Wayne Sumner, in the first six chapters of his excellent book Welfare, Happiness and Ethics, argues for what he calls an authentic life satisfaction theory of welfare. Somewhat generally, Sumner's theory of welfare is a sophisticated subjective account that treats one's happiness of a certain sort, in the right conditions, as enhancing one's welfare. Sumner spends the final, seventh chapter of his book arguing for welfarism – the thesis that welfare and only welfare matters in morality. What concerns me in this essay is not welfarism, but Sumner's theory of welfare. Sumner develops and defends what I think is the best theory of welfare that treats one's welfare as depending, at least in part, on one's mental states. In this essay, I shall critically explore Sumner's authentic life satisfaction account of welfare. After a discussion of Sumner's account of welfare, I conclude that it is inadequate.

I. SUMNER'S ACCOUNT OF WELFARE

There are two main kinds of subjective accounts: mental state accounts and desire-satisfaction accounts. Simple mental state accounts make one's welfare depend entirely on one's having certain mental states. Simple desire-satisfaction accounts make one's welfare depend entirely on the satisfaction of the relevant set of one's desires. Sumner's account of welfare on this taxonomy is a mental state account, but it is sophisticated because it does not make one's welfare depend entirely on one's mental states.

Sumner identifies four different kinds of happiness that one might think relevant to welfare: (1) being happy with or about something, e.g. being happy that the Blue Jays won the World Series; (2) feeling happy, e.g. blissful feeling, (3) having a happy disposition, i.e. tendency to have positive moods; and (4) being happy/having a happy life, i.e. having a positive attitude towards one's life.Footnote ¹ Sumner treats (4) as the principal mental state in his account of welfare. Sumner does this because only the fourth kind is about one's life. This kind of happiness Sumner calls ‘life satisfaction’.

Sumner thinks that one can have global or particular life satisfaction. Global life satisfaction is satisfaction about the whole of one's life. Particular life satisfaction is satisfaction about some specific aspect of one's life, such as one's work or family. If one were to have either or both kinds of life satisfaction, Sumner thinks that one will find one's life or life projects ‘enriching’, ‘rewarding’ and fulfilling.Footnote ²

Sumner's defense of life satisfaction as the relevant mental state for a theory of welfare is an important contribution to the literature on welfare. Mental state theorists generally have not been careful enough in explaining the mental state relevant to welfare. It seems clear that my happiness for someone that she got tickets to a play is simply not, in itself, relevant to my welfare. More must be said about what makes the relevant mental state, such as happiness, self-regarding.Footnote ³ More plausibly, Sumner claims that life satisfaction is the mental state relevant to welfare.

Simple mental state accounts run into serious problems. When the relevant mental state, such as life satisfaction, or the basis for such a state and one's actual circumstances diverge significantly, simple mental state accounts treat one's welfare as depending only on one's mental states, which is very counter-intuitive. Robert Nozick's example of the experience machine illustrates this problem for simple mental state accounts.Footnote ⁴

The exact details of Nozick's thought-experiment may mislead and are irrelevant to this essay. Instead, imagine that scientists have made a machine that can produce any possible sensation a person might have. The machine is so well developed that one could not tell whether one was in the machine or not. This machine produces sensations that could make someone in it satisfied with his or her life. However, the intuition goes, even if someone were to have a great deal of life satisfaction as a result of being in the machine, it is still possible that his or her life is not going well: in many cases, something in addition to life satisfaction is needed in order for a life to go well. On this point, Sumner and I agree.Footnote ⁵

Sumner develops his theory in such a way that it is supposed to be immune to problems stemming from experience machine-like examples. Roughly, he adds a requirement that one's life satisfaction be ‘authentic’ in order for it to enhance one's welfare. Sumner gives an admirably straightforward summary of his position in the following quotation:

According to Sumner, the fact that one's mental state merely is satisfaction with some or all of the conditions of one's life is not sufficient for that mental state to enhance one's welfare. Additionally, one's life satisfaction must be authentic, which includes being informed and autonomous.

