Tuesday, April 6, 2021

Book Review: The Feeling of Value by Sharon Hewitt Rawlette

See also this review's preview and postview.


The Feeling of Value, by Sharon Hewitt Rawlette, is a book proposing a moral realism, in which goodness and badness are instantiated in qualia of goodness and badness, respectively.

Her book is well-written -- maybe better than the average analytic philosophy book. Her topic is one that is important. People aren't always very motivated by reason (perhaps we are something like "reason resistant" or "reason anti-realist" when it departs from pragmatism on a personal level). But young people still expect the world to make sense, and a culture steeped in moral realism might behave more morally per person than the status quo. Certainly there's something about the spirit of realism, moral or otherwise, which is refreshing.

Arguably, we are all doing our best anyway, or it could be that causing everybody to want to do the right thing all at once will produce diminishing returns, so instead, people should just carry on as they are, and cultural change is not called for. I have tended to resist this idea. I think we can motivate at least a significant marginal change in culture through some kind of moral realism. While I ultimately find Rawlette's realism to be incomplete, I appreciate the spirit in which she may have pursued it, assuming that in some way she was hoping to make the world better through philosophy.

If I had no idea of such a thing as transformative AI, I might see a great field for cultural transformation to produce whatever good we're going to have in the world. There may still be something to this, even if there is such a thing as AI that may transform "everything". Maybe transformative AI will take longer to arrive than expected. Maybe it will never arrive, as much as it may seem to be on the way. It may be that AI will not be as transformative as we think. (For instance, its "hands" might be tied by the human political structures that govern them.)

Rawlette doesn't mention this, but one value of moral realism is the ability to say "Here is history-invariant right and wrong, so that civilization has a firm constitution, keeping it from drifting in bad directions." A futurist altruist might try their best to make sure that there is a good future in 2,000 years, or whatever time frame. If it seems that people could voluntarily drift their culture toward anti-natalism, or if humans become increasingly useless and are okay with that, choose uselessness, so that humans no longer need to exist for civilization's purposes (maybe AI decide that is the case), then such an altruist might want to establish a history-invariant constitution for human (or human-AI) culture which preserves human existence, just as they would take any other precaution to ensure that humans don't die out in 2,000 years. Also a good moral realism may help tell us what true human (or animal) well-being really is, which is helpful if the alignment of transformative AI is effectively the constitution of the future. We would want AI to be implementing truly good values, if we could manage to keep it aligned to whatever it is we want it to be aligned to.

I have written in more detail in my notes on The Feeling of Value (notes on first pass, notes on second pass). Also, I provide a somewhat incomplete summary of her book at the end of this review, which may help the reader of this review decide whether to read her book.

Here are a few thoughts overall.

I wondered a lot whether Rawlette adequately connected her normative qualia with pain and pleasure. So for her, goodness is instantiated in qualia of "ought-to-be-ness". It is not obvious to me that that is exactly the same as pleasure, as we ordinarily think of the term. Likewise with qualia of "ought-not-to-be-ness" and pain. (I listened to her 2020 interview with Gus Docker and it sounds like she intends "pleasure" to include any feeling that feels to us like it ought to be. So her underlying idea does not need to be connected with the word "pleasure", necessarily.)

Her book relies heavily on phenomenology. While phenomenology is a (relatively) basic element of philosophy, maybe a future edition could explain how phenomenology is supposed to work, insofar as it provides part of the foundation of her argument.

I thought about it a bit and think she has a case for normative qualia themselves being good and bad. But I think that value can be broader than that. So her theory covers a subset of value.

I don't agree that goodness and badness are inherent in those qualia, though, just that it's pretty hard to not see them as always being good and bad, respectively. That which makes them good and bad is not them themselves.

It's interesting reading her book given my readings on effective altruism. The effective altruists are one of the perhaps few groups of people who try to influence their action in the world with academic philosophy.

Writing a book like this in 2021 (as opposed to 2008, when it was), an author might engage more with futuristic scenarios, be less inclined to say "Weird things aren't going to happen in practice."


(Headings in italics are Rawlette's, from the book.)

Feeling of Value Summary, ch 1 - 3



Brief overview of the whole book.

Chaper 1:

pp. 7 - 10

People doubt whether moral realism is worth thinking about. Rawlette says she will show how the extra effort of thinking about it helps us to overcome some of our biases and motivate us more.

The realism/anti-realism distinction

pp. 10 - 15

Rawlette gives a definition for realism vs. anti-realism in normative affairs. She discusses constructivism and projectivism as actually being anti-realist. She explains why she prefers "judgment-independence" to "mind-independence" as criterion for realism. (When you experience pain or pleasure, that very experience in your mind grounds right and wrong on her view.)

