Pascal’s Mugging

Nov 10 JDN 2458798

In the Singularitarian community there is a paradox known as “Pascal’s Mugging”. The name is an intentional reference to Pascal’s Wager (and the link is quite apt, for reasons I’ll discuss in a later post.)

There are a few different versions of the argument; Yudkowsky’s original argument in which he came up with the name “Pascal’s Mugging” relies upon the concept of the universe as a simulation and an understanding of esoteric mathematical notation. So here is a more intuitive version:

A strange man in a dark hood comes up to you on the street. “Give me five dollars,” he says, “or I will destroy an entire planet filled with ten billion innocent people. I cannot prove to you that I have this power, but how much is an innocent life worth to you? Even if it is as little as $5,000, are you really willing to bet on ten trillion to one odds that I am lying?”

Do you give him the five dollars? I suspect that you do not. Indeed, I suspect that you’d be less likely to give him the five dollars than if he had merely said he was homeless and asked for five dollars to help pay for food. (Also, you may have objected that you value innocent lives, even faraway strangers you’ll never meet, at more than $5,000 each—but if that’s the case, you should probably be donating more, because the world’s best charities can save a live for about $3,000.)

But therein lies the paradox: Are you really willing to bet on ten trillion to one odds?

This argument gives me much the same feeling as the Ontological Argument; as Russell said of the latter, “it is much easier to be persuaded that ontological arguments are no good than it is to say exactly what is wrong with them.” It wasn’t until I read this post on GiveWell that I could really formulate the answer clearly enough to explain it.

The apparent force of Pascal’s Mugging comes from the idea of expected utility: Even if the probability of an event is very small, if it has a sufficiently great impact, the expected utility can still be large.

The problem with this argument is that extraordinary claims require extraordinary evidence. If a man held a gun to your head and said he’d shoot you if you didn’t give him five dollars, you’d give him five dollars. This is a plausible claim and he has provided ample evidence. If he were instead wearing a bomb vest (or even just really puffy clothing that could conceal a bomb vest), and he threatened to blow up a building unless you gave him five dollars, you’d probably do the same. This is less plausible (what kind of terrorist only demands five dollars?), but it’s not worth taking the chance.

But when he claims to have a Death Star parked in orbit of some distant planet, primed to make another Alderaan, you are right to be extremely skeptical. And if he claims to be a being from beyond our universe, primed to destroy so many lives that we couldn’t even write the number down with all the atoms in our universe (which was actually Yudkowsky’s original argument), to say that you are extremely skeptical seems a grievous understatement.

That GiveWell post provides a way to make this intuition mathematically precise in terms of Bayesian logic. If you have a normal prior with mean 0 and standard deviation 1, and you are presented with a likelihood with mean X and standard deviation X, what should you make your posterior distribution?

Normal priors are quite convenient; they conjugate nicely. The precision (inverse variance) of the posterior distribution is the sum of the two precisions, and the mean is a weighted average of the two means, weighted by their precision.

So the posterior variance is 1/(1 + 1/X^2).

The posterior mean is 1/(1+1/X^2)*(0) + (1/X^2)/(1+1/X^2)*(X) = X/(X^2+1).

That is, the mean of the posterior distribution is just barely higher than zero—and in fact, it is decreasing in X, if X > 1.

For those who don’t speak Bayesian: If someone says he’s going to have an effect of magnitude X, you should be less likely to believe him the larger that X is. And indeed this is precisely what our intuition said before: If he says he’s going to kill one person, believe him. If he says he’s going to destroy a planet, don’t believe him, unless he provides some really extraordinary evidence.

What sort of extraordinary evidence? To his credit, Yudkowsky imagined the sort of evidence that might actually be convincing:

If a poorly-dressed street person offers to save 10(10^100) lives (googolplex lives) for $5 using their Matrix Lord powers, and you claim to assign this scenario less than 10-(10^100) probability, then apparently you should continue to believe absolutely that their offer is bogus even after they snap their fingers and cause a giant silhouette of themselves to appear in the sky.

This post he called “Pascal’s Muggle”, after the term from the Harry Potter series, since some of the solutions that had been proposed for dealing with Pascal’s Mugging had resulted in a situation almost as absurd, in which the mugger could exhibit powers beyond our imagining and yet nevertheless we’d never have sufficient evidence to believe him.