II. SUMNER'S INFORMATION REQUIREMENT

Sumner considers three candidate information requirements. He rejects the first two and accepts the third. Of the first he says: ‘The strongest candidate would be a truth or reality requirement, which would stipulate that happiness counts as well-being only when it is based on a view of the conditions of our lives which is free from factual error.’Footnote ⁷ So, in order for happiness to count towards one's welfare, it must be based on true belief, according to this proposal. Sumner considers the reality requirement to be unreasonably ‘puritanical’. He says:

The reality requirement automatically, and too quickly, discounts life satisfaction that is not based on reality. The reality requirement undermines individual sovereignty, he says, because some people would still think that their lives were going well even under the belief that what they thought in the past was incorrect. For example, one could look back at time spent in a Nozick-style experience machine as increasing one's welfare.

The second information requirement Sumner considers is a justification requirement. The justification requirement is weaker than the reality requirement, since it requires only that one be justified in one's beliefs upon which one's happiness is based, which falls short of requiring truth, because one can have a justified false belief. Sumner finds the justification requirement slightly more plausible than the reality requirement, but he still finds the justification requirement too ‘arrogant’. On this Sumner says: ‘Once again it presumes to dictate to individuals how much their deviation from an ideal epistemic standpoint should matter to them. But that is for them to decide.’Footnote ⁹ However, it is not clear why Sumner is entertaining the justification requirement to begin with, since it does not solve the problems stemming from the experience machine-like case above. Surely, one could be justified in one's false belief that one is not in the experience machine.

Having rejected the reality requirement and the justification requirement, Sumner proposes his favored version of the information requirement. He suggests, provocatively, that the line of argument go in a somewhat different direction. He says:

Briefly, Sumner seems to favor the following thesis: A person has welfare (to a certain degree) only if both (i) the person has a mental state of life satisfaction and (ii) the person would approve of the conditions of her life, were she to be informed in the right way.Footnote ¹¹ It is of the utmost importance to get clear on the counterfactual nature of Sumner's theory. Sumner's thesis is not that any life satisfaction that is informed in the right way contributes to one's welfare. Any proposal along those lines runs afoul of Sumner's requirement that an account of welfare not be dogmatic. The counterfactual I speak of allows for an agent to choose her endorsement. No other proposal could do. It is important to grasp that, on Sumner's thesis, one can have a certain degree of life satisfaction, be completely wrong in one's beliefs about the grounds for one's life satisfaction, and yet still have a certain degree of welfare: this is provided, of course, that one would approve of the conditions of one's life. Unless Sumner's theory is counterfactual in this way, an individual is not the authority for her welfare.

Let us return to the amount and type of information one is to have when endorsing or not endorsing the conditions of one's life. At first glance, given what Sumner says, it might seem as though Sumner gives two different accounts of his information condition. He suggests, in one passage, that ‘the relevance of information for a person's well-being is a personal matter to be decided by personal priorities; there is here no authoritative public standard’.Footnote ¹² Sumner might be read as saying that each individual gets to choose whether some information is relevant to her choice or not.Footnote ¹³ But this cannot be right: for an agent may have different answers regarding which bits of information are relevant given different informational conditions. Absent an account of which epistemic viewpoint is superior in regard to an individual's endorsement of her circumstances, the truth-conditions of this construal of the information condition have not been given. Moreover, it is not clear how allowing an individual the choice about the amount and type of information he has will get around the Nozick-style experience machine cases described above.

Alternatively, Sumner suggests that an agent only be given information if his having it would make a difference in his evaluation of his life or some aspect of his life. Sumner says: ‘The place to start is with a (slightly) different question: when is (more) information relevant? The obvious answer, on a subjective account, is: Whenever it would make a difference to a subject's affective response to her life, given her priorities.’Footnote ¹⁴ On this construal of the information condition, one does not make a choice about whether some information is relevant. Instead, one's priorities play an indirect role in determining one's relevant information set. We can imagine someone who has the relevant mental state – life satisfaction – in abundance, and that the person is deeply misinformed about the real circumstances of his life. On Sumner's preferred version of the information requirement, the person's life satisfaction enhances his welfare only when correcting that misinformation would not make him somehow disqualify his life satisfaction from enhancing his welfare.

