last week. The people who came out to canvass were funny and smart and deeply decent. Working on a shared goal has a way of forging a deep and lasting bond. At the nightly hangouts, it was like a meetup of close and long-time friends. Most importantly, a lot of them were Bentham’s Bulldog readers, so that tells you a thing or two about the bottomless depths of their virtue.

They were also extremely smart. When I explained the extremely tangled suitcase argument against deontology to 4ish people, they all got it ridiculously quickly. They didn’t even think that it was an argument for naive consequentialism as a decision-procedure, unlike some people (cough, cough).

The other canvassers had extreme proficiency at getting things done. They’d plan events in full without needing outside input. Everything was amazingly well-organized given how complex logistics were. I was especially impressed by this because I am not agentic and am a bit of a disorganized fiasco—it was cool to be surrounded by an elite team of organized nerd geniuses.

I met people who told me how deeply they hated canvassing. That didn’t stop them from getting up at 7 a.m. and canvassing until 5 p.m., every day for a week. Despite being untrained unpaid volunteers, we worked really hard. We weren’t just clocking in for a 9-5 job. We cared about every single vote. And we tried hard to get them.

We weren’t in the handing out flyers business. We were in the persuasion business. We’d spend time looking at how other people canvassed, trying to do it better. When we handed out flyers—or “literature” as it is called, as if we were handing out copies of Moby Dick and Hemmingway rather than fliers with lines like “Alex Bores will fight for you”—it was generally after a conversation, rather than in place of one.

I loved hanging out with the other canvassers. Together we laughed, talked about how to canvass better, and talked about philosophy. We watched results on election day together. They tolerated my occasional (read: hourly) digs at functional decision-theory. I don’t know of a better group.

To the people I canvassed alongside who are reading this, if you ever need a favor, let me know. I consider all of you my friends, even if we didn’t talk very much. You’re all amazing.

4 But the dawn is breakin’ it’s early morn

On the day of the election, we all woke up at 4 a.m., got to our stations at 5 a.m., and canvassed until the polls closed at 9p.m. The classic 5-9 job! This meant I canvassed for about 16 hours, with only brief breaks to get lunch and dinner and go to the bathroom. So did most of the other people canvassing.

On election day, the people were unusually chatty. For many of them, had it been another day, they might have looked the other way. On other days, only about one or two people would tell me that I convinced them to vote for Bores. But on election day, seven different people told me that I’d gotten them to vote for Bores when they weren’t otherwise planning on voting. Two told me I’d convinced them to vote for Bores instead of Lasher. So at least 11 votes for Bores.

Sixteen hours, it turns out, is a long time. It is hard to pretend to be enthusiastic for sixteen hours. Six things helped:

1. The belief that what I was doing really mattered (more on that later).

2. A crap ton of caffeine.

3. Melatonin the night before, allowing me to get a non-trivial amount of sleep.

4. A firm desire to outcompete the various Lasher canvassers on adjacent corners.

5. The knowledge that the other Bores canvassers were doing the same and counting on me to do the same.

6. That morning, I’d listened to the song Ob-La-Di, Ob-La-Da by obscure music band The Beatles (I hear they’re popular in the impoverished, technologically primitive “nation” called the United Kingdom—which doesn’t even have air conditioning and is poorer than all 50 U.S. states). Hard to be in a bad mood listening to that song! In previous days, I had listened to more solemn music which left me more solemn, like It Was Love That Laid Us Low. The music you listen to has a big effect on your mood! Turns out it was “It Was Love That Laid Us Low” that laid us low.

By the end of the day, the world was a strange whirl of sights and sounds. I was in a haze, where all that was real was the random passersby and the line “Alex Bores sponsored the RAISE Act and was the most effective incoming legislator from New York City, which is why big tech companies spent $10 million…”

Finally—finally—9 o’clock rolled around and the polls closed, and we all went home to watch election results. At that point, the prediction markets had us at about 25% odds. We mostly thought the odds were a bit better than the prediction markets were saying, because we were hopeful that the canvassing might have raised the odds of a Bores victory by a few percent.

Immediately, everything went wrong.

The first thing that alerted me that something was wrong was the prediction markets. We were dropping precipitously. 15%, 10%, then 5%. I hadn’t even looked at the voting results, but I knew the prediction markets must be reacting to something. I #trustthescience (the prediction markets). It turned out that early results were in and they were not looking good for Bores.

The election party was fairly solemn. On the one hand, we mostly kind of suspected that this was going to happen. Alex had been an underdog throughout the race. On the other, it was hard to watch our guy getting crushed. In the end, he lost by about 4%. I’ll never forget when his odds in the prediction markets dipped below .5%, so that an Alex Bores victory became less likely than the truth of functional decision theory. RIP! (In all seriousness, this is an exaggeration. FDT doesn’t have enough content to be capable of truth or falsity).

Soon after, it was called for Lasher. We put up a decent fight and Lasher was always the likelier winner, but still, it was a bit painful to lose by 4%. There was, however, something a bit relieving about it—I kept imagining us losing by just a few votes and beating myself up about my errors being counterfactually responsible for a Bores loss. Perhaps this is immoral, but I was a bit relieved that it wasn’t close enough that we could have made any difference.

5 And I tell you now, they don’t mean a thing

In the end, we lost. Was it all a waste?

Not really. The election was reasonably close given that Lasher had all the big-name endorsements and years more of experience. You haven’t seen the last of Alex Bores, mark my words. We managed to make things close enough that the AI companies will think twice before trying to buy another election.

Alex put up enough of a fight that Lasher had to make his own AI policy. It’s pretty good. The fortunate reality is that Lasher is a really good candidate as well. In almost any other election, he would have been my pick. In his victory speech, he declared:

I have some news for the two big AI companies who’ve taken such an unusual interest in who won this congressional seat: I won’t be taking my cues from either of you when it comes to protecting our kids, our jobs, our environment…

Hopefully he won’t take cues from them either when it comes to averting existential risks.

It’s a good thing that New York 12 went with him over, say, sardine Jack, whose credentials are being a Kennedy, being handsome, and having no Super PAC backing (Super PACs only back candidates that win—they didn’t back Vermin Supreme either!) Jack referred to himself as “no PAC Jack,” and when he behaved aggressively while canvassing next to one of my friends, someone joked that he should be called “no tact Jack.”

The race was close enough that we got a flurry of positive media coverage. This emphasized the backfiring of the AI industry’s ploy. It made it clear that if they try to buy another election and crush another safety-conscious candidate, it will be a protracted slog, rather than an easy victory.

More importantly, you always have to act based on the information you have at the time. The Bores bet was valuable in expectation even if it didn’t work out. We were spending our time in the hopes that we’d have a small probability of making a difference. That’s what a lot of existential threat reduction looks like. Most of the time, you don’t make any difference, but it’s worth it because you just might make a huge difference.

So the campaign wasn’t a waste. We learned a lot. I made amazing friends who I will stay in touch with. I got to hear AI safety nerds try to give rousing speeches, with occasional profanity tossed in. Most of all, we fought against the forces of darkness and we gave it our all (that’s usually a cheesy line, but not when you canvassed for 16 hours). We helped show the big tech firms that they cannot silently crush whichever politicians support AI safety. The battle was lost, but the war may yet be won. Sometimes, it is rational to do what you’ll predictably regret, so I have no regrets.

To see what is in front of one's nose needs a constant struggle.

—George Orwell.

Many good arguments are simple and straightforward. For example, the argument for veganism depends on the obvious premise that you shouldn’t hurt others a lot to benefit yourself a little—and yet, given that nearly all animals are horribly tortured in factory farms, this is what you do when you consume animal products in the standard way. Oftentimes, however, people have a way of shutting their brain off, so that they are unpersuadable even with very straightforward arguments. I thought I’d discuss some of these ways.

Obviously I won’t provide an exhaustive list. There are many different ways not to appreciate straightforward arguments (e.g. you might stop paying attention to the argument and start thinking about what you’re going to have for lunch). What the methods of brain anesthetization that I’ll discuss have in common is primarily that they allow, in response to a straightforward argument, one to begin thinking of other things instead of the argument. Often these other things are related, but are different in crucial respects from the original argument. It is hard to be convinced by an argument if, rather than thinking about the argument, you instead think about other things.

1 Assuming a broader ideology

Here is one common way that people shut off their brains. They hear some premise in an argument. That premise reminds them of some other broader ideology that they reject. Then they start talking about the broader ideology instead of the premise.

is not a utilitarian. You can tell because he writes posts with titles like Why I Am Not a Utilitarian. However, Huemer sometimes makes arguments that contain some words that utilitarians use sometimes. When arguing against meat consumption, he appeals to the premise that it’s wrong to cause others a lot of pain and suffering for the sake of comparatively minor benefits to oneself.

This claim, you’ll note, is not utilitarianism. Utilitarianism says that one ought always and everywhere to take the action that maximizes total utility. That is different from the claim that you shouldn’t hurt others a lot for small personal benefit. And yet for some reason, Huemer’s critics love to accuse him of assuming utilitarianism.

Walter Block, when responding to Huemer, goes on at length about why utilitarianism is false, even though that is not a premise needed in Huemer’s arguments—and it’s not even one he accepts. Aeon Skoble accuses Huemer of presenting “a utilitarian argument of the pre-Millian, Benthamite sort: pains and pleasures do not differ in any qualitative way but are substantively the same and only need to be added up.” As Huemer put it:

Here is an analogy. Suppose you see a libertarian philosopher burning a baby with a cigarette lighter. You tell the libertarian that he should stop doing that, because he is hurting the baby. Now imagine this libertarian replies: “Aha, you appealed to the badness of pain; therefore, you are assuming utilitarianism! Furthermore, since the baby’s pain is purely sensory, you must be assuming a pre-Millian utilitarianism, wherein all pleasure and pain is of the same kind. But I reject utilitarianism; therefore, I am free to go on burning the baby!” My point here is not that animals are exactly like babies. My point is that you don’t have to be a utilitarian to be against inflicting great suffering for trivial reasons.

Here is what I suspect happens: when debating ideas we disagree with, we often start out trying to prove them wrong. This leaves us to latch on to whichever idea we have a rebuttal to that is in the nearest vicinity of the thing that our interlocutor is saying. So if our interlocutor has a premise that vaguely reminds us of utilitarianism, even if it does not assume utilitarianism, this might induce us to stop thinking about his argument and start thinking about utilitarianism.

But this is an error. To see if an argument is right, you have to think about the argument. You can’t just think about some other idea that repeats some of the same words as the argument. The claims “you shouldn’t throw baseballs,” and “you shouldn’t throw babies,” contain mostly the same words—even many of the letters overlap—but aren’t equally plausible.

Alternatively, people often treat simple and straightforward arguments as proxies in some broader ideological battle. If you appeal to a moral premise for a conclusion that people reject, they will often accuse you of assuming moral realism. But you don’t have to be a moral realist to make moral arguments. Nobody thinks that, say, arguing about abortion requires that one be a moral realist. It’s only when people make moral arguments for positions outside the societal Overton window that they’re accused of making contentious meta-ethical claims.

Or alternatively, if you make an argument about how your theory renders some fact more likely, you’ll be accused of assuming Bayesian ideology. This is bad reasoning. While Bayesian language might be the most precise way to operationalize standard claims of differential improbability, other theories are perfectly happy admitting the evidential force of facts that are likelier given one theory than another. It would not be reasonable for a young earth creationist to object to the claim that old earth better predicts convergent fossil dating by rejecting Bayesianism!

2 Thinking only about what your theory says

Suppose you are given an objection to your theory. One way to shut your brain off and not fully absorb it is to think only about what your theory says, rather than about whether the objection is substantively plausible.

Let’s say you are an average utilitarian, meaning you think the best action to take is the one that maximizes how well off people are on average. You learn that your theory implies that it would be good to proliferate miserable people in hell provided that they were marginally less miserable than the existing people. This seems like quite a forceful objection. It should cause you to abandon your view.

But here is one thing that you might think: “well, if I’m an average utilitarian, what I care about is average utility. Proliferating miserable people in hell raises the average, so by the lights of my theory, it’s good, rather than bad.” Instead of thinking about what is plausible, you start thinking about what your theory says.

