Types of subjective welfare

[-]Dagon2y31

Interesting exploration. I have some confusion, or maybe disagreement, about the classification. The first three (hedonic, felt desires, and cognitive beliefs) are distinguishable, but not necessarily distinct. They're so interrelated that they may simply be different views of the same thing. Or they may be different focuses, but deeply influencing each other in ways that are hard to disentangle. Felt desires, for instance, seem a lot like anticipation of hedonistic states, and belief-like preferences are a recursive modeling of the other two, in oneself and others.

The last (Choices and behavior-revealed preferences) is, presumaby, a result of the other three, not a distinct kind of welfare. Interestingly, it's the ONLY view that other people have of someone's internal state. Especially if speech is noticed as an action, and the choice of whether and what to say is just another choice, from which we can infer a bit about internal states.

[-]MichaelStJules2y10

Thanks!

I would say experiments, introspection and consideration of cases in humans have pretty convincingly established the dissociation between the types of welfare (e.g. see my section on it, although I didn't go into a lot of detail), but they are highly interrelated and often or even typically build on each other like you suggest.

I'd add that the fact that they sometimes dissociate seems morally important, because it makes it more ambiguous what's best for someone if multiple types seem to matter, and there are possible beings with some types but not others.

[-]sweenesm2y20

Thanks for the post. I’d like to propose another possible type of (or really, way of measuring) subjective welfare: self-esteem-influenced experience states. I believe having higher self-esteem generally translates to assigning more of our experiences as “positive.” For instance, someone with low self-esteem may hate exercise and deem the pain of it to be a highly negative experience. Someone with high self-esteem, on the other hand, may consider a particularly hard (painful) workout to be a “positive” experience as they focus on how it’s going to build their fitness to the next level and make them stronger.

Further, I believe that our self-esteem depends on to what degree we take responsibility for our emotions and actions - more responsibility translates to higher self-esteem (see “The Six Pillars of Self-Esteem” by Nathaniel Branden for thoughts along these lines). At low self-esteem levels, "experience states" basically translate directly to hedonic states, in that only pleasure and pain can seem to matter as "positive experiences" and "negative experiences" to a person with low self-esteem (the exception may be if someone's depressed, when not much at all seems to matter). At high self-esteems, hedonic states play a role in experience states, but they’re effectively seen through a lens of responsibility, such as the pain of exercise seen through the lens of one’s own responsibility for getting oneself in shape, and deciding to feel good emotionally about pushing through the physical pain (here we could perhaps be considered to be getting closer to belief-like preferences).

^{^}

Including possibly the relationships between those perspectives or mental states and the world, e.g. that things actually go as someone prefers, not just that they believe things are going as they prefer, as in the experience machine thought experiment (Wikipedia, Buscicchi, 2014, Carlsmith, 2021).

^{^}

I use ‘preference’ over ‘desire’ for belief-like and choice-based preferences, because desire typically has connotations of being directed towards (or away from, e.g. aversive) specific objects or features (or states of affairs), but preferences can also just be comparative or pairwise (Schroeder, 2015, sections 2.1 and 3.3), so are more general. Felt desires, on the other hand, do seem to be directed.

The ‘felt’ in ‘felt desire’ is based on Helm’s (2002) concept of “felt evaluations”. I don’t want to commit to “felt desires” being types of “felt evaluations” in case that Helm’s account turns out to be wrong, but it’s clear that there are desires that feel like something, and if Helm’s account is correct, then felt desires are probably types of felt evaluations, along with hedonic states.

For belief-like preferences, ‘belief-like’ is more specific than ‘cognitive’, which has also come to be used fairly broadly to include study of consciousness and emotion. There’s also no sharp line between the cognitive and affective (Pessoa, 2008, Shackman et al., 2011) and felt desires, understood as motivational salience, are therefore characterized by their effects on attention, a cognitive/executive function, and so felt desires are a form of cognitive control. On the other hand, there are various accounts of what beliefs are and require, and I don’t want to commit to these types of preferences necessarily coinciding with whatever account of belief we eventually settle on, so I use ‘belief-like’, not, say, ‘belief-based’.