It is hard to see how Sumner's account of the information condition differs from a requirement that an agent be fully informed. Let me explain. As I interpret Sumner, he is trying to lay out the truth-conditions of his account of welfare, not a procedure one could use to actually figure out whether someone's life satisfaction enhances his welfare. Sumner is trying to provide an account of the conditions under which an agent's life satisfaction enhances her welfare. Sumner proposes that one's life satisfaction enhances one's welfare only if one would endorse the conditions of one's life were one to be informed in the right way. When we try to find out the conditions of what it is to be informed in the right way, we are given yet another counterfactual – an agent is to be given more information only if it would make a difference in her evaluation of the circumstances of her life. It must be possible for one to be fully informed, and ‘work back’ from there. If all of this is correct, then Sumner is committed to the possibility that one can be fully informed.

Even if a counterfactual full information requirement is the best among the range of alternatives with respect to an information condition, requiring full information could be problematic for Sumner's account of welfare. Connie Rosati, David Sobel, and others target a species of desire-satisfaction theories of welfare, so-called ideal advisor theories, according to which one's welfare varies with the extent to which the desires one would want for oneself, were one to be fully informed, are satisfied.Footnote ¹⁵ Rosati's and Sobel's arguments apply to any account that depends on the possibility of an agent being fully informed, even an account such as Sumner's. Now, I do not here say that Rosati's and Sobel's arguments work, but if they do, then I think that they would work against Sumner's account as well. Rather than pursue the prospects of full information theories generally, I shall explore other aspects of Sumner's account of welfare.

Recall that Sumner's information requirement is that one's life satisfaction enhances one's welfare only if one would endorse the conditions of one's life were one to be informed in the right way. There is another aspect of the information condition worthy of significant exploration; namely, the details of the requirement that one would endorse the conditions of one's life.

Given what Sumner says about what it is for an agent to endorse the conditions of his life, I find three distinctly different readings. The three accounts of what it is for an agent, A, to endorse the conditions of A's life are: (a) A makes a judgment that his life is going well, with regard to either a specific issue or his life as a whole, (b) A makes a judgment that his life satisfaction contributes to his welfare, and (c) A values either his life satisfaction or the grounds for his life satisfaction.

Consider first account (a). Sometimes Sumner seems to assert that what it is for one to endorse the conditions of one's life is merely to make a judgment that one's life is going well, with respect to some issue, or more generally. For example, Sumner says that there are two options one can take once undeceived about the lack of faithfulness of one's partner. Here they are:

It seems in this passage that Sumner has in mind that one merely makes a judgment about one's welfare, once one has the relevant information. The hypothetical agent here merely makes a brute judgment of her own welfare at a given time, once she gets straight on the facts of the matter. Elsewhere, Sumner says: ‘After all, what we are seeking is an adequate subjective theory of welfare, one on which the subject's point of view on her life is authoritative for determining when that life is going well for her.’Footnote ¹⁷ In this passage, one straightforward reading is that an agent's judgment is about, at least to some extent, her welfare. Moreover, Sumner says of each individual that: ‘Their self assessments are therefore determinative of their well-being unless they can be shown to be inauthentic, i.e. not truly theirs. The requirements that these assessments be informed and autonomous spell out the conditions of authenticity.’Footnote ¹⁸ If Sumner here means ‘self assessments of their own well-being’ then this would be reading (a). I argue, however, that (a) cannot be what Sumner has in mind, even if several passages seem to support such an interpretation.