Notice that this is a fully general response to any argument against any position. For something to be an argument against a position, it has to be inconsistent with (or at least lower the odds of) the position. If you treat your theory as the ultimate arbiter of what is true, outside argument can never get you to stop believing your theory.

Meat eaters also often employ this strategy. They will start with a dogmatic insistence that meat-eating must be okay. Then they will come up with some totally ad hoc theory about why animals’ interests matter infinitesimally compared to human interests.

Then, each time you give them a counterexample, they will simply think about what their new theory says about it. For example, they might claim that the badness of your pain depends on how smart other members of your species are. Mentally disabled people and babies matter, but animals don’t, because babies and the mentally disabled belong to a smart species. Then, you point out that:

1. It doesn’t seem like how bad your pain is depends on facts about other people.

2. This view implies that if we learned that mentally disabled people and babies weren’t the same species as the rest of us, it wouldn’t be a big deal if they were tortured.

3. It’s not at all clear why this makes any difference to the badness of your pain. The badness of a headache seems to depend on how it feels, not all this other stuff.

In response, meat eaters often just dig in their heels and don’t really think through any of these counterexamples. Instead, they treat their theory as a self-evident truth and automatically reject any counterexample. Functional decision theorists do this a lot too. When you show that their theory licenses exploding yourself for no benefit, they just keep thinking about why their theory gives the recommendation—noting that it timelessly gets you more expected utility—without thinking if the judgment is substantively plausible.

Along these lines, people often just reiterate their theory in response to objections. When presented with an objection, they stop listening and start thinking about new ways to explain their view—as if you were too stupid to get it before. This is very common among non-philosophically-trained physicalists.

In response to the Mary’s room argument, they mostly just keep repeating their position without responding to the objection. They don’t explain which premise they reject, but instead act like you haven’t gotten their position and seek out new ways to explain it. People who think torture is worse than any number of dust specks are also fond of this technique—in response to the spectrum argument, they often just keep repeating their view that badness scales non-linearly with intensity of pain. You’ll notice that this is not a response to any premise in the argument.

3 Asking for definitions

Arguments are typically composed of words. Most words are a bit hard to precisely define. So one way to turn your brain off is to focus on the minutiae of the words used rather than the argument, even when the finer points of the definition don’t really matter for the core argument.

For example, Huemer’s argument for vegetarianism appeals to the premise: you shouldn’t cause others a lot of pain and suffering for small benefit to yourself. But what is suffering? This is an interesting question, but it’s irrelevant to the argument. Whatever suffering is, being castrated, locked in a cage, and forced to live in feces for months on end without being able to turn around definitely qualifies. When you castrate the pigs, they do not act like they’re having a nice time or having a satisfying workout. So if you start airily musing about how to define suffering, you are missing the point.

Similarly, you should not start quibbling over whether pain can be good if it’s enjoyed—e.g. the pain from working out is often thought to be enjoyable. Maybe this is right, but it doesn’t apply to castrating pigs without anesthetic. It’s irrelevant to the actual argument at hand.

You can ask for definitions if you’re genuinely unclear on how words that matter to the argument are being used. But you shouldn’t ask for definitions if the finer points of the definitions don’t matter to the argument. We use words all the time without having a precise definition of them. It’s fine to say things like “I know that the earth is round,” even if you haven’t nailed down the exact definition of knowledge.

4 Thinking about the effect of the argument being true

Suppose there’s some argument A. If A is true, it implies B. Imagine that B is logically downstream of A, so that B depends on A but A doesn’t depend on B. One way to shut off your brain is to think about B without thinking about A, feel skeptical about B, and then use that to reject A without thinking about A again.

For example, suppose that you are trying to decide whether insect welfare matters. You might think the following: “if insect welfare matters, then all sorts of ordinary human behavior might be really bad.” Intuitively, you feel skeptical of the claim that all sorts of ordinary human behavior might be really bad. So therefore you reject the claim that insect welfare matters.

Here, you have stopped thinking about the essential claim that if insects can hurt, their hurting matters for the same basic reason humans hurting matters. Instead, you have begun thinking about a surprising effect that would have. You feel skeptical of that surprising effect, so therefore you reject the claim that insect hurting is bad.

But this gets things the wrong way around. Your judgment about whether ordinary actions are bad depends on your judgment about the significance of insect pain. You shouldn’t start with the dogmatic assurance that ordinary action is fine, and then use that to tell you that insect pain doesn’t matter. Instead you should start by thinking about if insect pain matters and use that to decide if ordinary action is fine.

If your judgments about A should be derivative on your judgments about B, then you can’t use judgments about B to inform A. You don’t have direct access to whether ordinary actions are fine. Rather, your judgment about whether ordinary actions are fine should depend on whether they routinely lead to something bad.

If you could start with moral judgments about acts, and then use those to inform downstream judgments about acts, then you’d be able to deduce how the world is from moral claims. Suppose someone tells you that when you purchase certain products, you contribute to slave labor. Using this methodology, one could start with the assurance that buying the products is okay—and then from this learn that the products don’t use slave labor. Obviously this is wrong. You can’t get an is from an ought.

It is especially bad if you start thinking about how accepting some judgment would go against widespread societal attitudes. For most of history, the view that slavery is wrong ran counter to what most people thought. If you reject plausible moral premises when they imply that society is behaving badly, you will never be able to morally improve on the accepted practices of your society—and for most of history, you’d have tolerated conquering, genocide, and slavery.

5 Conclusion

Humans reason in a very motivated way. For this reason, if an argument is for a conclusion that we reject, our brain contrives clever little ways to avoid seeing its force. I’ve discussed some of these ways here. There are surely others.

How do you identify if you are shutting your brain off? I’m not sure if there’s any perfect universal method. But here is one thing that should tip you off. If you notice that, upon hearing some argument, you have stopped thinking about the argument itself and instead begun thinking about other related ideas, that is an alarming sign. The kind of shutting off your brain that I describe here just consists in not thinking about the argument in front of you. Good arguments do not come with tanks and artillery—they cannot grab you by the lapel. They cannot move you if you do not think about them.

is one of my favorite philosophers. His book Dialogues on Ethical Vegetarianism is one of the best books I’ve ever read, and it’s a big part of how I talk about veganism. The book manages to be both hilarious and devastating—decisively refuting all the common arguments against veganism. But there’s one big point in it that I disagree with.

In the book, Huemer argues that insects aren’t sentient. He gives three arguments for this position that came mostly from a paper by Eisemann in 1984. I know the book has been influential at convincing people that insects aren’t sentient, so I thought it would be worth explaining why I don’t buy these arguments. Here’s Huemer’s first argument:

One, they don’t have nociceptors…The kind of nerve cells that sense pain.

This is false. Nociceptors have been found in many different species of insects: flies, moths, bees, and various others. We’ve even found nociceptors in small worms with just a few hundred neurons, so it seems unlikely that insects are too primitive to have nociceptors. The claim Huemer made looked reasonable in 1984, but it turns out that we’ve learned some new things since the Reagan administration.

Second, they have drastically simpler central nervous systems. Like a hundred thousand times simpler.

It’s true that insects have smaller and simpler nervous systems. But small insect nervous systems are enough to sustain the capacity for seeing, problem-solving, memory, and smell. In light of this, it doesn’t seem that unlikely that they could sustain the capacity for pain.

We know that insects’ nervous systems facilitate sophisticated tradeoffs between pain and reward, painkiller response, and conditioned learning. Insect physiological response increases as external damage increases. The most interesting study on insect pain genetically modified insects to have capsaicin receptors which let them taste spicy food. When their food was laced with capsaicin, they stopped eating it—taking rare occasional nibbles and writhing around in agony, with their reaction scaling proportionally with the amount of capsaicin. But when they were given a painkiller, this aversive behavior stopped.

So there’s no dispute about whether insects display a range of behavioral responses characteristic of pain. But insofar as they do, it doesn’t seem that unlikely that they feel pain. If their nervous systems can lead to something that closely resembles pain, why are we so confident that it can’t lead to pain?

An insect with a crushed leg keeps applying the same force to that leg. Insects will keep eating, mating, or whatever they’re doing, even when badly injured – even while another creature is eating them.

This is the most memorable claim in the famous Eisemann 1984 paper. But it’s seriously oversimplified.

1. Insects are very different from humans in a lot of ways. For them, mechanical damage is generally less dangerous given their exoskeletons. They’re also small enough that there isn’t much benefit to getting insects to put less weight on an inured body part. So it isn’t that surprising that they sometimes walk normally even when injured—evolution gives them reason to care a lot less about bodily injury. In response to excessive heat, which is a lot more threatening to creatures with exoskeletons, they behave a lot more like vertebrates in pain.

2. Insects have smaller and simpler brains. It’s harder for them to focus on multiple things. So an insect’s mental life might be a lot more monolithic than a human’s. Thus, an insect mating might simply lack mental room to focus on anything else.

3. Being injured doesn’t always produce aversive behavior. Sometimes soldiers keep fighting after injury without acting like they were hurt. Boxers, when they’re hit in the face, keep fighting normally. Often mammals don’t feel pain in response to damaging stimuli—e.g. naked mole rats don’t react aversively to excessive CO2. So if there isn’t much evolutionary reason for insects to prioritize responses to particular kinds of damaging stimuli, we shouldn’t take their non-response to be particularly evidentially significant. And insects often change their behavior a lot in response to injury: they often stop mating, avoid putting weight on an injured body part, and groom injured areas.

4. Insects heal unusually quickly. Often studies injure insects and then wait to see how it affects their behavior. But by then, their injury might have healed. In addition, insects might experience pain more like fish—fish mostly feel sharp, shooting pain rather than dull aching pain. If so, it makes sense that they wouldn’t feel lasting pain in response to injury.

In short, the evidence against insect pain is that in some cases, simple-brained insects’ brains drown out responses to mating when injured, and they sometimes put weight on injured body parts. But I’m not that shocked, even conditional on insects feeling pain, that insects with their tiny little brains go crazy when mating and don’t notice anything else. Similarly, given that there isn’t much selection against insects putting weight on an injured leg, I’m not that surprised that they sometimes do this.

To be clear, I don’t think it’s any kind of guarantee that insects are conscious. But it’s reasonably plausible. The arguments against insect pain are far from decisive. The sensible position to have at this point is deep agnosticism. Given that there are about a hundred million of them per person, and the arguments for ignoring their interests are bad, insect welfare is a pretty big deal.

M: When we were last here, you convinced me of veganism. Now, I have been vegan for (checks date when book was published) about ten years. Are you going to try to convince me of of something new this time?

V:

Specifically, I am going to try to convince you of Longtermism.

M: Oh, what’s that?

V: The basic idea of Longtermism is that making the far future go well is super important. There’s a strong form and a weak form. The weak form says: the world should be doing a lot more to make the far future go well. The strong form says: making the world go better is by far the most important global priority, so that how our actions affect the far future is much more important than how they affect present people. I accept both.

M: And why do you hold this view?

V: The argument for it is very straightforward. The future could have a lot of people. Some estimates put the number of future people there could be at 10^58, and others think it’s a lot more. If we survive for billions of years and fill the surrounding galaxies, the number of people could be truly enormous. If this is right, then in expectation, almost all of those our actions affect reside in the far future.

If future people matter, and they’re most of who we affect, then what matters most is how we affect future people. As an analogy, if there was a giant undersea civilization quintillions of times more numerous than humanity, and our actions affected them in great numbers, even threatening to extinguish them entirely, then our impacts on them would seem to be the most important impacts of our actions.

M: But can we really affect the far future? I mean, sure, maybe the far future matters a lot, but imagine people ancient Egypt trying to benefit us. They wouldn’t have been able to. Probably their guesses for how to benefit us would have been weird stuff like “pray to Ra for the continuation of future humanity.” As more and more time passes, it becomes harder and harder to predict the impacts of your actions. So over long time scales, our ability to predict decays to almost nothing.

V: I think that’s basically right for lots of effects of our actions. In many cases, you can’t really predict how an action will affect the future. But there are some cases where you can. These are the most important impacts of our actions.