Arguments that belief in antirealism shouldn't affect our first-order normative judgments

pp. 15 - 18

Rawlette discusses some of the views of anti-realists, which claim that coming to be an anti-realist does not affect what you care about. She asserts that it does, because you lose the sense that you need to conform your moral judgments to what it judgment-independent.

Perspectival bias

pp. 19 - 22

Rawlette discusses perspectival bias ("human beings, as we actually are, are not equally motivated to attend to the interests of all other human beings"). Then considers realism vs. anti-realism with respect to how their metaethical views ought to affect their perspectival bias.

Motivation to remove perspectival bias

pp. 22 - 25

Rawlette tries to show how judgment-independence itself causes us to work against our perspectival biases. We see that we experience the value of pleasure / disvalue of pain, and think, if that is true for me, it must be true for anyone else's pleasure / pain.


pp. 25 - 26

Summary of Chapter 1


Chapter 2:

pp. 27 - 28

Brief overview of chapter 2. There are realisms that Rawlette rejects, and her realism is built as a response to their flaws.

pp. 28 - 40 Intuitionism

Rawlette criticizes intuitionism, on the grounds that people's moral intuitions can contradict. Therefore what seems self-evident (or in her terms, "self-justifying") can't be trusted in.

pp. 40 - 48 Minimal realism

Rawlette rejects minimal realism (the view that metaphysical questions do not apply to ethics because there is nothing "out there" in the world that informs them) because it implies that moral facts have no connection to moral judgments. If we have moral judgments, they are not evidence of moral facts. But if this is so, then we are effectively anti-realists, so this position is not effectively a form of realism.

pp. 48 - 58 Ideal-observer theory

Rawlette rejects ideal-observer theories, which say that a person can find judgment-independent moral truth by imagining their idealized, more or less omniscient self's desires, because these theories depend on desires, which can be arbitrary and contingent, and are thus really anti-realisms.

pp. 58 - 63 Synthetic naturalism

Rawlette rejects synthetic naturalism as a route to moral realism because it does not offer an account of what goodness is. Synthetic naturalism says that certain physical properties are the same as what is good, perhaps ones which causally regulate our saying "good" in a homeostatic way. (I don't fully understand this section, but this is my understanding.) Unfortunately, there are different ways that we could have evolved to say "good", perhaps due to being on different planets or in different cultures. So we aren't talking about the same thing when we say 'good', talking to someone from a different background. So this can't be the foundation of judgment-independent moral realism. We haven't figured out what "good" means in a way that justifies it against some kind of disagreement.

(I think I would understand this section better if I knew what was meant by "causal regulation". A quick search didn't reveal a definition, and I didn't feel like digging to find one.)

pp. 64 - 66 Criteria for a plausible realism

Rawlette gives four criteria for a plausible realism:

1) "An account of our concept of goodness". It may have to be a circular definition, in the end, or a description rather than a definition. Have some idea of what this "goodness" idea is.

2) "An explanation of the way in which things in the world objectively satisfy our concept of goodness." There has to be some connection with the way things are (as far as we can perceive them) and our judgment that they are good.

3) "An explanation of the way in which we can come to know which things objectively satisfy our concept of goodness." We have to be able to know the reasons why we connect "good" to a thing or state of affairs.

4) "An explanation of why we are often mistaken about what is good." Given that we have some way to say "this is good", or for any thing, "this is good", and have reason to know it, why do people disagree on moral matters? Why is our view both true and not obvious, or not universally adopted?

Rawlette points out that the attempted realisms given earlier in the chapter do not pass her test, then mentions analytic descriptivism as a promising candidate, to consider in the next chapter.

Ch. 3

pp. 69 - 71

Rawlette introduces the subsequent chapter. Her argument will connect the feeling of goodness or badness to good and bad. In this way, there is something outside human judgment -- but still accessible to the human mind -- which can be the foundation of valuing and disvaluing.

pp. 72 - 75 What are normative qualia?

Normative qualia are the very qualia of goodness or badness, in their purest, most basic form. Rawlette wants to claim that the feeling of goodness is goodness itself, and the feeling of badness is badness itself.

pp. 75 - 79 The experience of pain

Pain can be split into nociception and the feeling of badness. This feeling is not a judgment, but rather the perception of a quale of badness.

p. 79 - 86 Phenomenology or behavioral disposition?