Deep problems beset construal (a) of what it is for one to endorse the conditions of one's life. First, note that (a) requires that an agent use the concept of welfare in her endorsement of the conditions of her life, or lack thereof. Imagine that A is deciding whether to endorse the conditions of her life, when she is informed in the right way. Now, either A believes Sumner's account of welfare or she does not. If she does not, then the theory seems oddly self-effacing. Let me illustrate. Imagine that A believes a perfectionist account of welfare to be correct, according to which one's welfare is enhanced when one develops or exercises one's essential capacities. It would be an extraordinarily odd feature of Sumner's theory if it were to allow that A's judgments that she has developed her essential capacities and her corresponding judgment that her welfare has thereby been enhanced somehow result in her actually enhancing her welfare. That would be unacceptably self-effacing. Alternatively, A might believe Sumner's account of welfare. However, this would be unacceptable as well, since A cannot even ideally use the theory in making her decisions about what is good for her. Let me illustrate. Imagine A's thought process when she considers whether to endorse the conditions of her life. A thinks: ‘My welfare depends on my life satisfaction and my judgment of my level of welfare if I were to be informed in the right way. Now, I have some life satisfaction and I am informed in the right way. All that is needed now is my judgment about my welfare. Given my situation as it is, I have a certain level of welfare if I judge that I do. Well, do I? If I judge that I do.’ Certainly, A is in an odd predicament. It is unclear just what the agent who believes Sumner's account of welfare could possibly use as a criterion to judge her level of welfare when she is attempting to decide whether she endorses the conditions of her life or not. An agent who endorses the conditions of her life could not, it seems, believe Sumner's account of welfare and treat her endorsement as robust or substantive.Footnote ¹⁹

There are other problems with (a). It seems to require that all those who have some level of welfare possess the concept of welfare. But this seems wrong. First, Sumner's theory would seem to commit him to the claim that if some culture did not have the concept of welfare, none of its members could endorse the conditions of their lives in the right way and, in virtue of that fact, none could have any amount of welfare. This is an absurd conclusion. Moreover, we make judgments about the welfare of many creatures, even humans, such as infants and small children, who do not have the concept of welfare. It would be absurd, I take it, to claim that they do not have any welfare because they do not possess the concept of welfare, or are not capable of making judgments using the concept.

Reading (b) of the endorsement condition is slightly more plausible than (a). According to (b), one endorses the conditions of one's life only if one makes a judgment that one's life satisfaction contributes to one's welfare. Reading (b) has the advantage over reading (a), because (b) at least links up one's relevant mental state – life satisfaction – with one's welfare. Note that (a) does not require this link. However, (b) has the same devastating objections that confront (a). Reading (b) requires that an agent make a decision about whether to endorse the conditions of her life using the concept of welfare. The dilemma is not resolved by linking up one's life satisfaction with one's welfare.

I think that there is a plausible explanation for why Sumner often chooses to write as if an agent's endorsement of the conditions of her life depended on her using the concept of welfare. Really, in the passages that suggest (a), Sumner is using a kind of shortened expression for something more complex. He chooses to put ‘welfare’ within his statement of the content of an agent's judgment of endorsement, when, technically, it is problematic, but moving ‘welfare’ out of his explanation of the content of the agent's judgment would make for too long and complex an expression for what Sumner really has in mind.

Reading (c) is much more plausible than the other two. According to (c), loosely, one endorses the conditions of one's life when one values one's life satisfaction or the conditions that gave rise to one's life satisfaction. Reading (c) connects one's life satisfaction with one's endorsement of the conditions of one's life. This would seem to be an important connection to maintain through all of the counterfactual conditions of the information condition. Moreover, reading (c) does not require that an agent use the concept of welfare in his decisions about whether to endorse the conditions of his life or not.

Perhaps there is not a great deal of textual support for reading (c), but reading (c) is a charitable interpretation. Sometimes Sumner speaks loosely of an endorsement of the conditions of one's life as consisting in an ‘affirmation’ of one's life – which seems to support reading (c).Footnote ²⁰

With option (c) A would value, if relevantly informed, either (i) some life satisfaction or (ii) the fact that her grounds (her belief set, whether the beliefs be true or false) for some life satisfaction give rise to her life satisfaction. Although there are a few ways of fleshing out the details of (c), all interpretations along the lines of (c) share the idea that endorsement of the sort Sumner has in mind is a desire of a certain sort – a desire for the conditions of one's life. Let me explain. Endorsement can easily be understood as a desire, on my broad understanding of desires. Desires can be thought of as pro-attitudes generally speaking. One might think of a value for an existing state of affairs as a desire for an existing state of affairs. Normally one can desire what is the case. Valuing of this sort could very well be an endorsement of the conditions of one's life. So I think that the best way to construe Sumner's so-called endorsing is as a way of valuing which, in turn, is a way of desiring.