A persistent state is one that, if entered, will stick around. Extinction is a persistent state. If we go extinct, we’ll stay extinct. So by influencing the probability of extinction, we can affect the far future predictably. Other persistent states include: lock in (where we cement our values in stone for ever), extreme power concentration, and institutional development of, for example, space.

In fact, lots of experts think that AI might produce lots of rapid economic growth, and give us the ability to preserve current institutions forever. So affecting the trajectory of AI—ensuring the AIs shaping the future have good values and care about everyone’s welfare—could have enormous and lasting effects. If other actions are comparable, it’s because they benefit far future people.

M: None of that sounds crazy, I guess, but is that obviously more effective than other charitable opportunities.

V: Yes! If a trillion dollars would reduce existential threats by one in a billion, and you think the earlier 10^58 number if off by 10 orders of magnitude, then the expected number of life years added by a single dollar is 10^27. That’s a ridiculous amount—far more than the total number of people on Earth. It would be extremely surprising if there weren’t any spending opportunities that could reduce ex risks by more than one in a billion for a trillion dollars!

M: Sure, but actions could backfire as well. Maybe you’ll try to regulate AI and actually end up making things worse. History is replete with examples of people trying to do good and it backfiring. So perhaps we just don’t have enough information to reliably improve the future.

V: Sure, it’s always possible our actions will backfire. But note: that doesn’t just apply to actions to benefit the far future. It applies equally to actions to benefit present people. So if you think that some actions done for the sake of present people are good in expectation, then the same should be true of future people.

Let’s be a bit more concrete. Suppose you get a career working to stop deadly pandemics. Might that backfire? Sure, I guess. But while it might have bad effects you didn’t think of, it might also have good effects that you didn’t think of. Having good effects you didn’t think of seems, if anything, more likely. So then I don’t think this should undermine your confidence in this being a good career choice.

I don’t think you can really be completely in the dark about whether any action benefits the far future. If you were, then you’d think that working to cause a pandemic was just as likely to benefit the future as not to. That doesn’t seem right. So while we should remain cautious and think seriously about backfire risks, they aren’t reasons for total inaction. And if you think causing deadly pandemics is bad for the future in expectation, you should also think preventing deadly pandemics is good for the future in expectation.

M: That might be right. But doesn’t this assume utilitarianism? What if I don’t care about maximizing total welfare?

V: No, it doesn’t assume utilitarianism. It just assumes that the interests of the overwhelming majority of those affected by your actions matter.

M: And doesn’t it assume moral realism? Which I reject completely—people say it’s intuitive, but it’s certainly not intuitive to me. And whether it’s intuitive to most people is an empirical question.

V: No, why would it assume that? It’s just a standard moral claim. You don’t have to think morality is objective to make moral claims sometimes.

M: And doesn’t it assume disjunctivism about perception?

V: Um, no, why would you think that?

M: And doesn’t it assume that the French renaissance was highly overrated as a period of historical inquiry.

V: No.

M: Okay, okay, but I think I might have something it does assume. Doesn’t it assume that creating happy people is valuable? I reject that view. I was reading Jan Narveson, and he said “make people happy, not happy people.” That really resonated with me. The slogan just sounded so true.

V: And why would it assume that?

M: Well, you kept talking about how many future people there might be—about how extinction might decrease the number of future people. But if I adopt the person-affecting view, then I don’t care about increasing the number of happy people.

V: Remember what Longtermism says: making the future better is really important. It doesn’t, on its own, say anything about which things make the future better. So even if you think filling the future with happy people doesn’t matter, you can still be a Longtermist. Then you should think that preventing the future from going badly and preventing future suffering is important.

M: And how might we do that?

V: A number of ways. You might try to fight for moral progress, because moral errors might lead to lots of suffering. You could try to increase the representation of digital minds or work to give AIs better values. Alternatively, you can try to prevent extreme power concentration.

On top of this, the person-affecting view is very implausible—for more on this, see a really nice blog post by the dashingly handsome blogger Bentham’s Bulldog, rivaled in handsomeness only by his cleverness and most of all, his humility.

M: Okay but what about this. As John Broome has shown, the person-affecting view holds that creating people is greedy. This means it can soak up value in either direction. So if you create a happy person and make someone slightly worse off, that neither makes the world better nor worse. Similarly, if you create a happy person and make someone slightly better off, that makes the world neither better nor worse.

So by changing the distribution of future people, won’t this soak up all the value of Longtermist interventions?

V: Well, all of our actions change the distribution of future people. So really this is a fully general objection against any actions being for the best.

M: But isn’t this a very strange moral argument to make? It’s very counterintuitive that the far future matters this much. Just as we can toss out an ethical theory if it says slavery is okay, can’t we toss out one for saying that all of our actions’ impacts on future people are a rounding error? On this view, if the world could voluntarily take a 90% reduction in present welfare, all to improve the welfare of people alive more than a million years in the future, we should. That’s very counterintuitive.

V: I agree it’s a bit surprising. But sometimes moral claims are surprising because the world is surprising, not because the ethical claim is surprising.

Remember when we talked about meat-eating. You said it was counterintuitive that factory farming was the worst thing in the world. But then we agreed that wasn’t really counterintuitive. When we broke down the core argument, it was:

1. If something is the cause of more pain and suffering in a few years than all the suffering in human history, then it’s one of the worst things in the world.

2. Factory farming is the cause of more pain and suffering in a few years than all the suffering in human history.

3. So factory farming is one of the worst things in the world.

It’s not premise 1 that’s counterintuitive but premise 2. If a surprising moral claim arises because the world is surprising, that’s no reason to reject the ethical theory underpinning it. When the world is surprising, our judgments about it should be as well. Let’s do the same for the case for strong Longtermism. The core argument is:

1. For every present person our actions affect, there are in expectation trillions of future people that our actions affect.

2. If an action affects trillions of times more expected future people than present people, then the main thing that determines whether the action should be performed is how it affects future people.

3. Therefore, the main thing that determines whether actions should be performed is how they affect future people.

The surprisingness resides in premise 1, not premise 2. So this is no reason to reject strong Longtermism. When the world is weird, the ethical claims that apply to it will be weird too.

M: You’ve been talking about how many future people there will be. But the estimates you’ve been given have been assuming that there could be a lot of digital people. I reject that! I think consciousness is substrate dependent—no life, no mind, no person.

V: First of all, the case for strong Longtermism doesn’t depend at all on whether there can be digital people. The future can be really big even if there can’t be lots of happy digital minds. Second, you should have some serious uncertainty about whether there can be digital minds. Many of the best philosophers in the world think there can be. Even on standard views according to which consciousness is substrate dependent, there can still be digital minds if they’re configured correctly. But even if you think there’s only a 1% chance that there might be digital people, there are still enormous numbers of them in expectation.

M: But doesn’t this all assume Fanaticism? Your odds of making any major difference to the far future—preventing extinction, say—are very low. So then isn’t this a Pascal’s mugging? It’s a tiny chance of a big impact.

V: No, being a non-fanatic isn’t any help, and this isn’t a Pascal’s mugging. What alternative to fanaticism do you have in mind?

M: I think you should discount low risks. If someone comes to you and offers you infinite utility, and you’re 99.99999999999999999999% they’re lying, then you should value their offer at nothing even if its expected value is high.

V: Do you mean you simply ignore every event with low probability?

M: Yes.

V: That’s very implausible

M: Why?

V: Suppose that someone has a quintillion ways to torture you. They’ll pick a random one. Should you say “well, the odds of every particular one being picked are one in a quintillion, so this is no big deal—every specific scenario gets counted for nothing?”

M: Maybe the threshold is below one in a quintillion.

V: Whatever the threshold is, just change the number to that. Also, obviously the odds of Longtermist interventions making the far future a lot better are way more than one in a quintillion.

M: Hmm, fair point, I had never thought of that and have never thought about philosophy of risk at all. So then I guess the better view is the one adopted by the academics called “tail discounting,” rather than naive Nicolausian discounting that’s fallen significantly out of favor. Tail discounting involves discounting the best and worst slices of probability space. So when you’re tail discounting, you ignore the best and worst .000000001% of scenarios and only focus on the middling ones.

V: But Longtermist interventions nudge things outside of the best .000000001% of scenarios. So tail discounting is no help.

M: Hmm, good point, I’ll have to think more on that one. In light of this, are there any other considerations that undermine in serious ways the risk discounting objection to Longtermism?

V: Indeed there are. But they don’t lend themselves great to dialogue format, so I’ll just point you in the direction of section six of this article by dashingly handsome blogger Bentham’s Bulldog—whose good looks are matched only by his unwillingness to put self-praise in the mouth of dialogue characters.

M: Okay, now that all of my objections have been refuted, I am convinced! I’m now a strong Longtermist. But what should I do about it?

V: Well, a lot of impactful careers are advertised on the 80,000 hours job board! They also have more general career advising—and see here for some high-impact Longtermist projects. So go there and try to get a high-impact career that makes the world better.

And thus it was that M became a polyamorous vegan alignment researcher in a bay area polycule.

Every day, hundreds of millions of animals are slaughtered after undergoing a lifetime of intense suffering. Even just counting land animals, that’s about 1.4 billion animals slaughtered every week. In just one year, we kill more animals than people have lived throughout the entirety of human history.

The animals are treated horribly. They’re selectively bred so they can barely move. Their bones are broken repeatedly. They live in filth and feces, choking on ammonia all day and getting burned by the acidic feces of those around them. Billions of baby male chicks are ground up alive. The lives of the animals are worse than those of any group of humans on Earth, unless those humans live in torture chambers. You’d be arrested if you treated a dog this way.

Words cannot do justice to the horror. More suffering is caused in a few years by factory farms than all the suffering in human history. This is much worse than anything we do to other humans—billions of innocent, defenseless creatures tortured for their whole lives just because we enjoy the taste of their flesh.

But it’s still not the worst thing.

No, the worst thing would be if factory farms stuck around and spread. If their hellish, cancerous rot spread throughout the galaxy—so that their cries would ring out on a quadrillion worlds, for a trillion years. The worst thing would be if we did this entirely unnecessarily, for no benefit. At that point, even an all merciful God could not resist obliterating creation.

At that point, we wouldn’t just have built hell, but filled the universe with it—bringing screams to a thousand silent worlds. We would have brought those dark satanic mills through the void of space, so that no reachable world was free from vulnerable creatures crying out in pain. At that point, there would be some deep sense in which the fight for good was lost, where evil ultimately triumphed, on a cosmic scale.

If factory farming occurred at the scale it does on Earth across 1,000 worlds, then around 80 trillion land animals would be tortured annually. It would take half a day to slaughter a population equal to the number of people who have ever lived. And if meat consumption rises with abundance, the real numbers might be far more.

But 1,000 worlds is a dramatic underestimate of the expected number of future worlds where we’ll factory farm animals. There is some probability that we’ll spread humans to and terraform a non-trivial portion of the reachable universe. Toby Ord estimates that we could reach about 10^21-23 stars. If these stick around for a trillion years and we assume that there are 1,000 planets per star (given the possibility of constructed planets), we end up with 8x10^46 factory farmed animals.

We may be entering that world.

If factory farming isn’t replaced, then by default, it will spread to the stars. We won’t automatically stop farming cows, pigs, and chickens just because we venture away from Earth. So the question is whether we’ll replace factory farming.

Now, the good news: there’s a perfectly good alternative. It’s called lab grown meat. It involves growing meat without needing to torture and slaughter an animal. It has the taste and consistency of meat—it will be identical at the cellular level to standard meat. The only difference is that it did not come off the flesh of a tortured and murdered animal who thrashed in agony and terror as it was bolt-gunned in the skull, who didn’t want to die.

You’d think support for it would be a no-brainer. Humans enjoy the taste and consistency of meat. With lab-grown meat, we’ll be able to produce it without the torture. This has all the upside of ordinary meat and none of the downside. So what has the world done about it? Have we invested billions of dollar trying to produce it? Has there been a great research project taken by popular governments?

