The Hoover Hog

Editor's note: This is the fourth in a series of Hoover Hog interviews.

____________________

INTRODUCTION

____________________

WORLDS OF SUFFERING: AN INTERVIEW WITH ALAN DAWRST

HOOVER HOG: You approach moral problems from a the perspective of utilitarianism. Why? What attracts you to the utilitarian pedigree, when there are so many other ways to sort through ethical questions?

ALAN DAWRST: I'm not terribly interested in a lot of aspects of ethics as it's traditionally construed: virtue, rules for good conduct, judgement of actions, etc. I'm more concerned with getting things done. Here's a simple illustration, extended from the example of the drowning child that Peter Singer gives in his essay, Famine, Affluence, and Morality. Suppose you come across a pond in which 15 children are in the process of drowning. Do you sit and ponder whether it's part of the good life for you to rescue them? Do you ask whether there's a universal moral law commanding you to pull them out? I suppose you could. But my response would be to think, Wow, if I were in their position, I wouldn't want to drown. Let me see what I can do to help them! That simple feeling of empathy seems to me the most natural, basic response to the situation. You can talk about all kinds of abstract moral values and principles of righteousness, but at the end of the day, those don't have anything to do with why I actually care.

In practice, the world really is a big pond with kids drowning all the time: There are billions of people suffering from preventable poverty, disease, and violence, billions of animals enduring dreadful lives on factory farms, and orders of magnitude more animals in the wild that are sick, hungry, or being eaten alive. My response is, "That's terrible! What can I do to make things better?"

Your writing is overwhelmingly concerned with the problem of suffering. Why should people care about suffering, other than their own?

The most I can say is that many people are concerned with the suffering of others, and I'm thankful for that. There's nothing I could say to a paperclip maximizer that would make it start caring about suffering, for example.

Your emphasis on reducing suffering  leads you to consider problems that may never occur to most people, such as the possibility of invertebrate sentience and the harm-multiplying potential of "laboratory universes." Why are you drawn to areas that are so widely ignored, even among ethicists?

Well, to some extent, it's because areas that have been ignored can potentially offer high expected returns from being investigated (sort of like undervalued securities). But unlike the stock market, where most investors care a lot about their expected returns, there aren't many people who try to consistently maximize the expected value of their actions even when the probabilities involved become really small and the payoffs really big. So there's not much of an efficient market hypothesis for prevention of suffering.

I think the possibility that insects can suffer is a good example. The scientific jury is still out as to whether invertebrates can feel pain, and in light of the human tendancy to be overconfident, I wouldn't assign a probability lower than, say, 0.001 to insect suffering. (My personal probability is closer to 0.2, but I'm not an expert. I know one entomologist whose probability is close to 1.0.) The number of insects in the world is vast (10^18, a billion billion), so the expected amount of insect suffering is still huge. Even if your probability of insect sentience were one in a million, the expected number of suffering insects would still outnumber humans 100 times over.

Yet, I'm unaware of any animal-welfare organizations addressing this issue, except for one spoof website and perhaps a few vegans who want to avoid honey for reasons of bee suffering. From a PR perspective, this may be sensible: If a lot of people are still resistant to caring about, say, chicken welfare, they'll be put off altogether by the suggestion that they should care about the feelings of the flies in their house. (I should add that I don't think concern for insect welfare necessarily implies we should avoid killing insects. Indeed, the opposite may be true: If I were a fly, I might prefer being squashed by a human as quickly as possible rather than dying slowly by disease or parasites a few weeks later.)

Still, it's not the case that humans are powerless to prevent insect suffering. To take one example, animal-welfare organizations could pay farmers to replace their current insecticides with other insecticides thought to be faster-acting and less painful. I don't necessarily recommend this as an optimal use of resources, but the fact that this crude proposal competes in cost-effectiveness with, say, distributing literature on factory farming is a proof-of-concept for why insect welfare shouldn't be taken off the table.

Getting back to your question, I think part of the reason for neglect of these issues is scope insensitivity: The fact that preventing 1000 times as much suffering doesn't feel 1000 times as satisfying. There's a natural desire to feel like you're doing something to help, but exactly what that is may not seem as important. Yet, even using basic measures like US$/DALY, the cost-effectiveness of something as tame as different population-based health interventions can vary by several orders of magnitude. There are a few organizations (e.g., GiveWell and, hopefully soon, Giving What We Can) that aim to help donors choose more efficient causes. And of course, cost-effectiveness analysis is what many economists do all the time, though often in crude and misleading terms where costs and benefits have to be translated into dollars. (A QALY in developed countries is worth, say, $60,000, while interventions in the developing world for tens or hundreds of dollars per QALY aren't funded.)