I use 'choice-based' over 'behavioral' to emphasize voluntary choice.

^{^}

I avoid the term suffering as definitional for negative hedonic states, because it often has other connotations. Brady (2022) defines suffering as “negative affect that we mind, where minding is cashed out in terms of an occurrent desire that the negative affect not be occurring”. Corns (2021) and McClelland (2019, preprint) define suffering at least partially in terms of disruption of agency. Dennett (1995) also distinguished suffering from negatively-signed experience:

While the distinction between pain and suffering is, like most everyday, nonscientific distinctions, somewhat blurred at the edges, it is, nevertheless, a valuable and intuitively satisfying mark or measure of moral importance. When I step on your toe, causing a brief but definite (and definitely conscious) pain, I do you scant harm -- typically none at all. The pain, though intense, is too brief to matter, and I have done no long-term damage to your foot. The idea that you "suffer" for a second or two is a risible misapplication of that important notion, and even when we grant that my causing you a few seconds pain may irritate you a few more seconds or even minutes -- especially if you think I did it deliberately -- the pain itself, as a brief, negatively-signed experience, is of vanishing moral significance. (If in stepping on your toe I have interrupted your singing of the aria, thereby ruining your operatic career, that is quite another matter.)

^{^}

To highlight some important possibilities, it could turn out to be the case:

1. that the correct account of belief is pluralistic, i.e. it suffices to meet one of multiple sets of requirements,

2. that some accounts should be combined and to have beliefs means meeting the combined requirements, or

3. that an individual has beliefs in a more robust sense the more such requirements they meet.

^{^}

Some have argued that there’s no way to distinguish them and in fact entertaining is a kind of believing (e.g. Mandelbaum, 2014). However, believing could be a predisposition to judge, and judging could just mean entertaining + affirming (Kriegel, 2013), so we would just need to characterize affirming. Believing and entertaining might feel different (see Dietrich et al., 2020 on epistemic feelings) or otherwise have different effects on us, our motivations or actions, e.g. we only act as if the things we actually believe are true (see Schwitzgebel, 2023 on accounts of belief). Or, believing might require greater consistency with other beliefs and entertainments than mere entertaining can achieve (see Dennett, 2013 on holism). Or, we actively apply built-up concepts for belief or truth vs entertaining/hypothetical or false (or just one of the two concepts).

These concepts may be built up differentially related to capacities like surprise, reality monitoring, i.e. distinguishing between externally derived and internally generated information or perceptions, among others.

For beings unable to entertain, they may have no need to distinguish, and may just believe all of their perceptions.

One way might be via recruiting motivational salience, intentions and/or action towards satisfying the preference, but sometimes we are not motivated this way by what we genuinely believe matters. So, an account more general to belief seems necessary, or one that allows multiple ways of judging.

^{^}

A fairly successful utilitarian could judge their own life as far better than the typical person’s, judging based on their expected impact on total welfare. A dutiful absolutist deontologist could also judge their own life as far better than the typical person’s, because the typical person violates some of their duties, and such violations are absolutely impermissible. The utilitarian and deontologist could each judge their own life as far better than the other’s, if the utilitarian violates duties according to the deontologist, and the deontologist has only a small impact on the welfare of others. Each may see the stakes as far larger on their own views than on the other’s (MacAskill, Bykvist & Ord, 2020, 6.I). Different utilitarians can similarly disagree with one another, and different deontologists can similarly disagree with one another.

Barrett (2019) illustrates similarly:

Suppose that, in [the outcome] x, i is a political activist in jail and k is a corrupt politician at a fancy resort. i might personally prefer to be at the resort than in jail, but loathe k and everything k stands for so much that i would rather be herself in jail than k at the resort.

There were also huge discrepancies in preferences between community cash transfers vs saving the life of a child in the community according to a survey of respondents from Ghana and Kenya by IDInsight, with 38% preferring to save the life up to or past the maximum surveyed ($10 million), and of those preferring community cash transfers at any point below the maximum, 54% preferred community cash transfers at values below $30,000 (Redfern et al., 2019, p.18). IDInsight (2019) also wrote

For example, a number of respondents placed extremely high value on averting death relative to increasing income using deontological arguments arguing that the two just cannot morally be compared, which is in direct contrast to GiveWell’s utilitarian approach.