So, let me go on record as saying this: Yes, if someone snaps his fingers and causes the sky to rip open and reveal a silhouette of himself, I’ll do whatever that person says. The odds are still higher that I’m dreaming or hallucinating than that this is really a being from beyond our universe, but if I’m dreaming, it makes no difference, and if someone can make me hallucinate that vividly he can probably cajole the money out of me in other ways. And there might be just enough chance that this could be real that I’m willing to give up that five bucks.

These seem like really strange thought experiments, because they are. But like many good thought experiments, they can provide us with some important insights. In this case, I think they are telling us something about the way human reasoning can fail when faced with impacts beyond our normal experience: We are in danger of both over-estimating and under-estimating their effects, because our brains aren’t equipped to deal with magnitudes and probabilities on that scale. This has made me realize something rather important about both Singularitarianism and religion, but I’ll save that for next week’s post.

What can we do to make the world a better place?

JDN 2457475

There are an awful lot of big problems in the world: war, poverty, oppression, disease, terrorism, crime… I could go on for awhile, but I think you get the idea. Solving or even mitigating these huge global problems could improve or even save the lives of millions of people.

But precisely because these problems are so big, they can also make us feel powerless. What can one person, or even a hundred people, do against problems on this scale?

The answer is quite simple: Do your share.

No one person can solve any of these problems—not even someone like Bill Gates, though he for one at least can have a significant impact on poverty and disease because he is so spectacularly mind-bogglingly rich; the Gates Foundation has a huge impact because it has as much wealth as the annual budget of the NIH.

But all of us together can have an enormous impact. This post today is about helping you see just how cheap and easy it would be to end world hunger and cure poverty-related diseases, if we simply got enough people to contribute.

The Against Malaria Foundation releases annual reports for all their regular donors. I recently got a report that my donations personally account for 1/100,000 of their total assets. That’s terrible. The global population is 7 billion people; in the First World alone it’s over 1 billion. I am the 0.01%, at least when it comes to donations to the Against Malaria Foundation.

I’ve given them only $850. Their total assets are only $80 million. They shouldn’t have $80 million—they should have $80 billion. So, please, if you do nothing else as a result of this post, go make a donation to the Against Malaria Foundation. I am entirely serious; if you think you might forget or change your mind, do it right now. Even a dollar would be worth it. If everyone in the First World gave $1, they would get 12 times as much as they currently have.

GiveWell is an excellent source for other places you should donate; they rate charities around the world for their cost-effectiveness in the only way worth doing: Lives saved per dollar donated. They don’t just naively look at what percentage goes to administrative costs; they look at how everything is being spent and how many children have their diseases cured.

Until the end of April, UNICEF is offering an astonishing five times matching funds—meaning that if you donate $10, a full $50 goes to UNICEF projects. I have really mixed feelings about donors that offer matching funds (So what you’re saying is, you won’t give if we don’t?), but when they are being offered, use them.

All those charities are focused on immediate poverty reduction; if you’re looking for somewhere to give that fights Existential Risk, I highly recommend the Union of Concerned Scientists—one of the few Existential Risk organizations that uses evidence-based projections and recognizes that nuclear weapons and climate change are the threats we need to worry about.

And let’s not be too anthropocentrist; there are a lot of other sentient beings on this planet, and Animal Charity Evaluator can help you find which charities will best improve the lives of other animals.

I’ve just listed a whole bunch of ways you can give money—and that probably is the best thing for you to give; your time is probably most efficiently used working in your own profession whatever that may be—but there are other ways you can contribute as well.

One simple but important change you can make, if you haven’t already, is to become vegetarian. Even aside from the horrific treatment of animals in industrial farming, you don’t have to believe that animals deserve rights to understand that meat is murder. Meat production is a larger contributor to global greenhouse gas emissions than transportation, so everyone becoming vegetarian would have a larger impact against climate change than taking literally every car and truck in the world off the road. Since the world population is less than 10 billion, meat is 18% of greenhouse emissions and the IPCC projects that climate change will kill between 10 and 100 million people over the next century, every 500 to 5000 new vegetarians saves a life.

You can move your money from a bank to a credit union, as even the worst credit unions are generally better than the best for-profit banks, and the worst for-profit banks are very, very bad. The actual transition can be fairly inconvenient, but a good credit union will provide you with all the same services, and most credit unions link their networks and have online banking, so for example I can still deposit and withdraw from my University of Michigan Credit Union account while in California.

