It is reasonable to expect that, if all goes well, astronomically more people will be alive in the future than are alive today. So, the argument goes, ‘the utilitarian imperative “Maximize expected aggregate utility!” can be simplified to the maxim “Minimize existential risk!”’.
But, it was later objected, the fact that the number of future people may be very large does not by itself mean that we should focus on minimizing existential risk. In some situations, it might be more effective to try to make smaller improvements to the expected welfare of future people. Trajectory changes are changes that improve the value of the long-term future through some mechanism other than preventing existential catastrophe. Trajectory changes have to be:
(1) Sticky; to count, their effects must be extremely long lasting.
(2) Not inevitable; bringing something about that would have happened anyway a bit later does not count as a trajectory change.
(3) Morally significant; events that are unimportant cannot be trajectory changes.
Whether trajectory change or existential risk mitigation is more effective obviously depends on the magnitude of existential risk. More fundamentally, it depends on how smooth or jumpy the curve of increase in the expected value of the future is. To the degree that the future is not completely determined yet, variation in human choices will result in variation in the ultimately amount of realized moral value. Good choices will result in more value than bad choices. Different worldviews imply different functions mapping quality of choices to amount of value. For instance, one might think that there are really only two equilbiria in the long-run: extinction and utopia. If this is your view, your function mapping performance to realized value would look something like this:
Given this function, you should probably focus on existential risk reduction. Smaller changes are precluded. Another, I think somewhat less popular, view is that extinction is quite unlikely but that realized value in the future varies significantly with performance:
Finally, you might think that existential risk is high and that the variation in value between different futures without existential catastrophe is large:
If we are in the world described by the second graph or the third graph, it might make sense to pursue trajectory changes in addition to or instead of existential risk reduction. But it can be hard to imagine exactly what kind of changes those would be. It is very easy to see why a nuclear war that killed everyone on earth would curtail humanity’s future. What kind of event might reduce the value of the long run future by, say, 1%? In order to build intuition, I looked into a few examples of morally significant and long lasting historical changes.
Historically Attested Trajectory Changes
The Caste System
In India, nearly everyone belongs to a traditionally endogamous group that historically occupied a specialized economic niche. Genetic evidence shows that caste endogamy is thousands of years old. The rate of intermarriage between at least some Indian sub-castes and their neighbors in the last several millennia must have been less than one percent:
People tend to think of India, with its more than 1.3 billion people, as having a tremendously large population, and indeed many Indians as well as foreigners see it this way. But genetically, this is an incorrect way to view the situation. The Han Chinese are truly a large population. They have been mixing freely for thousands of years. In contrast, there are few if any Indian groups that are demographically very large, and the degree of genetic differentiation among Indian jati [sub-caste] groups living side by side in the same village is typically two to three times higher than the genetic differentiation between northern and southern Europeans. The truth is that India is composed of a large number of small populations.
David Reich, Who We are and How We Got Here, 145-146.
In David Reich’s analysis, fully one third of studied groups were as endogamous as or more endogamous than Ashkenazi Jews.
Textual evidence shows that hierarchical ideas of caste are also thousands of years old. In the Rig Veda, a collection of hymns composed some time in the second millennium B.C., there is a hymn in which a god, Purusha, is sacrificed and his body divided to form the basis of the castes. Purusha’s mouth formed the priestly caste, his arms the warrior caste, his legs the farmer caste, and his feet the laborer caste. Later ancient texts, like the Arthashastra and the Manusmirti, prescribe laws and policies for maintaining caste hierarchy.
Finally, Reich presents genetic evidence that the highest caste, the Brahmins, are disproportionately descended from Steppe people who conquered the Indian subcontinent in ancient times.