There is an additional benefit of this interpretation. Sumner requires that one's endorsement of the conditions of one's life be ‘up to one’ in the relevant way. Many desires are up to us in the relevant way and endorsing, as I construe it, can be up to us as well. Other mental states, such as beliefs, are not up to us in the way Sumner requires of his condition of endorsing. So, this interpretation of what it is to endorse the conditions of one's life seems to avoid the charge of presumptuous dogmatism.Footnote ²¹

With this rough idea of Sumner's information condition – including both the information condition and the endorsement condition, let us now look to Sumner's arguments for his information condition. Perhaps surprisingly, I argue that Sumner's lines of reasoning, when followed to their logical conclusion, support a desire-satisfaction theory.

Recall that Sumner says the truth and justification information conditions both are unacceptably dogmatic because they do not leave a person's welfare up to her. Sumner says, variously, that one's welfare depends on one's ‘priorities’, ‘personal priorities’, ‘point of view’ and ‘self-assessments’. However, one's priorities, self-assessments, and the like are only relevant to one's welfare when there is actual life satisfaction in the picture. When there is no actual life satisfaction in the picture, one's priorities, point of view, and self-assessments, are irrelevant to one's welfare. Sumner accuses those who endorse information conditions other than his own of being unreasonably puritanical because they fail to give significant weight to a person's own priorities, yet Sumner's final account of welfare is guilty of the same fault, I say.

Sumner says: ‘On a subjective theory, individuals are the ultimate authorities concerning their own welfare.’Footnote ²² He says that an individual is ‘authoritative for determining when that life is going well for her’.Footnote ²³ He says that a defining characteristic of subjective theories is that on such theories, ‘the individual is sovereign over her well-being’.Footnote ²⁴ In such passages, there is no hint that Sumner wishes to restrict or somehow qualify his claims about the subjective nature of welfare.

Clearly, Sumner thinks that an individual is to be sovereign over her welfare. On Sumner's authentic life satisfaction account of welfare an individual is not sovereign over her welfare. Even on option (c), Sumner's account requires actual life satisfaction to be in place in order for one to enjoy any degree of welfare. Here is the relevant question: Why restrict the authority of the informed people exclusively to matters of life satisfaction? On many accounts of welfare, such as several kinds of desire-satisfaction theory, an informed person's opinion in matters of happiness is on par with his judgments about other elements that make up his life. On Sumner's theory, only judgments relevantly related to happiness are important. Of course, if we presuppose that a happiness theory of a certain sort is the best account of welfare, then the restriction is natural. But, such a presupposition is unwarranted and we are left wanting an argument. In Sumner's arguments for his information condition, he gives no reason for thinking that an individual's authority and sovereignty over her welfare should be restricted to those cases where she has some actual life satisfaction. Yet on his account, an individual indeed is not truly sovereign over her welfare.

Sumner's arguments for his information condition really support some form of desire-satisfaction theory rather than Sumner's sophisticated mental state theory. Let me explain. If what it is for one to endorse the conditions of one's life on Sumner's account is as I claim above then endorsing involves desiring some or other conditions of one's life. Given Sumner's objection to ‘puritanism’ about welfare, we ought to reject the requirement that one have life satisfaction in order for one's endorsement of the conditions of one's life to enhance one's welfare. If this is correct, then what seems to be left is a desire-satisfaction theory of sorts, not a sophisticated mental state theory.

Interpreting the counterfactual condition in Sumner's theory as (c) seems to get around problems stemming from experience machine-like cases. Moreover, given what Sumner says in the text, with an eye to charitable interpretation, I suspect it is what he has in mind. However, the premises Sumner uses to support his information condition really support desire-satisfaction theories rather than his sophisticated mental state theory. His arguments support the wrong kind of subjective theory given his purposes.