No. In fact, it’s been banned in a number of U.S. states. The Save Our Bacon Act, currently being battled over in Congress, prohibits state bans on meat products, but leaves out any prohibition on lab meat bans. Lab grown meat became a culture war issue. Some people find it icky. They don’t want to eat it, even if it’s cheaper. They only eat flesh that came off a living sentient being.

So already, we’re doing nowhere near enough to accelerate lab grown meat. But still, it seems reasonably likely that we’ll develop it at some point. Maybe 80% odds, I’d guess, that we find some way to produce meat without needing to torture billions of animals.

But after that, we have to actually eat lab meat instead of the flesh of mistreated animals, in order for it to replace ordinary meat. And that’s where the deeper problem lies. AI might solve the technical problems of lab meat, but there’s no guarantee people would eat it. While most people in the U.S. and UK say that they would try cultivated meat, only about 13% said they’d prefer to eat it over regular meat.

Now, who knows how this would really go? People’s trust in a technology tends to increase after the technology is created. Before computers, people generally didn’t see much use for them—now, almost everyone has one. Surveys provide only imperfect evidence. But still, if people have such profound contempt and fear for lab meat that in a number of states it’s illegal to make a cultivated hamburger, it’s hard to be extremely confident that factory farming will end. If there’s a 70% chance that cultivated meat would displace factory farming if it was developed, and an 80% chance that it’s developed, then the odds that it will replace factory farming are only 56%.

It could, of course, end in other ways. But I find it hard to be very confident that factory farming will end through any other mechanism. My guess is that it’s about a coinflip whether factory farming spreads to the stars—whether hell goes intergalactic. Maybe 60% odds that factory farms are eventually abolished. , one of the most experienced and careful animal activists, thinks that the end of factory farming is neither impossible nor inevitable.

In What We Owe The Future, argues that the end of slavery wasn’t historically inevitable as we often now think. The end of slavery depended, in large part, on persuasive moral appeals from activists. The end of slavery didn’t depend primarily on economic factors—when it was abolished, the slave industry was highly profitable and was growing rather than contracting. In fact, the British government spent about 40% of their annual treasury expenditure paying off slave owners. And after abolishing slavery internally, Britain worked hard to get it abolished in other European countries, which makes no sense if they had primarily economic motivations.

Moral shifts don’t just happen. Sometimes they happen by default, but we shouldn’t be so confident that they will happen by default that this gives us license to stop fighting. Even if the odds that factory farms would spread intergalactically were only 10%, this would still be a moral emergency of cosmic proportions.

And those fighting for farmed animals have been amazingly successful. Each dollar spent on corporate campaigns has spared animals from many years in a cage. About 50 full-time activists have managed to shift the practices of a multi-billion dollar industry, affecting hundreds of millions of animals every year. This comes out to millions of animals affected per full-time activist pushing for these reforms.

How does knowing that factory farms might persist change the priorities we should have?

Some Longtermists think that fighting against factory farming isn’t a big deal from a Longtermist perspective. But this becomes hard to maintain once you realize that the factory farms might stick around for trillions of years and spread through space—that the next few decades might be the most important time in history for farmed animals, that will determine their eternal fate.

It becomes especially hard to maintain when one thinks about how a world where most biological organisms rot in hellish torture chambers is likely to treat other sentient beings. If we can’t even treat biological beings with any modicum of compassion, what hope is there for digital ones?

It is terrifying that the end of factory farms may not come until humanity is extinguished entirely. It makes the fight all the more urgent. The hell that exists on Earth today may soon spread to the stars.

https://partiful.com/e/dwazevmFszwJwKPhqtc0?c=u5bQcrxT

Effective altruists should be a lot more excited about politics.

People say lots of things about voting that don’t make any sense. For instance, you’ll often hear that we should vote because our forefathers fought for our right to vote, and it’s a very important right. But the right to offensive public speech is also important—this doesn’t mean we should speak offensively in public. That it’s important to have a right doesn’t mean it’s important to exercise it. That it’s important to have a third amendment doesn’t mean you should deliberately go out of your way not to let troops into your house.

There’s a better argument for voting: elections are really important. Government is powerful, so even making it a bit better is valuable in expectation. But in response, people argue: your vote won’t make a difference. Millions of people are voting. What are the odds that yours will be decisive?

Now, it’s certainly true that individual votes don’t flip most elections. But flipping an election is a huge thing—it changes who governs millions of people. A career spent on anti-proliferation probably won’t prevent even one country from getting a bomb, but it’s an impactful career in expectation because stopping a country from getting a bomb is so important.

The best estimates I saw before the last presidential election were that your odds of flipping the election if you were in a swing state were around one in two million. If we assume the better candidate would spend federal money 0.1% more effectively, and ignore all other effects of having a better candidate, then a vote for the better candidate is worth 13,000 dollars. Libertarians are fond of suggesting that voting is a waste of time, and while the cynical libertarian takes are often correct, here they’re just wrong. Voting is a great use of time.

For some slightly fancy technical math reasons, if there’s a competitive race with n voters, your odds of flipping it are about 1/n. So if a race is ten times larger, then the odds of flipping the election are ten times less. This is an in-principle reason for the size of the election not to generally impact how valuable voting is. Your vote gets less likely to change the outcomes of the election as there are more voters, but this is proportionately canceled out by the election being more important.

So what determines if your vote is valuable? Which elections is it best to vote in? There are a few big conditions.

First, if an election is close, your vote is more impactful. If you’re in a safe state in a federal election, then probably your vote doesn’t matter. Maybe it has a small effect for setting a precedent, but it’s not a huge deal. In contrast, if an election is really close, your vote makes a bigger difference.

Second, if there’s low turnout, your vote makes a big difference. Simple intuition: imagine that there were only five people voting in the presidential election. Your vote would then be huge! You’d be able to influence who assumes an important office with high likelihood. Similarly, the fewer people are voting for an important position, the more impactful your vote becomes. It means you can have a greater proportional influence on a more important candidate.

Third, if an election is important, all else equal, that will make your vote more impactful. This should be obvious.

Anything that makes it so that the election is more important or that you’re likelier to influence the election increases the power of your vote. Your vote starts out pretty powerful. About 2% of open-seat primaries come down to 100 votes or less.

2 NY-12

To see these in action, let’s apply them to the New York 12 race that I’ve been banging on about recently. Probably about 85,000 people will vote in the election. So your odds of flipping the election are around 1/85,000, by conservative estimates (probably less given how close this election is, but here we’re being conservative).

The election matters: Bores is the single most important candidate on AI safety anywhere ever and the election has important precedent-setting effects for future candidates’ stances on AI. Peter Wildeford’s more detailed calculation is that a vote for Bores has a 1/20,000 chance of flipping the election. And you should believe it, because he’s some kind of weird supergenius and also one of the best forecasters in the world.

Let’s assume very conservatively that by being serious about AI, Bores reduces existential risks by 1/100,000. Is this conservative? Bores is poised to be the best representative in the country on AI, and this election is a broader referendum on whether big AI companies can crush those who support AI safety. Here’s how Peter Wildeford put it:

• When they started spending, Alex Bores was third in the polls. The goal was to make him lose overwhelmingly, thus scaring away anyone from regulating AI.

• The good news is that this seems to be backfiring. Alex is now basically tied in the polls.

• But if Alex loses, the OpenAI SuperPAC will declare victory anyways and no one will really remember the details. This would create bad momentum for AI safety.

• But if Alex wins, industry will have been very visibly defeated and AI safety will have very visibly won. This will create huge momentum for AI safety!

Combine this with a few other important facts:

1. Bores is just 35. He may have an influential political career in the future. So when voting for Bores, you raise the odds of him getting elected to higher offices than the house.

2. AI will become an increasingly important issue. This will both increase the value of AI legislation and make it easier to pass AI legislation.

Combining these numbers, a vote for Bores has about a 1/2 billion chance of stopping human extinction (1/20,000 chance of flipping x 1/100,000 of stopping extinction=1/2 billion). Thus, it saves about four lives in expectation (there are 8 billion people in the world).. That’s incredible! It’s an amazingly good use of a few minutes! If your friend had a bomb in his basement with a one in 2 billion chance of blowing up Earth, and you could deactivate it in a few minutes, that would be a great use of a few minutes. The scale of impact is equivalent to the average marginal impact had if a quarter of Earth took action that collectively stopped human extinction.

For another point of comparison, assume that Bores winning is equal, in value, to 50 million dollars being spent on AI safety. This seems conservative—having extra people in government is more important than throwing more money into an already money-flooded zone. This would mean that a vote for Bores is equivalent to donating $2,500 to AI safety. Given that it costs around $700 to add a vote, donations to Bores’s campaign are probably about 3x better than other AI-safety donations (though they decline in value a lot after today, so donate soon). If you think having Bores in Congress is more like $100 million in AI safety, it might be more like 6-7x more effective.

Canvassing probably gets about 3 votes a day. So the impact of canvassing is conservatively equivalent to making and donating more than $7,500 in a day. If ten people canvass for Bores for ten days, that’s equivalent to giving $750,000 to AI safety. Alternatively, if a vote for Bores is equivalent to four lives saved, it’s akin to saving twelve lives a day. Phone-banking is also probably very impactful.

Thus, even going by very conservative estimates, voting for Bores, telling your friends to do the same, canvassing, and donating is a very good use of time. As Tom Smith put it:

In the past year, as I’ve done more research and (more recently) volunteered on the ground to help Alex Bores’s campaign in NY-12, I’ve developed a gears-level understanding of how electoral politics in the US works.

I now believe that working on US electoral politics is one of the highest impact areas from the general AIS perspective. I feel like I was a fool for not thinking it through sooner.

And note: here I’ve just talked about AI-related existential risks. If you add in the fact that Bores is good on animal welfare, stopping authoritarianism, and just generally passing good bills, the impact is even greater.

Now, maybe you’re not sure if I’m right. Maybe you read something written by Jason Brennan a while ago, where he suggested voting was low expected value, and you don’t know who is right. Well, even if there’s only a 10% chance I’m right about voting, that still means its expected value is 10% what I suggested (assuming the rest of the calculation is right). Still, expected value is crazy high!

So if you’re in a primary where one candidate is noticeably better than the other, get out and vote. Turnout is often between 10% and 20% in congressional primaries. New York 12 is an especially important election, but there are other important ones too. Voting in high-leverage primaries, canvassing, and convincing others to do the same might be your most impactful use of time all year—so do it!

One of the most alarming trends in the nation is a shift towards Trumpian authoritarianism. Corruption lives in the open. Trump prosecutes his political opponents. He fires people for producing statistics he doesn’t like. Resisting Trumpian authoritarianism—and the slow corrosion of democracy more broadly—is one of the most important priorities.

This is a big reason I’m supporting Alex Bores in New York’s District 12 in the election happening on the 23rd (if you’re in Manhattan: make a plan to vote—if not, contact your friends and tell them to vote for Bores). I’ve written recently about why Bores is a great candidate. He’s sponsored one of the most comprehensive pieces of AI safety legislation in the country, stuck up for animals, and is good at getting things done.

But here I want to make a narrower case: Bores is the best candidate in the race for resisting Trump and authoritarianism, because he understands the technology that is about to make authoritarianism much easier to pull off.

Even people not voting for him tend to agree that Bores is the best candidate on AI regulation. Taking into account both his unique focus on AI and the precedent-setting effects of this election, I’d feel confident saying that Bores is by far the best candidate on AI. Congress is currently very inept when it comes to new technology.

During Mark Zuckerberg’s 2018 hearing, senators asked in befuddlement how Facebook makes money if it’s free. The people regulating what might be the most influential technology in history are mindblown by the abstract concept of advertisements. It’s hard to get sensible AI policy if politicians don’t know anything about AI, just like it would be hard to get sensible car regulations if politicians didn’t know what cars were, how they worked, or how they differed from trains.

AI safety people generally focus on risks around loss of control. But there are other serious threats from AI, including its use for facilitating authoritarianism and surveillance. Right now, the government and private companies have collected personal data on a massive scale, but they don’t have enough workers to analyze it in detail.