It's somewhat puzzling why economists, who are trained to think in terms of cost-effectiveness, aren't interested in speculative high-risk high-expected-return topics like insect welfare. I guess the reason is just that they have a different objective function. It's interesting that a lot of utilitarian decision-making principles are widely taught in the business world, in the context of strategic management and corporate finance -- only there, the decision rule says Maximize expected net present value, rather than, say, Maximize expected reduction in suffering. (Businesses are also sometimes risk-averse, while altruists should try not to be.)

You are a former atheist who converted to Christianity on account of Pascal's Wager. This is rare, to say the least. Can you describe how it happened?

I like Eliezer Yudkowsky's description of utilitarianism: Shut up and multiply. It's very much a bullet-biting mindset where, if someone can show that action A clearly has higher expected value than action B, you're obligated to go with action A. As a result of this, my beliefs have been (and remain) highly volatile, subject to potentially dramatic changes as I learn new information or hear new arguments.

Ever since I was about six years old, I had been a materialist atheist, and I regarded religion in the same way as, say, ghosts and UFOs. But as I thought about the situation from a shut up and multiply standpoint, I decided that a simple analysis in terms of factual probability was inappropriate, because of the magnitude of the payoffs involved. In simple terms, any nonzero probability of Christianity (or, more precisely, differential in probability between Christianity and hypotheses according to which I'll be punished eternally iff I convert to Christianity) amounts to infinite expected suffering, which ought to outweigh any finite costs. (In practice, this is impossible on account of akrasia. Christians would just call that the problem of human fallenness, I suppose.)

What I described, of course, is just Pascal's wager, though Pascal himself didn't mention hell but merely losing out on heaven. It's phrased in fancy decision-theoretic language, but for me, it amounts to nothing more than an intellectualized fear of hell: The intuition that, as one site notes, Eternity is a very long time to be wrong, and I'm not willing to risk it.

In any event, after about a year of playing around with the wager and trying to come up with rationalizations for why it didn't hold, I bit the bullet and decided to try becoming a Christian. The reasons for Christianity instead of, say, Islam were a combination of factual probability, non-factual Pascalian considerations, and convenience. This decision isn't necessarily set in stone: Given persuasive evidence in favor of another religion (including atheism), I could change my mind. But after almost two years of research into various other possibilities, this hasn't happened.

It seems safe to assume that most philosophical utilitarians are nonbelievers, but you have argued that the demands of infinite decision theory unite utilitarianism with theistically predicated moral reasoning, particularly as expressed in Christian apologetics. How so?

I should clarify that while Pascal's wager is a very natural idea within a utilitarian mindset, the argument itself is egoist, appealing to my own self-interested desire to reduce my chance of hell. (It's theoretically possible that utilitarianism could recommend defying God's will in order to prevent large numbers of other people from going to hell, for instance.)

But it is true that utilitarianism shares some commonalities with Christianity (and religion generally) to the extent both recognize that individual worldly happiness for an individual is trivial in comparison to the vast expected value of devoting effort to other things. (Think of the common charge that utilitarianism is too demanding.) As I mentioned before, in practice, neither utilitarians nor Christians can actually live up to this standard, but they can do better than not trying at all.

In reading your work, it is easy to come away with the sense that your commitment to Christianity derives almost exclusively from intellectual or logical considerations. Your reasoning is informed by surreal numbers and probabilistic inductions of potentially Vast or infinite consequence. Is there a spiritual dimension to your faith? Or are you simply cornered by a rational gambit?

Jerram Barrs made this comment: I often have people say to me that they are not a religious person, as if they belong to some kind of different category. My response is always to say, 'In the sense that you mean, I am not religious either. I am interested in what is true.' I assign a lower subjective probability to Christianity than Barrs does, but I agree with the sentiment: If I were convinced that Christianity had probability zero, or was far less probable than another religion threatening eternal punishment, I would have no interest in it. To quote Paul, if Christ has not been raised, our preaching is useless and so is your faith. (1 Corinthians 15:14).