Another thing you can do is reduce your consumption of sweatshop products in favor of products manufactured under fair labor standards. This is harder than it sounds; it can be very difficult to tell what a company’s true labor conditions are like, as the worst companies work very hard to hide them (now, if they worked half as hard to improve them… it reminds me of how many students seem willing to do twice as much work to cheat as they would to simply learn the material in the first place).

You should not simply stop buying products that say “Made in China”; in fact, this could be counterproductive. We want products to be made in China; we need products to be made in China. What we have to do is improve labor standards in China, so that products made in China are like products made in Japan or Korea—skilled workers with high-paying jobs in high-tech factories. Presumably it doesn’t bother you when something says “Made in Switzerland” or “Made in the UK”, because you know their labor standards are at least as high as our own; that’s where I’d like to get with “Made in China”.

The simplest way to do this is of course to buy Fair Trade products, particularly coffee and chocolate. But most products are not available Fair Trade (there are no Fair Trade computers, and only loose analogues for clothing and shoes).

Moreover, we must not let the perfect be the enemy of the good; companies that have done terrible things in the past may still be the best companies to support, because there are no alternatives that are any better. In order to incentivize improvement, we must buy from the least of all evils for awhile until the new competitive pressure makes non-evil corporations viable. With this in mind, the Fair Labor Association may not be wrong to endorse companies like Adidas and Apple, even though they surely have substantial room to improve. Similarly, few companies on the Ethisphere list are spotless, but they probably are genuinely better than their competitors. (Well, those that have competitors; Hasbro is on there. Name a well-known board game, and odds are it’s made by a Hasbro subsidiary: they own Parker Brothers, Milton Bradley, and Wizards of the Coast. Wikipedia has their own category, Hasbro subsidiaries. Maybe they’ve been trying to tell us something with all those versions of Monopoly?)

I’m not very happy with the current state of labor standards reporting (much less labor standards enforcement), so I don’t want to recommend any of these sources too highly. But if you are considering buying from one of three companies and only one of them is endorsed by the Fair Labor Association, it couldn’t hurt to buy from that one instead of the others.

Buying from ethical companies will generally be more expensive—but rarely prohibitively so, and this is part of how we use price signals to incentivize better behavior. For about a year, BP gasoline was clearly cheaper than other gasoline, because nobody wanted to buy from BP and they were forced to sell at a discount after the Deepwater Horizon disaster. Their profits tanked as a result. That’s the kind of outcome we want—preferably for a longer period of time.

I suppose you could also save money by buying cheaper products and then donate the difference, and in the short run this would actually be most cost-effective for global utility; but (1) nobody really does that; people who buy Fair Trade also tend to donate more, maybe just because they are more generous in general, and (2) in the long run what we actually want is more ethical businesses, not a system where businesses exploit everyone and then we rely upon private charity to compensate us for our exploitation. For similar reasons, philanthropy is a stopgap—and a much-needed one—but not a solution.

Of course, you can vote. And don’t just vote in the big name elections like President of the United States. Your personal impact may actually be larger from voting in legislatures and even local elections and ballot proposals. Certainly your probability of being a deciding vote is far larger, though this is compensated by the smaller effect of the resulting policies. Most US states have a website where you can look up any upcoming ballots you’ll be eligible to vote on, so you can plan out your decisions well in advance.

You may even want to consider running for office at the local level, though I realize this is a very large commitment. But most local officials run uncontested, which means there is no real democracy at work there at all.

Finally, you can contribute in some small way to making the world a better place simply by spreading the word, as I hope I’m doing right now.

Scope neglect and the question of optimal altruism

JDN 2457090 EDT 16:15.

We’re now on Eastern Daylight Time because of this bizarre tradition of shifting our time zone forward for half of the year. It’s supposed to save energy, but a natural experiment in India suggests it actually increases energy demand. So why do we do it? Like every ridiculous tradition (have you ever tried to explain Groundhog Day to someone from another country?), we do it because we’ve always done it.
This week’s topic is scope neglect, one of the most pervasive—and pernicious—cognitive heuristics human beings face. Scope neglect raises a great many challenges not only practically but also theoretically—it raises what I call the question of optimal altruism.