Does the above evidence prove that bad treatment of lower caste people in India dates back to ancient times? Well, it is hard to be sure because the Indian climate makes it very difficult for ancient texts to survive. However, it is at least very suggestive. Lower caste people have, in well-documented recent history, been relegated to lives of poverty and illiteracy. They also have been treated utterly without respect. It seems likely to me that people have diminishing marginal utility in status. That is, the gain in going from an extremely low and despised position to an average position is greater than the gain in moving from an average position to an extremely high position. If this is true, the caste system is negative sum in welfare terms—the misery of the low castes is not compensated in aggregate by the bliss of the high castes. All societies, even lots of groups of non-human animals, are hierarchical to some degree. While a moderate amount of inequality may not have significant welfare costs, the costs could be extreme in an extremely hierarchical society. And, at least in recent history, it is hard to think of more extreme examples than the caste system.
For the caste system to represent a historical trajectory change it has to be long-lasting, avoidable, and important. The hardest of these criteria to establish is that the caste system was avoidable. It seems unlikely that the caste system is a necessary result of military-economic competition. Talent for various jobs is unlikely to be perfectly correlated with caste (why would the correlation be perfect?). That means that, inevitably, there will be inefficiency when people perform the labor appropriate for their caste rather than the labor appropriate for their skills.
But might the caste system have been a necessary result of the ancient Indian political situation? The Indo-European invaders seem to have established weaker systems of endogamous classes or castes in ancient Persia, Rome, and Greece. But none of those countries has anything comparable to Indian caste. So caste systems stable on the scale of several millennia were not a universal result of the Indo-European conquests. The arrival of Islam in Persia is sometimes associated with the end of the ancient Persian caste system; perhaps the fact that Islam only became firmly established in India five hundred years after the first caliphate’s conquest of Persia allowed caste to entrench itself in India. It is also possible that something about the ancient Indian political situation made the persistence of caste almost inevitable. Either way, after the caste system became firmly established, it has proved very hard to dislodge.
Infanticide and Abortion
The Greeks, Romans, and Pre-Islamic Arabs all practiced widespread infanticide. I think the extent to which Christianity and Islam actually ended infanticide in these places, as opposed to just pushing it out of the literary sources, is not totally clear. However, there is good data on infanticide and abortion from Early Modern Japan. Ordinarily, in pre-industrial societies fertility rates were high, and population growth was slowed by high natural infant mortality. In Tokugawa Japan, fertility rates fell long before industrialization and rose again early in the industrial period, once infanticide was brought under control:
In Early Modern Japan:
Infanticide permitted a range of interpretations. Administrators worried about dwindling populations and falling revenues, and often thought that it was a love of luxury that prompted people to kill their children. Villagers complained that poverty left them no other resort, and sometimes helpfully suggested that lower taxes would do wonders for the safety of their newborns. Men of learning often believed that moral education could convince villagers to give up infanticide, but some thinkers argued that it would take a fundamental reform of the political system to achieve that goal. Men of substance who were content to work within the established order, meanwhile, reinvented themselves as moral leaders of their communities and wrote to their governments with offers to finance the eradication of infanticide. Most domains in Eastern Japan built expensive systems of welfare and surveillance. By 1850, the majority of women north and east of Edo were obliged to report their pregnancies to the authorities, and the majority of the poor could apply for subsidies to rear their children. Over the same years, a demographic revolution was set in motion. In the eighteenth century, the consensus of many villages in Eastern Japan was that parents could, and under many circumstances should, kill some of their newborns. Perhaps every third life ended in an infanticide, and the people of Eastern Japan brought up so few children that each generation was smaller than the one that went before it. By 1850, in contrast, a typical couple in the same region raised four or five children, and a long period of population growth began. By the 1920s, the average woman brought six children into the world, and in Eastern Japan, as elsewhere in the nation, overpopulation at home became an argument for expansion abroad. Eastern Japan, in other words, had experienced a reverse fertility transition.
Fabian Drixler, Mabiki, 2.