Still, it's true that Christianity has spiritual aspects, and I do try to live up to them. It's also emotionally helpful that I can sympathize with, say, Jesus's concern for the poor and outcast, though even there, Jesus made some statements that are tough, such as Mark 14:3-6 -- the other people present said exactly what I would have felt in that situation!

When I mentioned to my wife that I was corresponding with someone who had disavowed atheism because of Pascal's Wager, her response was to say, well, I'm glad it finally convinced someone. And it's true: the Wager is the sort of thing people may encounter in Philosophy 101, and never think of again. Since you find the argument compelling, do you have a theory as to why it isn't more persuasive?

Well, one reason is that it's frankly hard to live up to (I'm certainly not able to do so completely myself). Religion is difficult: Even people who assign 100% probability to the existence of their religion's hell sometimes can't avoid doing things that they think God considers sinful, and it's even easier to make up excuses when you're not certain if the religion is true. I do wonder how many more people would agree with Pascal's wager if it just required, say, pushing a button.

In addition, many people are not aiming to minimize expected suffering. Some value other things, such as personal honesty, more highly than avoiding eternal torment. Still others are genuinely convinced that there's more likely a god who would be angry when people try to appease him for selfish reasons than when they ignore him altogether. This is possible, but I at least don't think it describes the portraits of God in many religions, where my impression is that fear of punishment is a perfectly valid reason for obeying God's commands, even if it may not be as desirable as willing discipleship.

There are a number of other reasons given. I think some do have merit, but they often have the feel of being arguments mined for in order to reach a desired answer, rather than ideas one would naturally come up with in thinking about the situation. Perhaps the best argument is that there are simply better potential sources of massive suffering to be worried about, but I don't understand why one can't take Pascal's wager while continuing to explore those other possibilities. (Admittedly, though, practicing religion does take some time.)

There is a vast literature on the the Wager, and most of it is highly critical. Common objections center on the problem of choosing the correct god, or on the the alleged conceit that God should reward faith that is determined by fearful devotion rather than goodness. I realize this is an immense subject, but can you provide a sense of where and why you believe the usual criticisms fail?

I don't really see what the many gods objection is getting at, other than to point out that religion is a risky business because you could very easily be wrong. Yes, that's an extremely unpleasant state of affairs, but who says can't that be the way things are? The suffering of billions of animals in the wild is awful, but that doesn't make it untrue. We just have to accept it and do the best we can.

Some people talk about the thousands of possible religions as though they were equally good candidates, but even apart from factual-truth considerations, I don't think that's the case. Many religions teach only of a general underworld for the dead, like the early Jewish Sheol or the Greek Hades, where all people end up. A number of others (certain branches of Buddhism, Hinduism, Jainism, Sikhism, Taoism, and Zoroastrianism) do have punitive hells, but they're usually finite and for purgative purposes only. Many of these punishments, moreover, are for general wicked deeds, rather than refusal to follow a particular god. In fact, the only two religions of which I'm aware that really fit into the standard believe or face eternal punishment model are Christianity and Islam. (If readers are aware of others, I'd love to hear about them!)

In your view, people (like me) who assign zero or negligible probability to the possibility of there being a creator-god are biased by overconfidence. Can you explain what this means? And also, do you allow that a strong atheist argument could overcome the bias? Or does epistemological uncertainty imply that overconfidence is insoluble?

The overconfidence effect is the observation that people feel more certain about their beliefs than they should be. For instance, the Wikipedia article gives the example of answers that people rate as '99% certain' turn[ing] out to be wrong 40% of the time.

In my piece on Pascal's wager, I added this qualification:

In view of the last point, I don't think any argument could reduce the probability of Christianity (or any other religion) to zero -- epistemological uncertainty reigns. Of course, there's a point at which even nonzero probabilities have to be ignored due to finite computational resources, but in light of the immense potential consequences, religion for me doesn't fall below that level.

You mention Eliezer Yudkowsky. It seems that he may have had you in mind when he inconclusively sought to outline a strategy by which an Occam-abiding mind could avoid being dominated by tiny probabilities of vast utilities -- a problem he describes as Pascal's Mugging. Yudkowsky focuses on dangers that might arise if a Friendly AI, lacking an intuitive understanding of dominant mainline probabilities that we assign through reason, were to be mathematically pranked into destroying the world (or something) in a scenario where the probabalistic stakes were presented on sufficiently grave and vast computational terms.