Rawlette considers whether pain is simply nociception married to a disposition of aversion, apart from any phenomenology of badness, and says no, because we have empirical reasons to think that aversive behaviors and thoughts, and phenomenology, are not always connected. She points out that for some reason, aversive and attractive phenomenology affect decision-making, and so probably have an essential connection to reality, in themselves. We consider ourselves to have ethical responsibilities toward animals if we believe they feel pain, rather than merely exhibit aversive behavior.

pp. 87 - 94

The qualia of goodness and badness can be found in pleasures of all sorts, and pains of all sorts, in all pleasant and unpleasant experiences, as diverse as they may be (Rawlette quotes someone contrasting the pleasures of "a rock-and-roll band, or the taste of a great wine" (p. 92)). They are each one quale, or each a dimension of qualia (Rawlette doesn't think the distinction matters for her argument), distinct from whatever qualia make up an individual pleasant or unpleasant thing.

pp. 95 - 100 Objective normativity and non-intentionality

Rawlette points out that normativity does not have to be intentional, that it can apply directly to the normative qualia themselves, and not have to apply to the objects that occasion the normative qualia.

pp. 100 - 101 Conclusion

Summary of the chapter.


Ch. 4 intro pp. 103 - 106

Rawlette introduces the chapter, saying that she wants to come up with descriptive content for her normativity.

pp. 106 - 117 The Open Question Argument

The Open Question Argument (of Moore) says that if you say "What is good is what is pleasant" then if you ask "is pleasure good?" it's like asking "Are pleasant things pleasant?" But asking "is pleasure good?" makes sense. So then, what is good? Rawlette thinks that the way out of this is to say that goodness itself is a quale -- locate it in something outside the world of human judgment. Then she considers and rejects objections to this thought, particularly those against pleasure being intrinsically good, or pain intrinsically bad.

pp. 117 - 124 The advantages of analytic descriptivism

Rawlette evaluates her analytic descriptivism by the four criteria for a robust realism given in Chapter 2. She thinks that it passes the tests.

Ch. 5

pp. 125 - 127 intro

Rawlette wants to go from pro tanto goodness / badness to goodness / badness all-things-considered. She thinks that the way to do this is to add up the value from each quale of goodness occasioned by an action and then subtract the value from each quale of badness.

pp. 127 - 139 Are others' qualia normative for me?

Rawlette says that others' qualia are normative for one, because they "present as having agent-neutral value" and "they do not contain any reference to any particular action or agent."

pp. 128 - 131 (subsection) 1. Positive arguments

Rawlette says that when we experience pain, we do not think it is only for us to avoid it, that gives no one else a reason to help us. Similarly with pleasure. She claims that normative qualia are "ought-to-be-ness" or "ought-not-to-be-ness" and do not prescribe in themselves a way to alleviate or promote them, nor on whom it is incumbent to act to alleviate or promote them.

pp. 131 - 134 (subsection) 2. Objections

Rawlette considers and rejects reasons internalism as an objection to her theory.

pp. 134 - 139

Rawlette uses Parfit's Reasons and Persons to support an argument that says that subjects do not really exist, at least not absolutely enough to say that a quale could only be what it is for one subject.

p. 139 (subsection) 4. Conclusion

Rawlette gives a practical reason for ignoring most normative qualia (we are more effective in doing the good that we can if we do so).

pp. 140 - 142 Are goodness and badness additive?

Rawlette says normative qualia are additive because they are concrete, and thus add up like other concrete things.

pp. 142 - 143 (subsection) 1. Parfit's alternative proposal

Rawlette discusses Derek Parfit's rejection of the Impersonal Total Principle ("If other things are equal, the best outcome is the one in which there would be the greatest quantity of whatever makes life worth living.")

pp. 143 - 145 (subsection) 2. Against Parfit's proposal

Rawlette discusses "intrinsic good", highlighting its difference from instrumental and extrinsic good. Also from goods whose value depend on the existence of a separate moral fact.

She claims that purely intrinsic goods can't have diminishing marginal utility because there would have to be some higher good that could measure them as diminishing. (My understanding: So if something is of diminishing marginal value, we normally say it's worth less relative to something else. If there were only one good thing to be maximized, there could be only one kind of marginal value, one to one with each unit of "good thing" added.)

pp. 145 - 149 (subsection) 3. Objections to additivity

Rawlette considers and rejects the possibility of there really being multiple intrinsic values (anything in addition to normative qualia).

She considers and rejects Parfit's claim that happiness can be of diminishing marginal value to a world but not to an individual.

pp. 149 - 153 (subsection) 5. [sic] Concluding argument

Rawlette claims that it is the intrinsic qualitative nature of normative qualia which give a reason to act. Therefore we can't prefer average utilitarianism (we would claim that each unit of happiness in the "Repugnant" world is worth less than in the "nice" world), nor can we endorse concerns of equality that contradict total utilitarianism. She considers and rejects the possibility of the Repugnant Conclusion being something we could practically implement. She claims that we find "Repugnant" worlds distasteful because we feel sympathy for the individuals who would live in them, but this does not affect the math adding up normative qualia to form an overall quantity of good or bad.

pp. 153 - 158 Practical questions

Rawlette offers some solutions to problems in measuring and comparing normative qualia.

pp. 158 - 159 Conclusion

Rawlette summarizes the chapter.

pp. 163 - 165 Introduction to Part III

Rawlette introduces chapters 6 and 7.