AI changes that. Building a database that scores every citizen for “presumed loyalty” would be impossibly ambitious for a team of humans; with enough compute it becomes feasible. China is already trying it. As the legal scholar Alan Rozenshtein has argued, the limits that have always constrained state power were practical rather than constitutional. Point a capable model at the data the government already holds, and the model will be able to find anyone who criticized the government. We need elected officials who understand this new threat in detail, particularly in light of the ubiquitous Trumpian assault on democracy.

On June 11, federal agents executed a search warrant at the Cleveland offices of the Ohio Organizing Collaborative, a group who primarily registers voters. They claimed to be carrying out a voter fraud investigation, though there are no public, specific allegations. They questioned staff for hours, then spread out across the state to employees’ homes, in some cases carrying subpoenas and seizing phones and laptops. The investigation appeared, according to those present, to concern alleged voter fraud—a claim the organization flatly denies.

The timing carried its own message: a voter-registration group, in a state with a closely contested Senate race, raided by Kash Patel’s F.B.I., five months before the midterms.

A raid like that is a threat to America’s free and fair elections. As the human frictions of running operations like this become automatable, the same impulse behind a single raid could be pointed at everyone at once. We need comprehensive AI regulation to prevent AI from being used for surveillance.

The whole fight between Anthropic and the Department of War was over two issues: Anthropic didn’t want their technology to be used for mass surveillance and they didn’t want it to be used for autonomous weapons. The DOW went around Anthropic by signing deals with OpenAI and Google DeepMind which, despite some token efforts, effectively allow the government to use AI for domestic surveillance.

An even more alarming prospect for AI-facilitated authoritarianism is that AI could concentrate power into a small number of hands. If current rates of progress in AI continue, or even if they slow considerably, superintelligence could be imminent—AI much smarter than the smartest humans could arrive soon. But if AI is powerful enough that leading experts fear it could take over the world, we should also be worried about people using AI to seize power.

AI could be used to carry out a coup. This could either be done by the head of an AI company, who gives their AI secret loyalties, or by a political leader. Remember: the AI companies are building AI with the express purpose of creating a technology with power unlike any other in human history. Shouldn’t such a thing be regulated? Shouldn’t the default assumption be that a technology directed by a small number of people, already interwoven with government, could be a tool for authoritarianism? Shouldn’t a technology like this be regulated by people who know something about it?

Bores is the major enemy of the big Trump megadonors and the tech companies who are crucial facilitators of digital authoritarianism. Greg Brockman, OpenAI’s president, gave $25 million to Trump’s super PAC last September and another $25 million to Leading the Future, a super PAC that launched with over $100 million—now more than $140 million—to fight AI regulation. Its backers include Marc Andreessen and Ben Horowitz, whose firm has committed over $50 million, and Joe Lonsdale, a co-founder of Palantir, the company now building ICE’s deportation-targeting software.

Their aim, aligned with a December executive order, is to wipe out the state AI laws passed in 2025 and block new ones. The big AI companies who want totally unregulated AI no matter what existential risks it poses are attacking the strongest AI legislator. Their admitted aim is to make an example of him. If he loses, every other lawmaker will think twice before crossing the industry.

A helpful heuristic: Trump’s biggest backers in tech are opposing Bores and spending millions of dollars to do it. If some of Trump’s richest supporters in tech regard Bores as the biggest threat, this is some evidence that Bores is the biggest threat to Trumpian authoritarianism.

Bores has already successfully stood up to Trump. When Trump proposed a slush fund to bail out January 6th insurrectionists, Bores proposed a bill taxing bailouts at 100%. This was copied by a number of different states. After facing this and other obstacles, Trump backed down. This illustrates a willingness to stand up to Trump when he does illegal things. We need candidates who will actively resist Trump’s illegal actions.

Bores is an extremely effective legislator. The people who worked alongside both Bores and Lasher have largely endorsed Bores. Of the fifteen who have endorsed one of the candidates, thirteen endorsed Bores. He’s gotten more bills passed than any other incoming legislator in the state of New York—see here for all his bills. Part of the reason the big AI companies are massively spending to defeat him is that they expect him to be effective at getting things done.

There are important things even a freshman Congressperson can do. Congress still has oversight authority, and, if Democrats take the House in the midterms, they will have subpoenas, hearings, and the power of the purse to use it. That means lawmakers can step in to punish bad behavior and write new rules while it still matters.

Many of the optimal rules are subtle and non-obvious: getting rid of loopholes that allow AI surveillance, preventing secret loyalties in AI, and requiring AI companies assess and mitigate catastrophic risks arising from their products. This requires legislators who understand the technology well enough to regulate it and are willing to act. Despite these being issues which most Americans would likely support, there are currently no Congresspeople well placed to pull this off. Surely the optimal number of Congresspeople who know and care about AI is non-zero.

Here’s the story in a nutshell: we have an extremely effective candidate who happens to be by far the best candidate on AI in the race. He’s so good on AI that the big AI companies are spending $10 million trying to defeat him. AI is an extremely powerful technology that leading experts fear could take over the world on its own; if it can potentially take over on its own, there’s also a serious risk that it could be used by someone to take over.

The government has already fought to make sure that this historically powerful, almost entirely unregulated technology can be used in mass surveillance. In light of this, doesn’t it seem we want a candidate who is good on AI regulation? Doesn’t it seem the technology that could facilitate an authoritarian coup and unleash a historically unprecedented boost in productivity should be regulated—so that we all may reap the rewards from it? Authoritarianism is not new, but AI might enable and cement it like never before.

It’s easy to think authoritarianism is some relic of a bygone era. It has been aided historically by the widespread thought that it could not happen here—that in a civilized and developed western country like our own, we are beyond its clutches. We must be ever vigilant about the ways authoritarianism might take root. While I would not claim that Trump has instituted authoritarianism, he has, at the very least, taken a number of actions that move us in the direction of authoritarianism (to take one notable example, he tried to illegally seize power after losing the 2020 election).

All of this is to say that having someone in Congress who is really good on AI seems quite important. Alex Bores is such a candidate. If you live in New York 12, I’d highly encourage you to vote for him and to canvass on his behalf. If you know people who live in New York 12, tell them to vote for him (see here for more detailed advice describing what to say and helpful links for which of your Facebook friends and LinkedIn connections live in New York). Donate—but only until the 18th; after the 18th, money raised by the campaign has almost no value, because you need to book ads five days in advance.

Bores is the best candidate for fighting Trump and AI-facilitated authoritarianism. Let’s help him win.

The worst thing in the world today, in my view, is wild animal suffering. For each human, there are around 10,000 reptiles, 100,000 fish, and 100 million arthropods. These beings can plausibly suffer intensely. For this reason, nearly all of the world’s intense suffering is experienced by wild animals.

Think back to the worst experience you’ve had in your life. Probably it is less bad than starving to death or being eaten alive. At any given moment, countless animals are experiencing that—in a few days the suffering of the animals in nature may surpass the suffering of all humans in history so far. Across history there have been countless human cries of agony and terror; in a week, in nature, there may be more.

Yet even that issue is surpassed in importance by something else, as thoroughly as the Milky Way surpasses a grain of sand. That issue is how things go in the far future. If things go well, the galaxy could sustain large numbers of conscious beings. Their lives could go extraordinarily well, and yet whether they do depends on what we do today. If the world goes extinct, or fails to count future interests, or becomes doomed to dystopia, then all that would be lost.

Any numbers in this domain are invariably speculative, but they are helpful for getting a sense of the scale. Bostrom estimated that in the far future there could be 10^58 future happy people. Another calculation implied that the real number could be twenty orders of magnitude more. Joy and love and knowledge and achievements saturating the galaxy, stretched out across epochs.

There are about 10^50 atoms on Earth. Imagine if every atom on Earth was a supremely happy person, able to experience things so great that you could not even imagine them. That is what might be on the table if the future goes well. It is important that we not blow it. As Bostrom put it:

In other words, assuming that the observable universe is void of extraterrestrial civilizations, then what hangs in the balance is at least 10,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 human lives (though the true number is probably larger). If we represent all the happiness experienced during one entire such life with a single teardrop of joy, then the happiness of these souls could fill and refill the Earth’s oceans every second, and keep doing so for a hundred billion billion millennia. It is really important that we make sure these truly are tears of joy.

But humans aren’t very good at thinking about big numbers. People would pay approximately the same amount to save 2,000 birds as 200,000 birds. When numbers get big enough, no matter how many zeroes there are, our brain simply registers it as [generic big number]. We grow insensitive to scope. A trillion beings in agony sounds like a quadrillion beings in agony, even though a trillion is a thousand times less. Taking away 99.9% of those suffering intensely matters a great deal, even if our brains cannot see that.

This is a staggering error. When a problem gets a thousand or a million times worse, our intuitions about it don’t change at all. Saving 99.999% of the victims of some problem has no effect on our intuitions about the scale of the problem. It’s no wonder that we neglect the 10,000,000,000,000,000,000,000,000,000,000,000 people there could be in the future (I just shaved off like 20 zeroes from the earlier number—you noticed, right?)

For this reason, it is helpful to try to think comparatively about big numbers. Instead of thinking of a trillion as a number, think of it as ~100 times more than all the people on Earth. When you imagine this vividly, it makes clearer what is at stake.

Start with wild animal suffering. There are around 100,000 fish for every human. There are 500,000 people in Wyoming. So imagine a Wyoming full of fish being weighed against five people. Which matters more? Maybe you are inclined to say the people matter more.

But think vividly about what the fish’s lives may be like: about them fleeing from predators trying to tear them apart, consistently undergoing experiences far more terrible than the worst thing that has ever happened to you, and being hungry every day of their lives. Is a Wyoming full of intense suffering really less important than five random people?

Imagine the fish’s brains in human bodies. These creatures would be, in pertinent cognitive respects, like enfeebled humans. They would matter as much as standard fish—for the sort of body a being possesses is clearly morally irrelevant. But these seriously disabled people would obviously not matter 100,000 times less than standard humans.

Similarly, there are 100 million insects for every person. Now, we are biased against insects because they are small and weird-looking. So imagine making them large and look like humans. These humanoid creatures would be plausibly capable of suffering, though very mentally primitive, like extremely cognitively disabled people. Would it really be plausible that a hundred million of these humanoids would matter less than you?

An average human has about 30,000 days in their life. So this would mean that a single day of your life would matter more than 3,000 of these creatures’ lives—creatures, remember, for whom it is plausible that they can suffer, perhaps even intensely. Does this seem plausible when you think about what it’s like to be them, rather than dismissing their interests out of hand based on arbitrary prejudice?

The badness of pain doesn’t depend intrinsically on your intelligence. A headache wouldn’t be any less bad if you were stupid. A drug that lowered your intelligence during surgery wouldn’t be anesthesia—if you felt everything, stupidity would be no comfort. So when countless beings are in intense agony for every person, that matters a great deal. If suffering matters because it hurts, then a month of wild animal suffering is a worse thing than all the suffering in human history.

This isn’t because of any unusual ranking of wild animal suffering. Even if suffering of animals was only .01% as bad as a comparable amount of human suffering, it would still be the worst thing in the world. It’s just because the numbers are so large. If you care about everyone who suffers, you should care a great deal about those who suffer in the largest numbers. If you think that a million beings experiencing intense pain for every second of human life is a big deal, then wild animal suffering is a big deal. It takes about a sixth of a second for the number of insect deaths to surpass the number of humans who have ever lived.

I’m not the one getting sniped by big numbers. It’s the people who ignore these problems who let the presence of large numbers anesthetize their brains to the horror of what is going on—so they ignore almost all the suffering in the world. Not caring about large numbers simply involves not caring much about each additional person suffering a wretched fate, so long as many others are suffering similar fates. It involves seeing individuals crying out in pain, and then ignoring it because you notice many others are in a similar situation. Caring about numbers is simply caring about each individual behind the numbers.

The situation is even clearer for future generations. Let’s be generous and say that 10^58 is optimistic by 8 orders of magnitude. That still means that there might be 10^50 future generations. That means that for every person alive today, there could be 10^40 people in the future.

Do people today matter 10^40 times more than people in the future? No! They don’t matter any more, because when you are in time doesn’t affect how much you matter. If you went back in time, you wouldn’t start mattering way more. But even if you think they matter somewhat more, surely they do not matter 10^40 times more!