The question is simple to ask yet remarkably challenging to answer: How much should we be willing to sacrifice in order to benefit others? If we think of this as a number, your solidarity coefficient (s), it is equal to the cost you are willing to pay divided by the benefit your action has for someone else: s B > C.

This is analogous to the biological concept relatedness (r), on which Hamilton’s Rule applies: r B > C. Solidarity is the psychological analogue; instead of valuing people based on their genetic similarity to you, you value them based on… well, that’s the problem.

I can easily place upper and lower bounds: The lower bound is zero: You should definitely be willing to sacrifice something to help other people—otherwise you are a psychopath. The upper bound is one: There’s no point in paying more cost than you produce in benefit, and in fact even paying the same cost to yourself as you yield in benefits for other people doesn’t make a lot of sense, because it means that your own self-interest is meaningless and the fact that you understand your own needs better than the needs of others is also irrelevant.

But beyond that, it gets a lot harder—and that may explain why we suffer scope neglect in the first place. Should it be 90%? 50%? 10%? 1%? How should it vary between friends versus family versus strangers? It’s really hard to say. And this inability to precisely decide how much other people should be worth to us may be part of why we suffer scope neglect.

Scope neglect is the fact that we are not willing to expend effort or money in direct proportion to the benefit it would have. When different groups were asked how much they would be willing to donate in order to save the lives of 2,000 birds, 20,000 birds, or 200,000 birds, the answers they gave were statistically indistinguishable—always about $80. But however much a bird’s life is worth to you, shouldn’t 200,000 birds be worth, well, 200,000 times as much? In fact, more than that, because the marginal utility of wealth is decreasing, but I see no reason to think that the marginal utility of birds decreases nearly as fast.

But therein lies the problem: Usually we can’t pay 200,000 times as much. I’d feel like a horrible person if I weren’t willing to expend at least $10 or an equivalent amount of effort in order to save a bird. To save 200,000 birds that means I’d owe $2 million—and I simply don’t have $2 million.

You can get similar results to the bird experiment if you use children—though, as one might hope, the absolute numbers are a bit bigger, usually more like $500 to $1000. (And this, it turns out, is actually about how much it actually costs to save a child’s life by a particularly efficient means, such as anti-malaria nets, de-worming, or direct cash transfer. So please, by all means, give $1000 to UNICEF or the Against Malaria Foundation. If you can’t give $1000, give $100; if you can’t give $100, give $10.) It doesn’t much matter whether you say that the project will save 500 children, 5,000 children, or 50,000 children—people still will give about $500 to $1000. But once again, if I’m willing to spend $1000 to save a child—and I definitely am—how much should I be willing to spend to end malaria, which kills 500,000 children a year? Apparently $500 million, which not only do I not have, I almost certainly will not make that much money cumulatively through my entire life. ($2 million, on the other hand, I almost certainly will make cumulatively—the median income of an economist is $90,000 per year, so if I work for at least 22 years with that as my average income I’ll have cumulatively made $2 million. My net wealth may never be that high—though if I get better positions, or I’m lucky enough or clever enough with the stock market it might—but my cumulative income almost certainly will. Indeed, the average gain in cumulative income from a college degree is about $1 million. Because it takes time—time is money—and loans carry interest, this gives it a net present value of about $300,000.)

But maybe scope neglect isn’t such a bad thing after all. There is a very serious problem with these sort of moral dilemmas: The question didn’t say I would single-handedly save 200,000 birds—and indeed, that notion seems quite ridiculous. If I knew that I could actually save 200,000 birds and I were the only one who could do it, dammit, I would try to come up with that $2 million. I might not succeed, but I really would try as hard as I could.

And if I could single-handedly end malaria, I hereby vow that I would do anything it took to achieve that. Short of mass murder, anything I could do couldn’t be a higher cost to the world than malaria itself. I have no idea how I’d come up with $500 million, but I’d certainly try. Bill Gates could easily come up with that $500 million—so he did. In fact he endowed the Gates Foundation with $28 billion, and they’ve spent $1.3 billion of that on fighting malaria, saving hundreds of thousands of lives.