One reviewer of Mabiki acknowledged the reality of the pattern but attributed the shortfall in births to abortion rather than infanticide. Either way, the population was reduced and per capita standards of living would have been raised (assuming, what is almost certainly true, that Early Modern Japan was a Malthusian economy). In addition to the effects of infanticide on population size and average well-being, it may also have had negative psychological consequences for parents. Speaking personally, the deaths that seemed to weigh by far most heavily on older members of my family are the (unintended) deaths of children, even though those deaths all happened seventy or more years ago. Those deaths were not products of infanticide, and for all I know, the pain associated with infanticide might have been less. On the other hand, I think murder is often more troubling to the victim’s family than other causes of death. And in Roman law, the paterfamilias could decide unilaterally, without the mother’s permission, whether to accept a new baby into the family. I imagine the pain of the mothers in those situations must have been enormous. Finally, if we are going to establish moral side-constraints against anything, we probably should start with infanticide.
Human Sacrifice and Gladiatorial Combat
Rodney Stark is a Christian historical sociologist. He is not at all a neutral observer. But I think his discussion of what gladiatorial games reveal about pagan society appropriately describes the stakes:
But, perhaps above all else, Christianity brought a new conception of humanity to a world saturated with capricious cruelty and the vicarious love of death. Consider the account of the martyrdom of Perpetua. Here we learn the details of the long ordeal and gruesome death suffered by this tiny band of resolute Christians as they were attacked by wild beast in front of a delighted crowd assembled in the arena. But we also learn that had the Christians all given in to the demand to sacrifice to the emperor, and thereby been spared, someone else would have been thrown to the animals. After all, these were games held in honor of the birthday of the emperor’s young son. And whenever there were games, people had to die. Dozens of them, sometimes hundreds. Unlike the gladiators who were often paid volunteers, those thrown to the wild animals were frequently condemned criminals, of whom it might be argued that they had earned their fates. But the issue here is not capital punishment, not even very cruel forms of capital punishment. The issue is spectacle for the throngs in the stadia, watching people torn and devoured by beasts or killed in armed combat was the ultimate spectator sport, worthy of a boy’s birthday treat. It is difficult to comprehend the emotional life of such people. In any event, Christians condemned both the cruelties and the spectators. Thou shalt not kill , as Tertullian (De Spectaculis) reminded his readers. And, as they gained ascendancy, Christians prohibited such “games.” More important, Christians effectively promulgated a moral vision utterly incompatible with the casual cruelty of pagan custom.
Rodney Stark, The Rise of Christianity, 214-215.
Fourteen centuries of Christian rule provided ample cruelties of their own. Christian era executions, some of which took especially dramatic forms like breaking on the wheel, often served as public spectacles. And pagan Rome wasn’t necessarily all bad. While the Carthaginians famously burned children alive as offerings to their god Moloch, the Romans: “did not tolerate human sacrifice among the peoples they conquered[...] seriously curtailing the practice (if not actually eliminating it) among the Carthaginians and among the Celts.” Practices such as human sacrifice and gladiatorial combat were long-lasting side-constraint violations, even if that the Malthusian model implies the welfare effects of each individual game in the Colosseum or human sacrifice were transient.
Vegetarianism
Today, vegetarianism and other diets that limit meat intake are more common in India than almost anywhere else. Moral vegetarianism in India dates back to ancient times. My understanding, from a professor I had as an undergrad, is that early Jains were the first to become concerned about animal welfare and then this concern spread to Buddhists and Hindus. While many religions involve fasts from meat for spiritual purposes, I think Indian civilization is pretty distinctive in valuing animals intrinsically. This seems likely to be a contingent fact of India’s cultural history rather than an inevitable adaptation to the circumstances. Further, the animal welfare consequences of Indian vegetarianism over the broad sweep of history are likely to have been large. There also may be human population size consequences of foregoing meat.
Alcohol Prohibition
In the modern world alcohol might be a net benefit to humanity. Many people enjoy it, but many others become addicted to it. So it is hard (if not impossible) to say whether it is good or bad. In a Malthusian situation, the effects of alcohol seem more obviously negative. Either it is a form of luxury spending that can be replaced with other luxuries when the system is out of equilibrium. Or it is consumed at equilibrium as a substitute for subsistence goods. Who would trade subsistence for alcohol? Alcoholics.