III. SUMNER'S AUTONOMY REQUIREMENT

Recall that Sumner thinks that authentic life satisfaction constitutes welfare. Life satisfaction is authentic only if it is informed and autonomous. Sumner develops his autonomy condition as a response to the worries found in the work of Amartya Sen and others.Footnote ²⁵ Sen thinks that happiness theories and desire-satisfaction theories are all subject to a specific criticism. We can imagine cases where social conditioning might lead one to be happy or have a desire for a state of affairs, and yet there is something alien about the instance of happiness or preferences. A person could be happy and yet his happiness might not be autonomous. A classic example of this in the literature is someone who is a slave and yet is content with his position – the feeling of contentment having been formed as a result of social or other forms of conditioning. Sumner's way to get around examples such as that of the contented slave is to require that the happiness be autonomous.

Sumner discusses two main accounts of autonomy: the identification/hierarchy view and the process/formation view. A person's desire or preference is autonomous on the identification/hierarchy view when she has a higher-order desire for it. This kind of autonomy requires critical reflection on one's own desires and preferences. Though much of Sumner's language on this topic is drawn from the preference and desire literature, his target is one's endorsement of the conditions of one's life: on the interpretation I argue is most plausible, his target is higher-order desires about life satisfaction or the grounds of life satisfaction. Sumner points out that the main problem with the identification/hierarchy conception of autonomy is that higher-order values could themselves be socially influenced, and thus would not solve the central problem.Footnote ²⁶

The process/formation view of autonomy is concerned with formation, not identification. The basic worry is that there are autonomy-subverting processes, such as indoctrination, programming, brainwashing, role scripting, etc. The process/formation version of the autonomy requirement would make life satisfaction that is formed by an autonomy-subverting process irrelevant to welfare. Sumner finds the process/formation view unacceptable because it is possible for life satisfaction to be formed by an autonomy-subverting process and yet the agent later comes to identify with it.Footnote ²⁷

Sumner concludes that both of the main conceptions of autonomy have problems and he does not adopt either for his autonomy condition. Still, he thinks that however the details of the concept of autonomy are worked out, an autonomy condition, left unspecified, needs to be included in his account of welfare. Is this right?

Consider whether an autonomy condition is indeed necessary for Sumner's theory. Although Sen and others have argued against desire-satisfaction views, and by Sumner's extension, against a version of life satisfaction account, I have doubts about their allegedly problematic examples. The examples involve contented slaves, desires for material riches such as expensive cars, and so on. Such examples are tainted, I say, because they involve implicitly immoral activity, or at best morally debatable activity. Many accounts of welfare ostensibly are accounts of the non-moral good. This might strike the reader as an odd claim since I noted earlier that Sumner also argues for welfarism – the thesis that only welfare matters in morality. However, although Sumner argues for welfarism, he does not seem to think that the concept of welfare is an essentially moral concept. Since Sumner seems not to think that the concept of welfare is one that is essentially moral, examples such as the contented slave or expensive tastes might not be problematic for his account.

What would convince me that Sumner's account of informed life satisfaction needs an autonomy condition would be a series of clear examples involving clearly non-moral, or perhaps better, morally good life satisfaction about which it is counter-intuitive for one to say that having such life satisfaction contributes to one's welfare. Imagine that Mother Theresa's goal of feeding the starving was socially inculcated through her religion from childhood. Would we be so quick to claim that any life satisfaction which resulted from her feeding the starving should not contribute to her welfare? I suspect not. After all, we are social creatures and it would not surprise me if nearly all of our desires and life satisfaction were heavily influenced by society, family, friends, and so on. Something more is needed to show why ‘outside’ influences are somehow illegitimate.

I have another concern about Sumner's discussion of an autonomy condition: to wit, that the process/formation view is incompatible with Sumner's information requirement. Consider an example. Imagine that a super-duper neuroscientist manipulates A's brain in such a way that A now has an increased level of life satisfaction. Imagine that A, were he fully informed about the neuroscientist's manipulation of his mental states, nevertheless would endorse the conditions of his life. In such a case, if Sumner were to accept a process/formation autonomy condition, then he would have to deny that A's informed life satisfaction could enhance his welfare.

