Theories of Welfare and Welfare Range Estimates

Bob Fischer14 Nov 2022 13:52 UTC
91 points
23 comments9 min readEA link

Key Takeaways

  • Many theories of welfare imply that there are probably differences in animals’ welfare ranges. However, these theories do not agree about the sizes of those differences.

  • The Moral Weight Project assumes that hedonism is true. This post tries to estimate how different our welfare range estimates could be if we were to assume some other theory of welfare.

  • We argue that even if hedonic goods and bads (i.e., pleasures and pains) aren’t all of welfare, they’re a lot of it. So, probably, the choice of a theory of welfare will only have a modest (less than 10x) impact on the differences we estimate between humans’ and nonhumans’ welfare ranges.

Introduction

This is the third post in the Moral Weight Project Sequence. The aim of the sequence is to provide an overview of the research that Rethink Priorities conducted between May 2021 and October 2022 on interspecific cause prioritization. The aim of this post is to suggest a way to quantify the impact of assuming hedonism on welfare range estimates.

Motivations

Theories of welfare disagree about the determinants of welfare. According to hedonism, the determinants of welfare are positively and negatively valenced experiences. According to desire satisfaction theory, the determinants are satisfied and frustrated desires. According to a garden variety objective list theory, the determinants are something like knowledge, developing and maintaining friendships, engaging in meaningful activities, and so on. Now, some animals probably have more intense pains than others; some probably have richer, more complex desires; some are able to acquire more sophisticated knowledge of the world; others can make stronger, more complex relationships with others. If animals systematically vary with respect to their ability to realize the determinants of welfare, then they probably vary in their welfare ranges. That is, some of them can probably realize more positive welfare at a time than others; likewise, some of them can probably realize more negative welfare at a time than others. As a result, animals probably vary with respect to the differences between the best and worst welfare states they can realize. The upshot: many theories of welfare imply that there are probably differences in animals’ welfare ranges.

However, theories of welfare do not obviously agree about the sizes of those differences. Consider a garden variety objective list theory on which the following things contribute positively to welfare: acting autonomously, gaining knowledge, having friends, being in a loving relationship, doing meaningful work, creating valuable institutions, experiencing pleasure, and so on. Now consider a simple version of hedonism (i.e., one that rejects the higher /​ lower pleasure distinction) on which just one thing contributes positively to welfare: experiencing pleasure. Presumably, while many nonhuman animals (henceforth, animals) can experience pleasure, they can’t realize many of the other things that matter according to the objective list theory. Given as much, it’s plausible that if the objective list theory is true, there will be larger differences in welfare ranges between many humans and animals than there will be if hedonism is true.

For practical and theoretical reasons, the Moral Weight Project assumes that hedonism is true. On the practical side, we needed to make some assumptions to make any progress in the time we had available. On the theoretical side, there are powerful arguments for hedonism. Still, those who reject hedonism will rightly wonder about the impact of assuming hedonism. How different would our welfare range estimates be if we were to assume some other theory of welfare?

In the rest of this post, we argue that while assuming hedonism makes a difference to welfare range estimates, it makes a surprisingly small difference—likely less than 10x. Our central claim here is that even if hedonic goods and bads (pleasures and pains) aren’t all of welfare, they’re a lot of it.

Tortured Tim

The task here is to determine the impact of assuming hedonism on welfare range estimates. We don’t need to consider all possible theories of welfare to do this. Instead, we need to identify a theory that’s plausible and can serve as an upper bound. That is, we need to identify a theory of welfare that meets two conditions: first, those concerned with relative welfare range estimates would be willing to assign it a reasonably high credence; second, of the theories meeting the first condition, it supports the largest differences in welfare range estimates.

These conditions are fuzzy; we won’t try to make them precise here. Moreover, we won’t dwell on the challenge of showing that a given theory meets these conditions. Instead, we’ll simply propose a candidate that seems promising and argue from there: namely, the garden variety objective list theory that we mentioned earlier. Given as much, our strategy is to assess the impact that the choice between hedonism and our garden variety objective list theory is likely to have on our welfare range estimates for humans. Then, we can extrapolate to animals.

Round 1: Can Non-Hedonic Goods Outweigh Hedonic Bads?

Our garden variety objective list theory needs to specify how much the various items on the list contribute to an individual’s welfare range. Recall: our objective list theory says that there are many things that are good for individuals, including friendship, romantic love, acquiring knowledge, engaging in theoretical contemplation, doing meaningful work, developing practical skills, and so on. Now imagine Tortured Tim, an individual who is experiencing extraordinarily intense physical suffering. Nevertheless, he may have many other goods in his life: friendship, romantic love, knowledge, practical skills, and so on. However, it seems highly implausible that these goods put Tortured Tim into a net positive welfare state: no matter how valuable they are, they aren’t so valuable as to outweigh the welfare costs of intense physical suffering.