With this in mind, what is scope neglect really about? I think it’s about coordination. It’s not that people don’t care more about 200,000 birds than they do about 2,000; and it’s certainly not that they don’t care more about 50,000 children than they do about 500. Rather, the problem is that people don’t know how many other people are likely to donate, or how expensive the total project is likely to be; and we don’t know how much we should be willing to pay to save the life of a bird or a child.

Hence, what we basically do is give up; since we can’t actually assess the marginal utility of our donation dollars, we fall back on our automatic emotional response. Our mind focuses itself on visualizing that single bird covered in oil, or that single child suffering from malaria. We then hope that the representative heuristic will guide us in how much to give. Or we follow social norms, and give as much as we think others would expect us to give.

While many in the effective altruism community take this to be a failing, they never actually say what we should do—they never give us a figure for how much money we should be willing to donate to save the life of a child. Instead they retreat to abstraction, saying that whatever it is we’re willing to give to save a child, we should be willing to give 50,000 times as much to save 50,000 children.

But it’s not that simple. A bigger project may attract more supporters; if the two occur in direct proportion, then constant donation is the optimal response. Since it’s probably not actually proportional, you likely should give somewhat more to causes that affect more people; but exactly how much more is an astonishingly difficult question. I really don’t blame people—or myself—for only giving a little bit more to causes with larger impact, because actually getting the right answer is so incredibly hard. This is why it’s so important that we have institutions like GiveWell and Charity Navigator which do the hard work to research the effectiveness of charities and tell us which ones we should give to.

Yet even if we can properly prioritize which charities to give to first, that still leaves the question of how much each of us should give. 1% of our income? 5%? 10%? 20%? 50%? Should we give so much that we throw ourselves into the same poverty we are trying to save others from?

In his earlier work Peter Singer seemed to think we should give so much that it throws us into poverty ourselves; he asked us to literally compare every single purchase and ask ourselves whether a year of lattes or a nicer car is worth a child’s life. Of course even he doesn’t live that way, and in his later books Singer seems to have realized this, and now recommends the far more modest standard that everyone give at least 1% of their income. (He himself gives about 33%, but he’s also very rich so he doesn’t feel it nearly as much.) I think he may have overcompensated; while if literally everyone gave at least 1% that would be more than enough to end world hunger and solve many other problems—world nominal GDP is over $70 trillion, so 1% of that is $700 billion a year—we know that this won’t happen. Some will give more, others less; most will give nothing at all. Hence I think those of us who give should give more than our share; hence I lean toward figures more like 5% or 10%.

But then, why not 50% or 90%? It is very difficult for me to argue on principle why we shouldn’t be expected to give that much. Because my income is such a small proportion of the total donations, the marginal utility of each dollar I give is basically constant—and quite high; if it takes about $1000 to save a child’s life on average, and each of these children will then live about 60 more years at about half the world average happiness, that’s about 30 QALY per $1000, or about 30 milliQALY per dollar. Even at my current level of income (incidentally about as much as I think the US basic income should be), I’m benefiting myself only about 150 microQALY per dollar—so my money is worth about 200 times as much to those children as it is to me.

So now we have to ask ourselves the really uncomfortable question: How much do I value those children, relative to myself? If I am at all honest, the value is not 1; I’m not prepared to die for someone I’ve never met 10,000 kilometers away in a nation I’ve never even visited, nor am I prepared to give away all my possessions and throw myself into the same starvation I am hoping to save them from. I value my closest friends and family approximately the same as myself, but I have to admit that I value random strangers considerably less.

Do I really value them at less than 1%, as these figures would seem to imply? I feel like a monster saying that, but maybe it really isn’t so terrible—after all, most economists seem to think that the optimal solidarity coefficient is in fact zero. Maybe we need to become more comfortable admitting that random strangers aren’t worth that much to us, simply so that we can coherently acknowledge that they aren’t worth nothing. Very few of us actually give away all our possessions, after all.

Then again, what do we mean by worth? I can say from direct experience that a single migraine causes me vastly more pain than learning about the death of 200,000 people in an earthquake in Southeast Asia. And while I gave about $100 to the relief efforts involved in that earthquake, I’ve spent considerably more on migraine treatments—thousands, once you include health insurance. But given the chance, would I be willing to suffer a migraine to prevent such an earthquake? Without hesitation. So the amount of pain we feel is not the same as the amount of money we pay, which is not the same as what we would be willing to sacrifice. I think the latter is more indicative of how much people’s lives are really worth to us—but then, what we pay is what has the most direct effect on the world.