Do you have any thoughts on the potential problem that Yudkowsky identifies? And do you think it is relevant to your current reasoning regarding the Wager? When one's epistemic framework allows that supernatural forces may test vast or infinite degrees of suffering, how does one avoid being mugged?

I don't know if one should avoid being mugged. While it's not obvious to me that Solomonoff induction per se is the ideal way to arrive at one's prior probability distribution -- for example, why should there be a fundamental link between universal Turing machines and epistemology? -- I think something like it is the right way to go. (Unfortunately, Solomonoff induction itself is uncomputable, but we can try to approximate it.) If we had an accurate assessment of the probability distribution over possible outcomes by a Solomonoff-style AI, I think we should just shut up, multiply, and do what the calculation says. If that means handing over $5 to the mugger, so be it.

As far as the mugging, Nick Tarleton suggested a similar scenario in which an AI might devote vast resources searching for a magic button to generate vast amounts of happiness (or, of more visceral interest for me, prevent vast amounts of suffering). Carl Shulman agreed, noting that there may be ways to appeal to Dark Lords of the Matrix who, say, live in a universe with infinite computational power or different laws of physics. I would side with the AI here: Sure, these investigations will almost certainly prove fruitless and wasteful, but the potential consequences are too big to get wrong.

You have devoted considerable thought to the prospect of creating laboratory universes, which may or may not be possible. Assuming the idea of making a universe -- or an infinity of universes -- from scratch cannot be dismissed as crazy sci-fi speculation, why should this possibility be a cause for worry among those concerned with the problem of suffering?

If the probability of sentient life emerging in a random universe is greater than zero (and if other technical conditions are met), then creating infinitely many universes would almost surely create an infinite amount of suffering. Of course, the same could be said of happiness, but I think there are reasons to suggest that suffering would dominate. Briefly, the probability of intelligent sentient life emerging on a given planet seems considerably smaller than the probability of merely sentient life (at, say, the level of invertebrates or fish). Since the smallest, most numerous animals tend to breed far more offspring than can survive, most of these animals won't even live to maturity. And those that do will often have short lifespans: For instance, I would rather not exist than be born as a fly, live for a few weeks, get caught in a spider's web, and die by venom poisoning. (And even being the spider could be unpleasant at times.)

You write that suffering far outweighs happiness right now on earth, and you seem confident that a disproportionate burden of this suffering is experienced by animals. This perspective has impelled you to consider the plight of animals not merely in factory farms and vivisection labs -- which is par for the course among prominent utilitarian ethicists -- but in nature, where an average animal's life may be overwhelmingly defined by fear, starvation, and pain. If wild animal suffering is, as you suggest, an open problem, why do you think it is not taken more seriously among philosophers and animal welfare activists? And if natural animal suffering is a legitimate area of moral concern, what is to be done?

That's a good question -- I'd like to know the answer! The issue probably doesn't cross people's minds very much, and even when it does, they may not register it as a problem. I remember as a kid loving to watch videos of predacious dinosaurs chasing, catching, and biting the flesh of their prey; the videos were just exciting, and I didn't give a second thought to how agonizing the Stegosaurus must have found it. Some people who do recognize wild-animal suffering argue that cruelty is part of nature, so that's just the way things are, and we should get used to it. (Of course, cancer is entirely natural as well, and most people don't seem quite so willing just to get used to it.)

The absence of interest on the part of animal welfarists is more surprising. Part of it may result in a desire to avoid sounding too extreme -- meat-eaters sometimes use concern for wild animals as a reductio against concern for animals in general -- but fear of sounding extreme doesn't describe a significant fraction of the animal-welfare movement (unfortunately).

There's also the mentality that ethics begins at home: That humans need to get their own moral house in order before worrying about suffering for which they aren't responsible. In some cases, this may stem from a view of morality as being about keeping your own hands clean from blame, rather than, as I see it, trying to fix problems (focusing on the biggest problems first).

But why the silence on the part of utilitarians? There are some concerned with the problem, though I haven't heard much discussion in mainstream philosophical circles, or among welfare economists. Peter Singer argued:

Another subject that you stand virtually alone in taking seriously is the possibility of insect sentience. Among animal behaviorists and entomologists, there seems to be general consensus that is that this is impossible, or at least profoundly improbable. Why aren't you convinced? And again, if it should turn out that bees and fruit flies have the capacity to suffer, what is to be done?