There have been about 110 billion people so far in human history. That means the ratio between present people and future people utterly dwarfs the ratio between the very first human alive and all humans who have existed up until this point. The aggregate of all humans in history so far matters a good deal more than the very first human. For the same reason, the far future matters a lot more than the present. The most important thing—that which matters so much more, incomprehensibly more, than all else—is that the far future goes well. Making such a thing happen is in our hands.

We are blind to large numbers, because the human brain was not built for them. But when we try to decompose those large numbers into ratios in order to gauge relative importance, then we can see that there are things that are of gargantuan importance—whose significance dwarfs the problems in the world as a galaxy dwarfs a planet. If there are many beings of some class for every human, then if they matter even a tiny fraction as much as present humans, their aggregate significance surpasses our own.

An analogy I sometimes think of is a monarch who learns of the outside world. They were raised thinking that only the monarchy matters—that the outside world is not significant. Yet over time, they come to appreciate that people in the outside world matter as much as people in the monarchy.

Though their culture parochially fixates on what happens to their family, and only their family, in the outside world, there are a million times more things happening that are significant. There are a million times more non-monarchs than monarchs, and the non-monarchs matter as the monarchs do. Decency demands they not be blinded to most of what matters by their culture’s parochial fixation. It demands the same of us.

This is not to say that humans suffering today is a less big deal than we ordinarily think. It isn’t. It is to say that the far future and the suffering of the wild animals is a much bigger deal than we normally think. Human misery and death is exactly as bad as you think—and there are other things that matter trillions of times more.

This is a very tragic case. Two people’s lives were ruined that day. A young man was brutally stabbed to death, leaving his parents and loved ones to grieve. The sensible way to react to such a situation is with sadness. What you should not do—what you have lost a little bit of your soul if you do—is to treat an innocent teenager being brutally murdered as a convenient political football, while smearing him and his family. Could anyone stoop so low?

Yes.

, who has about 50,000 Substack readers, recently wrote a post called Dear Jeff Metcalf: Your Son Is Dead Because You Failed to Teach Him That Black Boys Have Boundaries. Jeff Metcalf is the father of Austin Metcalf. Patton is a college professor whose Substack bio describes her as, “Black, snarky, award-winning journalist, author, historian, child advocate, and college professor.” The piece is one of the most grotesque things I’ve ever read. Relevant bit:

Yesterday evening Jeff Metcalf, the father of Austin Metcalf, used his victim-impact statement to address Karmelo Anthony directly, after he was sentenced to 35 years after the court treated his act of survival as murder.

…

“You failed your parents, you failed yourself, and you failed society. You don’t belong in this community.” He also reportedly told him, “You’re going to prison,” and, “You can’t even look me in the eyes right now, but you can stab my fucking son in the heart.”

And right there, in his own words, Jeff Metcalf told on himself.

…

So let’s talk about Jeff Metcalf’s failure as a father.

Because when Trayvon Martin was killed, this country put a dead Black child on trial. They picked through his school records, his hoodie, his social media pics, his text messages, his teenage moods, his parents, his family structure, and every ordinary adolescent detail they could twist into evidence that he had somehow authored his own death.

…

So now, Jeff Metcalf, let us return this ugly ritual in kind.

…

You stood in that courtroom and told a Black teenager he failed his parents, himself, and society. But perhaps the harder truth is that you failed your son first.

…

YOU failed to teach your boy that Black children have boundaries. YOU failed teach humility, restraint, or the sacred fact that another person’s body is not your jurisdiction. YOU failed to teach him that another child’s space is not a challenge to be conquered. YOU failed to teach him that “community” does not mean white boys get to decide who belongs and who does not.

…

You failed, Jeff.

And I will not let you leave these truths out of the story.

…

You didn’t just say, “My son is dead and I am broken.” You said, “You do not belong among us.”

Among whom, exactly?

Whether you intended it or not, whether you want to admit it or not, “You don’t belong in this community” carries history. It carries sundown towns, schoolhouse doors, and white mobs outside courthouses. It carries the old civic theology of white space, where Black folks are tolerated only so long as they are quiet, grateful, submissive, and available for punishment. And I’m not about to sit quietly while a Black teenager is verbally exiled from a community that had already decided what he was before the facts could even breathe.

What a completely despicable sentiment.

Metcalf did not behave badly. Anthony was in his school’s tent. Metcalf asked Anthony to leave repeatedly. Anthony refused. So Metcalf tried to move him out, and was stabbed to death. The only one transgressing boundaries was Anthony, who entered the tent of a school that wasn’t his own and refused to leave when asked repeatedly.

But even if he had behaved badly, when a 17-year-old is murdered, you should not psychoanalyze them—trying somehow to tie their malfeasance to their father, and then blame a grieving father for the murder of his son. Evidence-free speculation on the misconduct of a grieving man whose son is murdered is evil. There’s little more despicable.

The reason Metcalf said that Anthony didn’t belong is not because he was black. It was because he murdered his son. If someone kills your son, harsh language is excused. In fact, it’s merited. Patton didn’t even bother discussing the facts of the case—in her mind, the fact that a white child was murdered by a black child is enough to settle who is at fault.

Imagine the races being reversed. Imagine that a white child entered a black child’s tent, was asked repeatedly to leave, and refused. Then, when the black child tried to remove him, the white child stabbed him to death. Would Patton think this was permissible conduct? Would she smear the black child’s father and blame him? Would she treat this as, rather than a tragic crime, a microcosm of radicalized grievance. Of course not. Excusing killing based on the race of the person carrying it out is the worst kind of racism.

Why did I write this article? To tell the truth, I don’t know if I should have. The sentiment is so grotesque that I feel a bit like I’m writing a reply to a statement put out by the Nazis. But here’s why I did it.

I’ve elsewhere complained about people on the right being bottomless pits of malevolence—gleefully cackling when children die from foreign aid cuts. It’s helpful to be reminded, every once and a while, that pure malevolence is not limited to one political party. Both sides have people willing to smear the families of the dead. There really are people who are cartoonishly evil, and some of them are ostensibly on the same political side as me.

But though Patton and I would likely vote for the same candidates, I don’t consider us to be on the same side in any deep sense. Patton’s cause is racialized grievance, and she’s willing to excuse any evil, however profound, in support of that cause. I want no part in that kind of ideology, and repudiate as fully as I would that of my political opponents. Evil is evil, even if it wears your side of the political spectrum as a skinsuit.

All the human suffering in the world is a rounding error compared to animal suffering.

Every two years, we torture and kill a population of farmed land animals more than the number of people who have ever lived. Counting farmed aquatic animals, it takes just a few months to exterminate a population more than the number of people who have ever lived. If we killed humans at the same rate we kill fish, the human race would go extinct in about a day. Animal farming routinely causes more suffering than all the suffering in human history so far.

And there is no end in sight. Once humans colonize the galaxy, factory farms might spread across countless planets. Factory farming may not end before we do—the human species may be eternally entwined with torment and carnage. Most people do not care very much about factory farming—they don’t regard it as a crucial moral issue and are unwilling to engage in fairly trivial sacrifices to mitigate the effect they have on it.

But even that is a rounding error compared to wild animal suffering. There are about 100,000 fish for every person, about 10,000 reptiles, and about 100 million insects. If humans died at the same rate as insects die naturally in the wild, we would go extinct in about a hundredth of a second. Despite it being plausible that insects can suffer intensely, the common societal view damns them on multiple fronts:

1. It holds that insects do not matter. While maybe shrimp and lobsters matter a bit, their land-based brethren, on account of being small and weird-looking, count for zero. Their experiencing intense suffering by the quintillions is not, on standard views, any cause for alarm. Even though insects are the plausibly conscious life form most affected, in expectation, by every decision ever made, the common view is that we should never pay any attention to the impact that our actions have on them.

2. It holds that wild animals don’t matter at all. Even though almost every conscious being is a wild animal, most people think we should never pay any attention to how our actions affect wild animals. So the wild insects are doubly discounted—both because they’re wild and because they’re insects.

3. It holds that we have strong moral reasons not to modify nature. So while most people don’t care at all about wild animal suffering, they think it would be very bad if you changed nature around in ways that reduced wild animal sufferings. If we had the technology to transform nature to be less cruel, most people would oppose using it.

Liberating animals is one of history’s great moral priorities. As I’ve argued repeatedly elsewhere, the badness of pain comes from how it feels. Animals can feel pain like us. Nearly all the world’s pain is experienced by animals. Making sure their lives are better and ending their massive suffering is a crucial moral priority. If you pressed a button which ended all human pain, this would have a negligible effect on the total amount of pain in the world.

Now, it’s true that, as I’ve argued elsewhere, digital minds might be even more numerous than biological ones. But this only makes the fight more urgent. A world where most biological sentient beings aren’t counted at all in decision-making is one where digital beings will likely be neglected too.

And in decades of animal activism, we’ve made some progress, but not overwhelming progress. We haven’t succeeded in banning the gigantic torture farms. In this post, I’ll lay out my best idea for how we win this fight. It’s an extremely tough one to win—for we battle both with giant corporations and human greed. But here is, at a high level, my view on what the animal movement should be doing to bring about the eventual liberation of all animals.

2 The vegans who hate animal welfare reforms

Some of the people who most hate animal welfare reforms aren’t factory farmers. They’re vegan activists.

They argue that welfarist reforms are merely incremental improvements that whitewash the cruelest abuses of the industry. If states ban battery cages, people can eat eggs guilt free, knowing that their eggs came from a chicken who wasn’t kept in a crate. The foundational problem, claim the abolitionists, is that we’re slaughtering animals for food, not how we do it. Focusing on how animals are murdered for food, rather than whether animals are murdered for food misses the forest for the trees, they claim. It would be like if abolitionists had worked to pass laws making sure you couldn’t whip slaves too brutally, rather than trying to ban slavery outright.

Proponents of this view often argue that welfare reforms counterintuitively make things worse for animals. The most successful welfare reforms in recent memory were cage-free reforms that made it so that animals can’t be caged. But critics like Gary Francione argue that this could make things worse for animals—instead of several chickens crammed into a tiny cage, huge numbers are crammed into a barn. This, they claim, is no better.

These people have things backwards.

Let’s start with the question of whether cage-free reforms really benefit animals. Fortunately, there’s been a fairly extensive body of research on this. The most detailed report on the effects of cage-free reforms came from the Welfare Footprint Institute. They analyzed how much time animals spend in different kinds of discomfort in caged vs. non-caged systems. Their conclusion?

Cage-free aviaries were found to be clearly superior to conventional cages and furnished cages even soon after a transition to cage-free environments. Overall, an average of at least 275 hours of disabling pain, 2,313 hours of hurtful pain and 4,645 hours of annoying pain are prevented for each hen kept in an aviary instead of [a conventional cage] during her laying life, and 1,410 hours of hurtful pain and 4,065 hours of annoying pain prevented for each hen kept in an aviary instead of a [furnished cage] during her laying life.

In other words, cage-free reforms are very substantial welfare improvements. All different kinds of pain are reduced. Suffering in cage-free systems is around half suffering of caged hens. Disabling pain—pain so intense that when you undergo it you can think of little else—is reduced by hundreds of hours per hen placed in a cage-free environment. This kind of research led my friend , one of the leading opponents of cage-free policies, to write an article lamenting his past opposition.

Now, is it true that cage-free reforms make the industry seem humane? The best thing I’ve read on this topic was by the Sentience Institute, which provides a number of strong objections to this concern about welfare reforms. Here are the considerations that strike me as most convincing:

1. When people vote for welfare reforms, they begin to see themselves as the sorts of people who care about animals. Getting a foot in the door with concern for the welfare of farm animals makes further change easier. Once people take some actions to promote animal welfare, caring about animals becomes part of their identity, and they’re likelier to do more. As writes:

Prop 12 was the California ballot initiative banning the sale of products from battery cages or gestation crates. After a year of intense campaigning and media coverage featuring endorsements from prominent stars and politicians, 7.5 million people voted in favor of the measure. I would guess that for at least 7 million of them, that was the first action they ever took in their lives consistent with the idea that farmed animals deserve even a shred of moral consideration. That’s a powerful step thanks to the consistency principle, the psychological pressure to act consistently with our past words and actions.