Orthodox forms of Islam prohibit the consumption of alcohol. And this prohibition seems to be effective enough, at least now, that it can be read off of a map:
Average alcohol consumption is, as a general matter, higher at extreme latitudes than close to the equator. While pre-existing variation may have something to do with these patterns, I find it very hard to believe that it explains everything–the differences are just too extreme. Thus the prohibition of alcohol was morally significant and has been long lasting. The content of religious laws also seems very contingent. If Muslims had won the Battle of Tours or lost the Battle of Talas, the contemporary borders of the Islamic world (and therefore also the region of minimal alcohol consumption) might be very different.
Moral Change within the Malthusian Trap
Some readers might have noticed that none of the suggested historically attested trajectory changes involve changes in per capita living standards. That is because nearly all economies prior to the Industrial Revolution were governed by Malthusian dynamics. In a Malthusian economy, growth in technology or the capture of new natural resources results in only a transient improvement in per capita living standards. This is because the population always grows to eat up the new surplus. Once the surplus is eaten up, living standards decline again. How far do they decline? Back to subsistence; that is, back to the point at which they could not decline any further without causing the population to fall. This is why per capita incomes did not grow in a sustained way between the rise of agriculture and the Industrial Revolution.
Unless the jaws of the Malthusian trap are broken, there is no way that changes in per capita economic living standards can be made to stick. But there is more to life than per capita economic living standards. Some changes along other dimensions have been significant and long lasting. That’s why I want to push back on the tendency I sometimes see in online discussions of macrohistory to assume that the only genuinely “macro” historical events are the invention of agriculture and the Industrial Revolution.
The examples I gave of historically attested trajectory changes fit into the categories of:
Changes in Population Size
Malthus allowed that, if fertility could be controlled artificially, the increase of population might not consume increases in productivity. Thus infanticide and abortion can change both population size and equilibrium economic living standards. The history of infanticide and abortion will therefore have different assessed consequences according to different population views of population ethics (apart from their inherent moral significance).
Non-Economic Welfare Changes
Everyone knows that it is possible to be poor and happy or rich and unhappy. This idea seems less relevant to the deep past because the level of poverty that nearly everyone was subjected to was so extreme. Similar levels of poverty are seen today in rich countries not among the average poor but only among the very poorest. It is hard to imagine how anyone could be happy while starving or freezing. But it is easy to imagine how someone could be made more miserable. Extreme disrespect or non-disabling physical torture could make the life of even someone living at subsistence harder, without killing him by reducing his income.
Animal Welfare
Human economies were Malthusian with respect to the human population. But they also had associated populations of domesticated animals, in varied living conditions. Those animal populations did not follow Malthusian laws because people consciously regulate the size of animal herds. Thus changes to human beliefs or practices related to animals can have long lasting welfare consequences, even in a Malthusian situation.
Violation of Side-Constraints
The short-run welfare effects of many atrocities (murders, wars, violent rampages) are obviously negative in any situation. In the medium term, in the Malthusian trap, the consequences become debatable. A war that killed off 10% of the population would reduce the intensity of cultivation and might allow farmers an easier life until the system returns to equilibrium. Or maybe the acute suffering and destruction of physical capital outweigh this effect. Either way, in a Malthusian situation, the welfare effects of deadly violence will be transient. Eventually, the population will return to equilibrium, regardless of whether there was more or less suffering in the meanwhile. However, some moral theories hold that certain actions are wrong apart from their welfare consequences. And even if we are committed consequentialists, we should still not be certain that our preferred moral theory is right. So we might regard events as historical trajectory changes if they established long-lasting practices that violate the rules of deontological or virtue theories.
Addiction
Normally, people prioritize survival for themselves and their children over all other goals. There is, however, a big exception. When people are addicted to a drug, they often prioritize access to the drug over access to goods needed for survival. Because, unlike most goods, demand for an addictive drug can compete with demand for food, the spread of an addiction can reduce the maximum population that can subsist at a given level of total wealth. Also, addictive drugs introduce new sources of non-economic suffering.