Imagine the objective list theorist denying this. Then, the objective list theorist would be committed to saying that it would be prudentially rational for Tim to choose to extend his life by a day of torture rather than die a day sooner, as long as that wouldn’t affect the other goods in his life. Or, if the arc of a life matters, the objective list theorist would be committed to saying that it would be prudentially rational for Tim to choose days of torture now for an equivalent number of normal days tacked onto the end of his life. It seems obvious, though, that it would not be prudentially rational for Tim to make either choice.

Let’s grant, then, that the non-hedonic goods in Tim’s life don’t put Tim into a net positive welfare state. Let’s suppose further that welfare ranges are symmetrical around the neutral point—that is, individuals can be made as well off as they can be made badly off. It follows that those non-hedonic goods, in the aggregate, increase Tim’s welfare less than 50% of the hedonic portion of his welfare range (his “hedonic capacity,” for short). That is, if the positive contributions of the various non-hedonic items on the objective list don’t jointly outweigh Tim’s suffering—i.e., they leave him in a net negative welfare state—then they aren’t contributing more than half of that portion of Tim’s welfare range that’s grounded in his capacity to experience positive and negative affective states. (See Figure 1.)

Figure 1: Tortured Tim

If that’s right, then we have a data point that’s relevant to our initial question: just based on the above, it looks like the choice of a theory of welfare won’t change our welfare range estimates by more than a factor of 1.5. That is, suppose that if hedonism is true, then chickens have a welfare range of 10 welfare units. It would follow that if our objective list theory is true, their welfare range will be less than 15 welfare units (depending on how well they can realize the relevant objective goods). In other words, when we “tack on” the portion of the welfare range associated with unrealized non-hedonic goods, it doesn’t change Tim’s total welfare range all that much.

Before going any further, let’s pause to note that nothing said thus far entails that interspecific variation in welfare ranges is limited to a factor of 1.5. To see why not, let’s assume that given hedonism, chickens have a welfare range of 10 welfare units, whereas given our objective list theory, they have a welfare range of 12. Now consider the abilities that permit securing various non-hedonic goods. For all we’ve said, those may significantly enhance an individual’s hedonic capacities too. That is, it could work out that the cognitive capacities that permit, say, theoretical contemplation also permit individuals to feel pleasures and pains more intensely than those without such capacities.

However, if hedonic capacities are enhanced by the same capacities that matter for securing various non-hedonic goods, then hedonism might imply that humans have a welfare range of 50, whereas given our objective list theory, they might have a welfare range of 74. So, the choice between hedonism and objective list theory could, in principle, be the difference between saying that humans have five times chickens’ welfare range vs. saying that humans have roughly seven times their welfare range. Still, there’s an intraspecific cap. Suppose that hedonic capacities are enhanced by the same capacities that matter for securing various non-hedonic goods. Then, given some set of empirical facts, the shift from one theory of welfare to another shouldn’t affect our estimate of one species’s welfare range by a factor greater than 1.5.

Tortured Tim, Round 2: Can Hedonic Bads Compromise Non-Hedonic Goods?

All that said, there’s a problem with our estimate. Our description of Tim’s case assumes that Tim can face extraordinary suffering while retaining all the relevant non-hedonic goods. However, it could well be that at least some non-hedonic goods are compromised as suffering increases. For instance, if exercising your autonomy enhances welfare and Tim is suffering to the point that he can’t exercise his autonomy, then he lacks that non-hedonic good.

By way of reply, let’s first note that plausible though this observation may be, it doesn’t seem to apply to every non-hedonic good. You can still have friends while being tortured; you can still have knowledge; indeed, you can actually be doing meaningful work by being tortured, as might be true of martyrs in the right (horrible) circumstances. So, while this observation may show that we’ve underestimated the ceiling on the difference that non-hedonic goods can make to welfare ranges, there are limits to its significance.

The second thing to say is more concessive. Suppose that there are 11 items on the objective list, with hedonic goods being one of them, and suppose that of the 10 remaining, half the items are zeroed out due to Tim’s suffering. Finally, suppose that all 10 of those remaining items make equal contributions to welfare. Originally, we concluded that if all the non-hedonic goods don’t outweigh Tim’s suffering, then they increase Tim’s welfare by less than 50% of his hedonic capacity—that is, his ability to have those goods means that his total welfare range is no more than 1.5 times his hedonic capacity. With these new assumptions, we can conclude that his ability to have all the objective goods no more than doubles his welfare range. (See Figure 2.)