There are several entomologists who take insect pain seriously. In her excellent review of the subject, Jane A. Smith explained:

What is to be done? Well, on a very small scale, people can make sure that when they kill insects they do so as quickly as possible. Injured insects (say, on the sidewalk) can be squashed to be put out of their misery. But while compassionate, these actions address a minute sliver of the expected insect suffering that exists. I remain open to suggestions for larger-scale action. In addition to humane insecticides, some have suggested ecosystem redesign, though given the immense complexity and fragility of biological systems (as noted before), I'm not convinced of its plausibility.

Your work is broadly informed by transhumanism. But you argue that fans of the singularity are biased by undue optimism. Why?

To a large extent, my skepticism about the Singularity is no different than skepticism about any other hypothesis on which there's widespread disagreement: Namely, that lots of intelligent people are doubtful of the feasibility or broad social acceptance of many of the technologies that transhumanists discuss.

In addition, some transhumanists approach their philosophy in the same way that many approach religion: As a source of optimism for the future (including, perhaps, immortality). There's a natural tendency to want to think that these desirable outcomes will happen. However, I don't want to paint with too broad a brush, because I know many hard-nosed transhumanists who are very deliberate in trying to avoid this. Some even agree that Singuarity scenarios aren't terribly likely -- it's just that, if they did happen, the consequences would be so enormous that the expected value of considering them is still huge.

While I can't bring myself to assign any probability to the reality of a Christian hell, my brush with your -- and Yudkowsky's -- writings leaves me wondering if a secular version might be a live possibility. I'm imagining a kind of sentient AI program that would create -- or simulate, or embody -- real and vast suffering. It just seems that if you take a physicalist view of nature -- and scientific enterprise -- that simulated consciousness has the force of inevitability. And unless human nature changes radically, it seems likely that from the safety of their cybernetic skinner boxes, at least a few sadistic hobbyists won't be able to resist the temptation to create worlds of interminable suffering. What is your thinking? Is the problem of simulated suffering something that merits exploration?

It's not at all obvious to me that computer-simulated consciousness is possible, but I maintain some probability that it is, and in that case, the possibility of torturing simulations or creating simulated hells could be very real. There's also a possibility of extraterrestrial civilizations that explicitly seek to cause pain to other organisms or, more likely (and less anthropomorphically), that are completely indifferent to whatever qualia they happen to create.

As Nick Bostrom has pointed out, if simulation is possible, we ourselves may very well be in one. In view of the vast amounts of suffering in the world, our simulators could very well be of the indifferent type. Perhaps they were trying to model the evolution of the universe, and the events on earth simply happened to be part of that (though unless one subscribes to something like functionalism, it's not obvious that mathematical simulation of the atoms composing a human brain would actually give rise to consciousness). Or perhaps our simulators were aiming to explore different types of minds to see how they react in various environments.

If we are simulations, it would be interesting to ask why the concept of hell is so widespread among a variety of cultures. Is it possible that our simulator introduced the idea to give people a foretaste of what s/he had in mind following some of their deaths? Even for those who can't bring themselves to assign nonzero probability to conventional religions, this possibility alone may introduce a non-negligible probability of eternal torment. The existence of notions about hell is a fact about the world that needs explaining, and this is one hypothesis that does so.

Whether people buy Christianity or not, I do think it's important to give serious thought to the possibility of hell in general. I like this quote from Pascal's Pensees (199-206):

Let us imagine a number of men in chains and all condemned to death, where some are killed each day in the sight of the others, and those who remain see their own fate in that of their fellows and wait their turn, looking at each other sorrowfully and without hope. It is an image of the condition of men.

A man in a dungeon, ignorant whether his sentence be pronounced and having only one hour to learn it, but this hour enough, if he knew that it is pronounced, to obtain its repeal, would act unnaturally in spending that hour, not in ascertaining his sentence, but in playing piquet. [...]

When I consider the short duration of my life, swallowed up in the eternity before and after, the little space which I fill and even can see, engulfed in the infinite immensity of spaces of which I am ignorant and which know me not, I am frightened and am astonished at being here rather than there; for there is no reason why here rather than there, why now rather than then. [...]

The eternal silence of these infinite spaces frightens me.