Ch. 6

pp. 167 - 171 intro

Chapter six is introduced. Rawlette will attempt to show that hedonic utilitarianism is not in conflict with our moral intuition, in the cases that we actually confront (as opposed to hypothetical ones used in thought experiments).

pp. 171 - 173 Act utilitarianism versus rule utilitarianism

Rawlette affirms act utilitarianism and rejects rule utilitarianism.

pp. 173 - 176 Preliminaries about the utility-maximizing decision procedure

Rawlette considers what she calls "Straightforward Utilitarian Decision Procedure" (know everything, add up all possible consequences to all possible actions, choose the utility-maximizing one) and says that this is impractical, but that there can be other decision procedures which are simpler and which do maximize utility, which recommend exactly what the Straightforward UDP would.

pp. 176 - 178 First general feature of actual situations: Uncertainty

Because we don't have perfect information, we have to guess what is the best way to go, based on previous experience. Rawlette cites Gerd Gigerenzer's Gut Feelings: the Intelligence of the Unconscious as a source of a decision procedure that does not require having a lot of evidence / past experience (consider one relevant variable for which you have a lot of data, because adding variables shrinks your sample size, making you more vulnerable to random effects).

pp. 178 - 184 Second general feature of actual situations: Need for coordination

Promises are a good way to help coordinate behavior. But Straightforward UDP does not allow promises to be formed. Rawlette offers a modification of Straightforward UDP which does allow promises to be kept (in cases where you're unsure if keeping a promise is utilitous, err on the side of keeping it). Deviating from Straightforward UDP can further utility (can fulfill Straightforward UDP's goal).

pp. 184 - 189 Third general feature of actual situations: Motivational limitations

People tend to find it hard to motivate themselves to increase utility unless they (or those they care about) can reap the benefits of their work toward it. So it's best to let people have some kind of property (literal or figurative) which is protected from interference from others (thus, "rights"). People will tend to let their selfish bias override the more utilitous, other-favoring decisions, unless they have principles ingrained into them beforehand, of how to act in certain circumstances. What this (and the previous two sections) suggest is that we have rules and rights which we only violate (according to the dictates of something like Straightforward UDP) in exceptional circumstances -- which Rawlette points out is the way ordinary morality seems to work.

pp. 189 - 194 Applying this decision procedure to Transplant cases

Rawlette considers thought experiments in which it is supposed beneficial to sacrifice one healthy person to be able to transplant their organs in multiple people in need of a transplant, and finds it not recommended by her decision procedure.

pp. 195 - 197 The probability of destroying useful expectations

Rawlette continues her discussion of Transplant cases, specifically whether knowledge of one disturbing transplant can significantly damage people's expectations.

pp. 197 - 200 Conclusion

Rawlette discusses other anti-utilitarian thought experiments besides Transplant cases, still concluding that utilitarianism is viable, for reasons given earlier in the chapter.

Ch. 7

pp. 201 - 204 intro

Rawlette discusses the fact that there is a history of discussing and also rejecting hedonism, and identifies Robert Nozick's Experience Machine thought experiment as the prime obstacle to people accepting hedonism.

pp. 204 - 208 Pleasure and pain as indicators

Pleasure and pain are indicators of reality. In the long run, we experience more pleasure and less pain if we respond to reality.

pp. 208 - 213 Why the experience machine is a bad idea

Rawlette gives hedonistic reasons to reject the Experience Machine. It is impractical and dangerous to cut yourself off from reality. If it is safe, we would have to have very good computers take care of us.

Also, if we found ourselves in a 100% safe world for using the Machine, there would be no point to being in touch with reality, so we would be so bored we might as well use it.

pp. 214 - 216 Our valuing something vs. its having objective value

Rawlette argues that perceptions of value or disvalue, as applied to objects, are not as reliable as the perception that the normative qualia themselves (as perceptions) are valuable or disvaluable.

pp. 216 - 220 Why taking something to be intrinsically valuable could be instrumentally valuable

Rawlette gives reasons why it could be useful from a hedonistic point of view to see things as having intrinsic value.

Feeling of Value, Ch. 7 con't

pp. 220 - 224 Human pleasure vs. animal pleasure

Rawlette claims that her view elevates the status of animals (because their pleasure is no better or worse than ours), but gives practical reasons why a human life might be instrumentally more valuable than a given animal's life.

p. 224 Conclusion

Summary of the chapter.


p. 225

Quick wrap-up of the whole book.

No comments:

Post a Comment