2. If welfare reforms lead to complacency, then we should expect them to be anti-correlated with one another. After one passes, it should be harder to pass other ones. Empirically, the opposite has been true. They’ve tended to snowball—once one happens, more happen.

3. Enshrining animal welfare in law normalizes it. A big part of neglect of animal welfare is cultural inertia—people don’t want to go vegan if the rest of their society isn’t vegan. The more that concern for animals becomes part of the culture, the easier it is to get further concern on the table.

4. Animal welfare reforms give rise to huge amounts of media coverage, focusing disproportionately on the evils of the industry, rather than suggesting that all is well. Being bombarded with constant TV ads about the industry that crams chickens into tiny crates where they can’t turn around doesn’t make people more sympathetic to the industry. As Aidan put it, “That means Prop 12 has without a doubt done more to increase the salience of farmed animal suffering than any of the abolitionist campaigns I’ve ever worked on. Sure enough, I’ve met more people who told me they were turned vegan by seeing a TV ad for the Prop 12 campaign than as a result of vegan street outreach.”

5. Empirically, learning about welfare reforms doesn’t seem to make people less opposed to animal farming.

6. The limited evidence that we have suggests that only a small portion of people become complacent in response to welfare reforms. The Sentience Institute article notes, “An analysis of social media responses to corporate welfare announcements found that only 3 of 1,617 comments explicitly suggested that the welfare reforms were sufficient and that no more action would be needed.”

7. Welfare reforms raise the costs of continued meat production. This makes alternatives to meat more economically viable.

8. These legislative reforms aren’t minor improvements with little overall effect on welfare. It’s much worse to be trapped in a tiny cage where you can’t move for your whole life than not to be. The main benefit of legislative reforms is that they lead directly to tens of millions of innocent creatures not suffering in horrible ways, every single year.

I wouldn’t just say “legislative reforms don’t hurt the cause of wholesale victory for animals,” just like I wouldn’t just say “avoiding lighting matches around oil-soaked rags doesn’t hurt the cause of fire prevention.” Corporate and legislative reforms are an integral part of how we win. Vegan advocates are fond of saying that we couldn’t eat animals in the numbers we do if we gave them high welfare. If that’s right, then to push for high animal welfare is to push for abolition of the industry.

3 How to end factory farming

Imagine if every year, in every state, we got a new welfare reform passed. First, we banned the hellish cages that trap animals for their whole lives. Next, we banned selective breeding for grotesquely large frankenchickens who can barely walk. Next we pass regulations ensuring that you can’t break the animals’ bones multiple times. Then we pass laws ensuring that they have adequate space.

After this, we push for more ambitious reforms. We ensure, under the law, that many of the things it would be illegal to do to a dog or a cat are also illegal to do to a cow, chicken, or pig. We require animals to have pasture access—so that they can go outside. Slowly, the price of animals begins to rise. Voters begin to think of themselves as the sorts of people who care about animals. They vote for a new pro-animal ballot initiative every year.

The price of animal farming rises as these reforms make it harder to provide the kind of animal torture that keeps animals extremely cheap. Lab meat becomes cheaper than normal meat, and also healthier.

I think having ambitious asks combined with serious community organizing is a key part of winning. It’s hard to get a mass popular movement going for cage-free reforms because, despite majorly reducing animal suffering, they’re not sufficiently inspiring. But we might be able to get a movement going demanding pasture access—demanding that the animals we eat can go outside. That’s hard to argue against, yet would deal a meaningful blow to cruel farming practices. Combining the mass popular efforts of extinction rebellion with specific policy asks is a recipe for success.

Now, here’s what the abolitionists tend to say in response to my proposals. Even if we abolished factory farms, that wouldn’t eliminate the problem. There’s an inherent moral issue with eating animals. If you give someone a great life, and then you slit his throat, you’re doing something evil. So even if we were able to get rid of factory farming, that wouldn’t be enough.

I have two responses.

First: are you kidding me? Sure, maybe ending factory farming wouldn’t be enough if we still slaughtered animals. But eliminating more suffering than all the suffering so far in human history would still be a very good thing. It is much worse to torture an animal and then kill it than just to kill it, no matter what you think about the ethics of killing happy animals.

Second, a world where we abolished factory farming would put us in a much, much better position to end the last remnants of animal farming. Once we think of ourselves as animal lovers who care deeply about the welfare of farm animals and have a viable replacement for it in the form of lab meat, we’ll be a lot more reluctant to continue killing and eating animals. The reason meat exists at a mass scale is factory farming.

So here’s step one in the fight for animal liberation: achieve as many welfare victories as we can. Ban the cruelest practices in factory farms. Do corporate reforms and protest aggressively when corporations don’t follow through on their corporate commitments. Fund cheap lab meat, so that there’s a viable alternative to factory farming.

I even think encouraging people to offset is an important part of this. To get a major social movement going for animals, we’ll need lots of people to participate. Unfortunately, the masses are not going to become vegan—doing so is personally costly and inconvenient. They should, but they won’t. So having a fairly low cost way that people can help animals without going vegan is a crucial step.

Offsetting also helps fund impactful projects that benefit animals. It both gives an opportunity to get the public on board and directly funds impactful projects. The best animal charities help multiple animals per dollar. If 1% of the U.S. population gave $25 per month, that would raise almost a billion dollars annually for animals. That’s several times more than all the current funding for farmed animals.

So activists should pursue a multi-part strategy. Get legislative reforms passed, create lab meat, and get the public on board with low-cost ways to help animals.

Is this the best plan in the world? Certainly not. But it’s a hell of a lot better than the plan “try to get the world to be vegan.” We’ve tried that one. We have nothing to show for it, but stagnant rates of veganism and the general public feeling reluctant to care about animal welfare because they feel if they do, they’d have to go vegan. We need a different strategy.

Ending factory farming is an extremely difficult problem. I don’t know if we’ll be able to do it. But here’s plan A.

4 The silent majority

Ending factory farming is, in some sense, the easy part. The hardest part is making serious progress on the suffering of other animals.

Here’s why this is a lot harder: when you tell people about what goes on in factory farms, they’re horrified. While people generally aren’t willing to go vegan, they typically think that the evil things that factory farms do should be outlawed. There are very few people who react with nonchalance to hearing that baby male chicks are ground up alive in the egg industry or that we sometimes roast pigs to death using 150-degree steam.

Contrast that with what happens to insects and wild animals. You can describe in detail a wild animal—even a mammal—getting slowly ripped apart, and the attitude most people have is indifference. They recognize that it’s brutal and painful for the animal, but they deny that it matters morally. With insects, the situation is even worse: people don’t even recognize that anything bad is happening at all.

Now, if we succeed in ending factory farming, my guess is that we’ll have a lot more success getting people concerned about wild animals. Once people think of factory farming as a serious moral atrocity akin to slavery and the holocaust, they’re likely to react more seriously to the suffering of animals in nature. But it’s far from clear that this would be enough. Most vegans, after all, don’t think wild animal suffering matters. And vegans generally don’t care much about insects.

We do have one advantage here: we’re right. Wild animal suffering does matter. A creature’s pain doesn’t stop being bad just because it lives in a grassy field. If a wild deer can suffer like a dog or a cat, this matters morally. Our neglect of insect welfare is almost entirely the byproduct of bias—if insects were large and cute, we’d care about them a lot more. Likewise, our neglect of wild animal suffering comes largely from the fact that we don’t naturally empathize with them, we don’t normally witness their suffering, and we’re biased in favor of the status quo. For more detail on why wild animal suffering matters, see here—for more on insect suffering mattering, see here.

A first step here is, of course, getting people concerned about the issue. I try to write regularly about the moral importance of wild animal and insect suffering in the hopes that someone will do something about it. It will be easier to get the world concerned about these issues if some people care about them.

Academics should write about this too. Imagine if in every philosophy class, discussion of wild animal suffering was a normal part of the curriculum. Research on insect sentience is valuable for this too—so that people cannot write off insect welfare by confidently declaring that insects aren’t sentient. An early paper by Eisemann arguing against insect sentience was a big part of getting people not to care very much about insects. The subsequent research that went into rebutting the early Eisemann paper strikes me as very valuable.

Fighting against factory farming is another important step in building concern for wild animals and bugs. As mentioned before, in a world that undergoes a major revolution with respect to concern for animals, we’re a lot likelier to care about the animals we currently neglect.

Next, we should try to push for incremental and popular reforms on behalf of these animals. The work done by the Shrimp Welfare Project is a great first step for getting people concerned about arthropods. When you tell most people that we crush the eyes of shrimp and suffocate or freeze them to death without stunning them first, they usually find this objectionable.

While the common canard about shrimp welfare is that it’s unpopular and damaging the EA brand, I don’t think that’s true. When you explain to most people that we can prevent thousands of shrimp from experiencing the pain of slowly freezing to death for just one dollar, they don’t find that so crazy. Heck, shrimp welfare even got on the daily show. It might be a thing that our opponents make fun of us for, but when you explain the ideas to people, they’re not really unpopular. And once people have already given some money to shrimp welfare, they’re a lot more sympathetic to the plight of the poor shrimp.

Along these lines, we should just destroy the bug farming industry. It has basically no upsides—it’s bad for the environment, workers, and the insects. Helping publicize the total shitshow that is the industry will make people a bit more opposed to insect farming, which might make them a bit more concerned about insects. If people are horrified about the industry microwaving and boiling live insects, that’s the first step on the road to total insect liberation. Once people think of themselves as the sorts of people who care a bit about insects, they’re likely to support low-effort actions to reduce insect suffering.

We should do the same for wild animals. There are a lot of wild animal welfare policies that are utterly commonsensical. You don’t have to be some weird animal welfare person to think it’s good to do things that prevent birds from crashing into windows and dying. For example, the new world screw worm lays maggots in the flesh of its victims. They eat their way from the inside out, causing its victims one of the most painful experiences that they can experience.

We’ve already eliminated this parasite from North America. We should eliminate it from South America too. And crucially, this is a pretty easy ask. Get rid of the weird worm creatures that leave maggots in the flesh of conscious creatures. The crucial test for the kinds of reforms we should push for: could you convince your grandmother of them?

Your grandmother, let us stipulate, has fairly normal priorities. She’s not an EA and to tell the truth finds the whole EA thing a bit kooky. If you explain to your grandmother: we’re trying to eradicate this horrible parasite that causes animals tons of suffering, while she might find it a little bit weird, she wouldn’t be repulsed and horrified. Odds are decent she’d think it was a good idea.

Similarly, if you tell her, “we’re trying to stop the shrimp companies from crushing the shrimps’ eyes,” probably she’ll think that makes sense. In contrast, if you say “we’re paving over nature to reduce nematode suffering,” she will not be pleased. If we want to make concern for bugs and wild animals popular, we’ll need to start with the kinds of proposals that make sense to people. Only later, after we get large-scale public buy-in, should we think about the more ambitious projects that AGI might carry out that might make animals in nature have mostly great lives rather than terrible ones.

5 Getting people to oppose torture is a lot like getting people to torture

So far, a lot of the steps I’ve discussed have involved gradual escalations in demands. There’s a reason for this: if your demands escalate gradually, it’s a lot easier to get people on board.

Before I first went vegan, I cut out the animal products that cause the most suffering from my diet. The only animal products I kept eating were wild-caught fish, beef, and dairy. But then, once I’d cut out the other animals products, cutting out the few remaining ones didn’t feel as difficult. It’s easy to make lots of consecutive little changes than one change all at once.

Probably the most famous example of this being successful was in getting people to torture the innocent.

There are two famous psychology experiments, where the headline takeaway is “wow, humans will do some pretty evil stuff in surprising numbers.” One is the Stanford prison experiment, though that one suffers from the minor defect of being total bullshit. The other is the Milgram experiment, and it was actually well done (plus, it has replicated repeatedly).

In this one, Stanley Milgram had scientific-looking people in white lab coats instruct participants to inflict an electric shock on a man when he got the wrong answers to questions. He repeatedly got the wrong answers, and would thrash and scream and beg to be released. The majority of people went along with this, continuing to inflict electric shocks on the victim even after he became despondent.