Counting only “personal” preferences doesn’t address the problem, because people’s personal preferences can be indirectly other-regarding with hugely varying stakes between them and between them and other personal preferences, e.g. being a good parent, fulfilling their moral duties, maximizing their net impact on total welfare. Discounting even the indirectly other-regarding preferences seems alienating. And even if we did discount them, we can still imagine hugely varying entirely personal stakes, e.g. someone could recognize absolutist moral duties to themself.

^{^}

It isn’t often necessary to know how much better one option is than another to make choices, so we don’t bother to represent that information, nor would our functions, dispositions or other consequences covary with it reliably across contexts. Schroeder (2015) wrote:

I can tell that I prefer securing my father's health to doing the laundry, and I know which I would choose if it came to a choice. But can I introspect the degree to which I prefer my father's health over clean laundry? Perhaps not. And if I cannot, perhaps this is because I have introspective access to the most basic psychological facts only, and these are facts about simple pairwise preferences.

See also Brandstätter et al., 2006 for the priority heuristic in decision-making.

There’s substantial neurological and behavioural evidence against representations of value in a “common currency” even within each person’s brain across tasks (e.g. Hayden and Niv, 2021, Garcia et al., 2023, Kellen et al., 2016), as well as evidence of intransitivity or dependence on irrelevant alternatives, including even in honeybees (Shafir, 1994).

We can off-load representations of value externally, e.g. to paper or computers, and even do so with computer programs without ever actually seeing the final representations of value, and instead just the recommendations, like an algorithm filtering and selecting stocks to buy or movies to watch. Or, we can defer to the judgements of other people.

The contents of the belief-like preferences can differ dramatically in type or scales to which they refer, like pleasure intensities and durations, QALYs saved, tonnes of CO₂ emissions averted, thousands of tonnes of CO₂ emissions averted, or RAM in a laptop. Only sufficiently similar features can be used to ground comparisons.

Global preferences can dissociate from other measures, like hedonic states, felt desires, choice-based preferences — as discussed in a later section —, and probably attention.

^{^}

For example, people probably don’t equally value a year of life at full health — which would follow from treating each QALY equally — because they probably tend to come to value their own lives more if they have young children or partners, for the sake of those children or partners or because they value their time with their children or partners. They tend to be less willing to sacrifice or risk years of their own life to live in better health (van der Pol & Shiell, 2007, Matza et al., 2014, Hansen et al., 2021, van Nooten et al., 2015).

^{^}

We could consider behaviours and behavioural dispositions in general, but this seems too trivial to me to warrant any significant concern. For example, electrons tend to avoid one another, so we might interpret them as having preferences to be away from each other. Choice-based preferences (and perhaps other kinds of welfare here) could be understood as far more sophisticated versions of such basic behavioural preferences, but choice seems like a morally significant line to draw. For more on how simple behaviours in fundamental physics might indicate welfare, see Tomasik, 2014-2020.

^{^}

Or, I prefer choosing the apple to choosing the orange for more indirect reasons related to my choice.

^{^}

Haggard (2005) summarized the neuroscience of intentions:

Instead, recent findings suggest that the conscious experience of intending to act arises from preparation for action in frontal and parietal brain areas. Intentional actions also involve a strong sense of agency, a sense of controlling events in the external world. Both intention and agency result from the brain processes for predictive motor control, not merely from retrospective inference.

^{^}

On the possibilities of unconscious emotions and unconscious affect, see, for example, Berridge & Winkielman, 2003 and Winkielman & Hofree, 2012 or Smith & Lane, 2016.