Figure 2: Factoring in Unrealized Non-Hedonic Goods

However, the third thing to say is that we’ve been charitable—and perhaps too charitable—in our estimate of the contribution of non-hedonic goods. For what it’s worth, we don’t think it’s plausible that those goods get Tim anywhere near a net positive welfare state. His net welfare state remains very, very low even if he has a significant number of non-hedonic goods. So, if we suppose that those non-hedonic goods do far less to offset Tim’s suffering, it follows that even if some non-hedonic goods are compromised as suffering increases, non-hedonic goods still make a much smaller contribution to individuals’ welfare ranges than hedonic bads—perhaps not even enough to justify our original, factor-of-1.5 estimate.

Round 3: What about Non-Hedonic Bads?

Unfortunately, there’s another problem with our estimate. Thus far, we’ve been ignoring non-hedonic bads. For instance, Tortured Tim has been tortured, which means that he’s been violated. Perhaps it’s objectively bad to be violated. If so, then we’ve been ignoring factors that can drag Tim’s welfare level down further, which means that his welfare range is larger.

How much larger, of course, depends on the list of objective bads and their relative weights. Let’s assume, however, that the non-hedonic bads mirror the non-hedonic goods—which may not be true, but seems like a friendly concession. That is, let’s assume that if there are 10 non-hedonic goods, then there are 10 non-hedonic bads. Moreover, let’s assume that Tim isn’t realizing every non-hedonic bad simply in virtue of being tortured. Then, factoring in both non-hedonic goods and bads might triple his total welfare range relative to his hedonic capacity. (See Figure 3.)

Figure 3: Factoring in Realized and Unrealized Non-Hedonic Bads

Of course, in saying this, we’re ignoring the point we made at the end of the last section: namely, that we’ve been too charitable in our estimate of the contribution of non-hedonic goods, which suggested that doubling was too high an estimate. If that’s right, then tripling is too high an estimate as well.

Applying Tortured Tim to animals

Still, let’s suppose that tripling is correct. Let’s also suppose that welfare invariabilism is true, according to which the same theory of welfare is true of every welfare subject. So, to illustrate, if our garden variety objective list theory is true of humans, then it’s true of all nonhumans. (This is a friendly assumption: if variabilism is true, we might not get any differences in welfare ranges at all.) Given invariabilism, we can say that if the choice between a simple version of hedonism and an objective list theory wouldn’t affect our estimate of humans’ welfare range by more than 3x, it shouldn’t affect our estimates of nonhumans’ welfare ranges by more than 3x.

Again, this is not the same thing as saying that interspecific variation in welfare ranges is limited to 3x. For all we’ve said here, many nonhumans may have much smaller hedonic capacities than many humans. If chickens, for instance, have just 0.1x humans’ hedonic capacity, then the choice between hedonism and an objective list theory could be the difference between the conclusion that chickens have 0.1x humans’ welfare range (given hedonism) and ~0.03x humans’ welfare range (given our objective list theory). So, the point is about the impact of theories of welfare on welfare range estimates. We aren’t making any claims here about welfare range estimates themselves, nor even trying to bound them.

Conclusion

Prior to investigating this issue, we would have guessed that disagreements about the correct theory of welfare could cause welfare range estimates to differ by orders of magnitude. If we’ve drawn the right lesson from poor Tortured Tim, that simply isn’t true. We suggest that, compared to hedonism, an objective list theory might 3x our estimate of the differences between humans’ and nonhumans’ welfare ranges. But just to be cautious, let’s suppose it’s 10x. While not insignificant, that multiplier makes it far from clear that the choice of a theory of welfare is going to be practically relevant. To see this, recall that Open Philanthropy once estimated that “[if] you value chicken life-years equally to human life-years, this implies that corporate campaigns do about 10,000x as much good per dollar as top [global health] charities.” We’ve argued that the “discount” from switching to an objective list theory will be relatively small. So, while it might matter alongside other ways of “discounting” chicken welfare, it would be surprising if a theory of welfare could alter the outcome of such analyses on its own.

Acknowledgments

This research is a project of Rethink Priorities. It was written by Bob Fischer. Thanks to Jason Schukraft, Adam Shriver, Rachel Norman, Martina Schiestl, Alex Schnell, and Anna Trevarthen for helpful feedback on earlier versions of this post. Thanks to Catarina Kissinger for the visuals. If you’re interested in RP’s work, you can learn more by visiting our research database. For regular updates, please consider subscribing to our newsletter.

Bob Fischer14 Nov 2022 13:52 UTC
91 points
23 comments9 min readEA link