The Steady State of the Future
Because of the expansion of the universe, there is only a finite amount of matter and energy that is in principle accessible from earth. The maximum amount of possible economic value per atom may be finite or it may be infinite. If it is finite, economic growth due to technological change will at some point cease. All the important, possible technology will have been invented already. If this “Technological Completion Conjecture” is correct, trajectory changes will have to act on some other mechanism than increasing the total wealth of the civilization of the future. Future civilizations would be analogous to past civilizations in that the only way they could increase per capita wealth would be population decline. The economic steady state of the future may or may not be Malthusian. Population could be artificially capped at some level above subsistence. But population and individual living standards will ultimately be capped, naturally or artificially. Thus changes to the moral trajectory of future civilization might have to take similar forms to changes to the trajectories of pre-industrial civilizations. My typology of trajectory changes possible given a fixed level of output is very likely to be incomplete. There are probably many other morally significant kinds of changes that can occur in the absence of changes to per capita income (most obviously, changes in the distribution of income might change welfare levels without changing total economic output).
What specific aspects of the trajectory of future civilization might it be important to change? I have a few ideas, but they are very speculative. If output is again capped and Malthusian dynamics return, population axiology may begin to seem a lot less dry and academic. Factory farming might either disappear or radically expand. Digital minds might be created, and treated either humanely or horrifically. Horrific treatment could be motivated by efficiency (as it is in factory farming) or by more perverse motivations (as in human sacrifice or gladiatorial games). And new addictive drugs (and counter-measures to addiction) are very likely to be invented.
Thanks to Applied Divinity Studies, Matthew Barnett, Skluug, Kenan, and Voxette for comments and discussion.
Just wanted to comment that this was a really thoughtful and enjoyable post. I learned a lot.
In particular, I loved the way the point about how the relative value of trajctory change should depend on the smoothness of your probability distribution over the value of the long-run future.
I’m also now curious to know more about the contingency of the caste system in India. My (original) impression was that the formation of the caste system was somewhat gradual and not especially contingent.
Thank you!
It was unclear to me whether by “existential risk” you were primarily only thinking about extinction risk, vs. including non-existential extinction risks as well.
For example, the leadup to your first diagram says:
Whereas the leadup to your third diagram says:
Personally, my actual best guess now is that there are many possible non-extinction, non-utopia equilibria, but almost all of them have ~0 value compared to the best possible future. I explore this briefly here. So the shape of my beliefs are closer to your first graph (albeit sharper), but (depending on your definition of existential risk), the verbal description of what I would introduce my belief as would fit closer to your introduction of the third graph.
As an intuition pump, imagine if our descendants aimed pretty hard towards making great lives for all sentient beings, but decided (wrongly) to be biological instead of digital. Even if the per-being utility is the same as their (optimal) digital alternatives, you could easily lose 3 OOMs of efficiency just by forcing a constraint of physical substrate. And this is just one possible disappointing future, and there are many other traps like that. The traps are disjunctive and any one of them can make the value of the future be ~0.0% the value of the best possible future.
I meant broad sense existential risk, not just extinction. The first graph is supposed to represent a specific worldview where the relevant form of existential risk is extinction, and extinction is reasonably likely. In particular I had Eliezer Yudkowsky’s views about AI in mind. (But I decided to draw a graph with the transition around 50% rather than his 99% or so, because I thought it would be clearer.) One could certainly draw many more graphs, or change the descriptions of the existing graphs, without representing everyone’s thoughts on the function mapping percentile performance to total realized value.
Thanks for explaining how you think about this issue, I will have to consider that more. My first thought is that I’m not utilitarian enough to say that a universe full of happy biological beings is ~0.0% as good as if they were digital, even conditioning on being biological being the wrong decision. But maybe I would agree on other possible disjunctive traps.