Recall that Sumner argues against the reality and justification versions of an information condition because he thinks that they are unreasonably puritanical and presumptuously dogmatic since they automatically discount an agent's welfare without leaving the matter up to her. Similarly, a process/formation autonomy condition does not leave the issue up to the agent: there is an automatic discounting that would be presumptuously dogmatic and puritanical. Sumner's arguments for his information condition would work against a process/formation view. If what I say above is right, and if Sumner is right in thinking that there are only two options for an autonomy requirement, then Sumner is left with an identification/hierarchy view. This is not a serious problem for Sumner, however, because identification with life satisfaction is up to us in the relevant way. Moreover, an identification/hierarchy view is not as problematic as Sumner seems to think. Let me explain.

Sumner worries that, on an identification/hierarchy view, higher-order mental states themselves can be subject to manipulation. This is true, but not problematic. If a neuroscientist were to manipulate A in such a way that the neuroscientist were to create (a) A's life satisfaction, (b) A's disposition to endorse the conditions of A's life, were he to be informed in the right way, and (c) A's identification with his life satisfaction, then I see very little reason for Sumner to say that A's welfare is not thereby enhanced. After all, the neuroscientist seems to have changed A's personality in a deep way. It should not at all be surprising if one's welfare could change with a deep change in one's personality. This seems like the sort of claim that most subjectivists about welfare should be happy to make (though perhaps not a welfarist). Moreover, the identification/hierarchy view is very similar to another feature of Sumner's theory – the requirement that one must endorse the conditions of one's life, since both are higher-order desires.

Now, I do not think that Sumner's authentic life satisfaction account of welfare is acceptable. As I argued, his arguments for his information condition, if followed to their logical conclusion, support a desire-satisfaction theory and not his sophisticated mental state theory. Even if I am wrong about that, I offer arguments against Sumner's account in the next section.

IV. A BLANKET TOO SMALL

Sumner's theory does a good job in discounting life satisfaction that is based on factors of which one would disprove, were one to be sufficiently informed. However, Sumner's theory fails spectacularly in cases in which there is no relevant life satisfaction to discount. Let me illustrate. Unfortunately, sometimes unconscious women are raped. Moreover, sometimes they never come to know that they are raped. To be sure, rape makes a woman's life go drastically worse than it otherwise would be, even if she never finds out about it.Footnote ²⁸ Simple mental state accounts cannot explain the loss of well-being in this sort of case, since there is no corresponding mental state to understand the loss of welfare, and such a failure is significant. Sumner's sophisticated account of welfare fails in this sort of case for similar reasons. If there is no life satisfaction tied to relevant beliefs, then there is no life satisfaction to discount, and thus no loss of welfare. So Sumner's account is implausible for a reason that simple mental states are implausible.

It is possible to modify Sumner's account to address this kind of criticism. One way is to rule out the need for any actual life satisfaction at all – the theory of welfare could, in a manner of speaking, go fully counterfactual. According to this new view, one's welfare is constituted by the life satisfaction one would have were one were informed in the proper way. At first glance, this might seem a promising response, since it does seem that if a victim of the sort described above were to be sufficiently informed, she would have a lower life satisfaction than she otherwise would have in ignorance (and this then explains her loss of welfare in being raped). However, this cannot be Sumner's response. It is not a view that Sumner can accept, since it is too far removed from his stated goal of requiring some individual to have some actual happiness in order for that person to have welfare – his so-called experience requirement. Moreover, this new account still does not allow the woman to be sovereign over her welfare in the way urged by Sumner in his book, since she still must have life satisfaction, at least counterfactually.

Sumner's theory seems a blanket too small. With his counterfactual condition, he can cover some of the implausible implications for pure mental state accounts of welfare, but his account cannot deal with all of the significant objections. Pure mental state accounts of welfare do not account for experience machine-type cases. Sumner's theory does. Pure mental state accounts of welfare do not account for examples such as the rape case discussed above. In such cases, Sumner's theory fails for the same reasons that the pure mental state accounts fail. Note that desire-satisfaction accounts have no such problems explaining the decrease in welfare in such cases, for surely women typically have a desire, at least dispositional, for sexual autonomy. As I have been suggesting throughout, for an individual to be truly sovereign over her welfare, as Sumner maintains, some kind of desire-satisfaction account of welfare would have to be true.Footnote ²⁹