It’s actually possible to justify not dying or selling all my possessions even if my solidarity coefficient is much higher—it just leads to some really questionable conclusions. Essentially the argument is this: I am an asset. I have what economists call “human capital”—my health, my intelligence, my education—that gives me the opportunity to affect the world in ways those children cannot. In my ideal imagined future (albeit improbable) in which I actually become President of the World Bank and have the authority to set global development policy, I myself could actually have a marginal impact of megaQALY—millions of person-years of better life. In the far more likely scenario in which I attain some mid-level research or advisory position, I could be one of thousands of people who together have that sort of impact—which still means my own marginal effect is on the order of kiloQALY. And clearly it’s true that if I died, or even if I sold all my possessions, these events would no longer be possible.

The problem with that reasoning is that it’s wildly implausible to say that everyone in the First World are in this same sort of position—Peter Singer can say that, and maybe I can say that, and indeed hundreds of development economists can say that—but at least 99.9% of the First World population are not development economists, nor are they physicists likely to invent cold fusion, nor biomedical engineers likely to cure HIV, nor aid workers who distribute anti-malaria nets and polio vaccines, nor politicians who set national policy, nor diplomats who influence international relations, nor authors whose bestselling books raise worldwide consciousness. Yet I am not comfortable saying that all the world’s teachers, secretaries, airline pilots and truck drivers should give away their possessions either. (Maybe all the world’s bankers and CEOs should—or at least most of them.)

Is it enough that our economy would collapse without teachers, secretaries, airline pilots and truck drivers? But this seems rather like the fact that if everyone in the world visited the same restaurant there wouldn’t be enough room. Surely we could do without any individual teacher, any individual truck driver? If everyone gave the same proportion of their income, 1% would be more than enough to end malaria and world hunger. But we know that everyone won’t give, and the job won’t get done if those of us who do give only 1%.

Moreover, it’s also clearly not the case that everything I spend money on makes me more likely to become a successful and influential development economist. Buying a suit and a car actually clearly does—it’s much easier to get good jobs that way. Even leisure can be justified to some extent, since human beings need leisure and there’s no sense burning myself out before I get anything done. But do I need both of my video game systems? Couldn’t I buy a bit less Coke Zero? What if I watched a 20-inch TV instead of a 40-inch one? I still have free time; could I get another job and donate that money? This is the sort of question Peter Singer tells us to ask ourselves, and it quickly leads to a painfully spartan existence in which most of our time is spent thinking about whether what we’re doing is advancing or damaging the cause of ending world hunger. But then the cost of that stress and cognitive effort must be included; but how do you optimize your own cognitive effort? You need to think about the cost of thinking about the cost of thinking… and on and on. This is why bounded rationality modeling is hard, even though it’s plainly essential to both cognitive science and computer science. (John Stuart Mill wrote an essay that resonates deeply with me about how the pressure to change the world drove him into depression, and how he learned to accept that he could still change the world even if he weren’t constantly pressuring himself to do so—and indeed he did. James Mill set out to create in his son, John Stuart Mill, the greatest philosopher in the history of the world—and I believe that he succeeded.)

Perhaps we should figure out what proportion of the world’s people are likely to give, and how much we need altogether, and then assign the amount we expect from each of them based on that? The more money you ask from each, the fewer people are likely to give. This creates an optimization problem akin to setting the price of a product under monopoly—monopolies maximize profits by carefully balancing the quantity sold with the price at which they sell, and perhaps a similar balance would allow us to maximize development aid. But wouldn’t it be better if we could simply increase the number of people who give, so that we don’t have to ask so much of those who are generous? That means tax-funded foreign aid is the way to go, because it ensures coordination. And indeed I do favor increasing foreign aid to about 1% of GDP—in the US it is currently about $50 billion, 0.3% of GDP, a little more than 1% of the Federal budget. (Most people who say we should “cut” foreign aid don’t realize how small it already is.) But foreign aid is coercive; wouldn’t it be better if people would give voluntarily?

I don’t have a simple answer. I don’t know how much other people’s lives ought to be worth to us, or what it means for our decisions once we assign that value. But I hope I’ve convinced you that this problem is an important one—and made you think a little more about scope neglect and why we have it.