In one sentence: most people were willing to torture an innocent person to death when instructed to by an authoritative man in a white lab coat.

The gradual buildup was a big part of the success of this strategy. People were wary about getting off at any particular stop, because if you get off at 250 volts, you feel guilty that you didn’t get off sooner. At 150 volts, the man begged to be released. Almost no one stopped the shocks after 150 volts. Once people were committed, they didn’t stop.

If you can get people to torture an innocent man to the point of death by gradually escalating your demands, can you get them to care about animals being tortured to death in the same way? It’s worth a shot.

People don’t like feeling bad about their past decisions. Once they’re a bit bought in to a cause, once you’ve gotten your foot in the door, it’s a lot easier to get them to take future actions in support of the cause. If people donate to help animals, they’re likelier to vote on behalf of animals. In a world where half of people offset their meat consumption, my guess is we could get pretty ambitious animal welfare laws passed in short order.

So the lessons of Milgram:

1. Escalate gradually.

2. Once people have already started down a path, they’re reluctant to go back, especially if going back would make their previous actions look like mistakes.

6 But what of the coming superintelligent AI?

We’re now 4,000 words into an article about the end of animal exploitation, and I’ve barely mentioned the biggest reason for optimism: the coming AI.

It seems pretty likely that in the near future, we’ll have superintelligent AI. Progress in AI has been ridiculously rapid and shows no sign of stopping. Leading experts generally think odds are non-trivial that we’ll have AI that can surpass humans across cognitive fields in the next few years—and odds are quite high that we’ll have this within decades.

This matters for animals for two reasons. First, it means we’ll have unprecedented ability to replace animal products. AI innovation could help crack lab meat. Second, if AI is smart, it might apply its smartness to morality and see that we don’t have good reason to neglect the interests of animals. The exact details here depend on which kinds of AI we get, but it’s not impossible. As AIs get more and more dominant, their moral beliefs will increasingly matter.

How should this affect our priorities?

As a first step, we should work hard to incorporate concern for animals in the AI model specs. Model specs describe what AIs’ values are. It’s important that the leading AI companies make sure that the powerful AIs that they’re building care about animals. If AI is concerned about animals and becomes superintelligent, it’s a lot likelier to be sympathetic to big, ambitious projects to reduce animal suffering. For example, I’m super happy that Claude’s constitution instructs it to consider the “welfare of animals and of all sentient beings.”

Second, we should try to accelerate AI for epistemics. This involves using AI to help people form true beliefs. Our current neglect of animal suffering comes from having lots of beliefs that are wrong and silly. So helping increase the reliability of our belief-forming process makes it likelier that people will care more about animals.

Third, it should make us a lot more sympathetic to efforts to turn over important decisions to AI. Humans don’t care about most animals at all. AI might do better. My guess is that AI will be a lot better for animals than humans on account of:

1. Having been trained in an environment to be compassionate, rather than to maximally adroitly pass on its genes—often through rape and murder.

2. Being superintelligent.

3. Already being nicer to animals than humans. For example, here’s what Grok said when I asked if insect welfare would be a big deal if it turned out that insects experienced most of the world’s suffering:

Yes, improving insect welfare would be an overwhelmingly big deal—potentially the single largest moral priority humanity has ever faced. If the vast majority of all suffering on Earth is borne by insects (due to their immense numbers, short lives filled with predation, starvation, disease, and environmental stress), then even modest reductions in that aggregate misery—through better pest management, habitat design that minimizes suffering, genetic interventions for painless existence, or ecosystem tweaks—would dwarf the impact of alleviating all human pain, poverty, and war combined.

Based!

Fourth, it means outreach to influential people at labs is more important. Influential people at the labs might be some of the most influential people in world history, building the robots who will control the future. Influencing them is important. This gives some reason to move to the Bay Area, where everything happens.

Fifth, this makes the other strategies that I endorsed earlier in this article likelier to succeed. If people have a superintelligent assistant, we’re likely to have more success in campaigning for animals—where our main enemy is irrationality and bias and error. If AI will allow people to think more intelligently, then if animals are more salient, that will get people thinking more intelligently about animals.

AI is the big thing happening in the world today. It is the most important factor with any ghost of a chance of enabling animal liberation. put it well:

Transformative AI should be at the center of all our strategic thinking. The large majority of our resources should go into interventions that have a very solid answer to the question:

How does this have a good chance of making AI go better for animals?

If you want to spend time and money on something that doesn’t have a precise, compelling answer, you should have a very good reason why it’s worth doing anyways.

7 Flesh and blood and powers and principalities

This is a hard fight to win. We are going up against an established multi-trillion dollar industry, entrenched human bias, and self-interest. On our side, we have just two weapons: goodness and truth. We fight for creatures who most people don’t care about because they look weird.

But still, the fight isn’t over. The case for caring about most of the world’s suffering is very strong. The counterarguments are weak. And we have a number of extremely smart and hard-working activists. The animal rights movement has spared billions of animals from cages. We’ve helped billions of shrimp. We’ve made concern for neglected animals increasingly mainstream.

History’s great moral revolutions have involved moral circle expansion—caring about more and more beings who were once neglected. The ultimate culmination of this march of progress would be growing our moral circle to include every sentient being. Once we do that, we’ll be in a position to build a bright and glorious future, where every tear is wiped away, where even the little animals are taken care of. That’s a future worth fighting for.

Now, it’s certainly true that any individual article won’t change the minds of most people. But how could it be otherwise? If most articles convinced most people, then readers would constantly be buffeted around by whatever they last read—one day a Catholic, the next a radical vegan, the next an atheist.

Most of the biggest ideas I believe today came to me through some singular article or video. I first encountered effective altruism through a TED Talk, wild animal suffering through an article, and animal welfare because of ’s video. I wasn’t immediately convinced. But these pieces introduced me to the central ideas that I came to see were correct, after I reflected more.

When Scott Alexander wrote a post arguing that people should pledge to give away 10% of their lifetime earnings, he was able to get 60 people to take the pledge. That probably saved about 120 lives. About 30 people who have taken the pledge mentioned my posts, which probably saved about 60 lives.

I am not Scott Alexander. I don’t have his amazing ability to write six hundred pages filled with subtle Kabbalistic analysis. My blogging approach has generally been “write things that I think are true, in ways that are clear, and give solid arguments for them.” Scott Alexander is some kind of weird blogging witch—I am not a witch. I’m you.

(Funniest campaign ad in history!)

But by writing a bunch of articles about EA topics, I’ve been able to raise enough money to save hundreds of lives and help billions of shrimp. When some bloggers got together and did a shrimp fundraiser, we were able to raise $137,000 for the Shrimp Welfare Project—enough to spare about 2 billion shrimp.

A recent survey of my readers revealed that:

• 29% said my articles had convinced them to donate somewhere they wouldn’t have donated to otherwise. In many cases, this was just a few dollars given once—but in just as many cases, people set up monthly donations to highly effective charities. Some people said it was counterfactually responsible for them giving thousands of dollars to FarmKind.

• 7.8% said the blog was responsible for them taking a different career or doing some other high-impact project that they wouldn't have otherwise done.

• 66.1% said I’d changed their mind on some topic.

• 7.5% said the blog was responsible for them taking the Giving What We Can pledge. If the real number is even 1%, then this would make the blog counterfactually responsible for 100 pledges and about 200 lives saved.

• 35% said the blog contributed to them going vegan, vegetarian, or being some kind of reducetarian.

I don’t think I’m particularly special in this regard. I’m not an especially naturally skilled writer (just look at my earliest blog posts). If you write true things reasonably clearly, you can probably have a pretty big impact. If you have a blog with 100 subscribers, and convince one of them to take the pledge, that’s about as much counterfactual impact as giving away $10,000.

A bunch of young interesting EA writers have started substacks: , , , , and . My guess is that on average each person who has done this has gotten lots of money donated to effective charities. This is an insane fact about our world. By starting a blog on the internet and posting regularly, you can save a bunch of people’s lives, in expectation.

This has some big implications.

First, if you are thinking about starting a blog, do it. Maybe, probably it will go nowhere. But even if it remains pretty small, you might just save someone’s life, or spare a bunch of sentient beings from extreme suffering. Blasting important ideas into the ether is really impactful. The only situation in which I’d recommend against this is if there’s some other super impactful thing that it would compete with. But if you’re a student or an academic, start writing on the internet about important topics!

Probably you should do this even if you haven’t given much thought to blogging before. Before I started blogging, I didn’t think of myself as a writer. I didn’t even like writing much. I just started it on a whim because had some objections to utilitarianism that I thought were wrong!

Second, if you have a blog, make sure to write about important topics. It’s easy to get lost in trying to be novel and only focus on cool intellectual topics. But you should try to regularly say things that are important, even if they’re not that novel. It isn’t a brilliant insight that you should give away your income to effective charities, but if you write that down in an article, and send it out to your readers, you might save a bunch of lives. When writing a post, have in the back of your mind the question: if I convinced every one of my readers that what I was saying was true, would this make the world better?

A number of pretty influential writers read this blog. If any of you people are reading this article, I am speaking to you: please write a post arguing that people should give money to effective charities. Suggest explicitly that they take the Giving What We Can Pledge, and explain what it is. Mention some specific highly-effective charities. Remember: when Scott Alexander did this, he prevented around 120 deaths! You can prevent large numbers of deaths too, just by writing an article.

Even if you already said something vaguely similar in 2018 or whatever, say it again. You should try to write reasonably frequently about impactful subjects. Maybe a few people will find it somewhat repetitive, but other people will hear it for the first time, and take the pledge, and extra children will not die! My sense is that basically every big writer writes way too infrequently about impactful subjects.

Return to the classic drowning child experiment. If a bunch of children were drowning in far-away ponds, and the only way to save them was to publish a blog post suggesting that people nearby pull them out of ponds, then obviously you should publish that blog post! This would be true even if you published a vaguely similar blogpost a bunch of years ago that some of them have read. On top of this, write about high impact careers!

Third, if you start writing, and your articles are good, don’t stop! A common failure mode is people start blogs, they go pretty far, and then they sort of run out of ideas and peter out (cough cough). Now, in some cases this is because they start working on other important things (e.g. I hear is out saving the world, so he has an excuse). But and —you are without excuse, sorry. You were both gifted with supernatural writing powers and should use them for good!1

Fourth, even if you are not a blogger, but instead a member of the hoi polloi, try to share articles on important topics with other people. Restack them, share them with friends and family, etc. It is very easy to hit the restack button when you come across important articles on EA topics—doing so enables these articles to get a wider reach.

Fifth, I think it’s plausible that becoming a paid subscriber for high impact EA blogs is pretty impactful. Now, obviously this is very self-serving, so you have some reason to be suspicious. But having more paid subscribers helps a blog do better in the subtack rankings, thus potentially moving large amounts of money to effective charities. It wouldn’t surprise me if one of the better things you can do with a few dollars a month is becoming a paid subscriber to ’s blog, for example, or ’s (especially on the off chance it would rouse him from his slumber).

Sixth, if you start a blog, try to ensure it isn’t just read by EAs! While you should sometimes write about the most important topics, you shouldn’t only write about the most important topics. If I only talked about shrimp, that wouldn’t even be good for the shrimp. Ideally you want to make EA ideas more salient among non-EAs.

Do other things to make sure your blog gets more exposure. For example, I just got Twitter premium and started cross-posting my articles on Twitter. My guess is that more people should do things like that. If doing that for every single article gets one extra person to take the pledge, it will have much more than paid off.

A common blog failure mode is getting nerd-sniped by interesting topics that no one else cares about. Now, I’m not above this, as my fifty billion articles about subtleties in anthropic reasoning attest. But writing about topics that are neither important nor interesting to most people should be thought of as akin to drinking: fine in moderation, objectionable in excess. That is, unless you are Scott Alexander, and you can write an engaging piece about literally anything.

There are real and significant problems in the world. Much can be done about them. It is easy to get lost investigating cool puzzles that no one cares about. But you should know that doing so trades off against convincing people of what matters most.