Incentive salience, the apparent basis for appetitive felt desires, can also be unconscious. Berridge and Aldridge (2009, ungated) write:

Importantly, incentive salience attributions are encapsulated and modular in the sense that people may not have direct conscious access to them and find them difficult to cognitively controls (Robinson & Berridge, 1993, 2003; Winkielman & Berridge, 2004). Cue-triggered “wanting” belongs to the class of automatic reactions that operate by their own rules under the surface of direct awareness (see Chapters 5 and 23; Bargh & Ferguson, 2000; Bargh, Gollwitzer, Lee-Chai, Barndollar, & Trötschel, 2001; Dijksterhuis, Bos, Nordgren, & van Baaren, 2006; Gilbert & Wilson, 2000; Wilson, Lindsey, & Schooler, 2000; Zajonc, 2000). People are sometimes aware of incentive salience as a product but never of the underlying process. And without an extra cognitive monitoring step, they may not even be always aware of the product. Sometimes incentive salience can be triggered and can control behavior with very little awareness of what has happened. For example, subliminal exposures to happy or angry facial expressions, too brief to see consciously, can cause people later to consume more or less of a beverage—without being at all aware their “wanting” has been manipulated (Winkielman & Berridge, 2004). Additional monitoring by brain systems of conscious awareness, likely cortical structures, is required to bring a basic “want” into a subjective feeling of wanting.

^{^}

However, in response to this last quoted paragraph, if we are functionalists, then motivation should be definable in purely functional terms. For example, the conscious states could have important effects on mental states worthy of identifying as motivation even without externally-directed behaviour, e.g. effects on attention in particular — like felt desires — or cognitive control in general.

^{^}

Or when a hedonic state and an aversive felt desire occur together, e.g. pain is often simultaneously unpleasant and aversive.

^{^}

Other effects could be unnecessary for and dissociable from the pleasure and unpleasantness themselves, e.g. optimism and pessimism bias, learning/reinforcement.

^{^}

Kammerer (2019), considering the normative implications of illusionism in general, not just the particular interpretation that assigns value to illusions, also writes:

The best option here for the illusionist would probably be to draw inspiration from desire-satisfaction views of well-being (Brandt 1979; Heathwood 2006) or from attitudinal theories of valenced states (Feldman 2002), and to say that pain is bad (even if it is not phenomenal) because it constitutively includes the frustration of a desire, or the having of a certain negative attitude of dislike. After all, when I am in pain, there is something awful which is that I want it to stop (and that my desire is frustrated); alternatively, one could insist on the fact that what is bad is that I dislike my pain. This frustration or this dislike are what makes pain a harm, which in turn grounds its negative value. This might be the most promising lead to an account of what makes pain bad.

^{^}

The word ‘intentionality’ is used to describe mental states being about something (Jacob, 2023). This is of course more general than an intention to do something, as a kind of plan or commitment, which is how I’ve used the word in the rest of this piece.

^{^}

We might think of hedonic states as necessarily occurrent by definition, but there are still relevant dispositions.

^{^}

Someone could report their previous judgement if nothing of importance has changed or use it as a baseline from which to judge again.

^{^}

We could in principle just manipulate or replace a person’s global preferences to be more satisfied after the fact, by directly changing how they evaluate their life. We need to count your original global preference(s) as less satisfied or further frustrated, and this shouldn’t be offset or outweighed by your new global preference(s). As far as I can tell, the only robust subjective welfare solutions to this are preference-affecting views, similar to a person-affecting view but applied to individual preferences (e.g. Bykvist, 2008), and antifrustrationism (Fehige, 1998), the axiology of negative preference utilitarianism. These could be applied to global preferences only.

People are often okay with some kinds of preference change or reasons for preference change, e.g. further reflection or experience, and this could already be reflected in their global preferences. In those cases, it wouldn’t be worse for their global preferences to change. To accommodate this, people’s global preferences could be understood as made up in part by placeholders:

I prefer whatever I prefer now or what I would prefer with further reflection or experience in typical ways, with my consent and on the basis of accurate beliefs. This excludes things like coercion in general, drugging, brain stimulation and brain reorganization without my informed consent, as well as deception and misleading. There could be other ways of preference change of which I’d disapprove that I haven’t thought of yet, but I could recognize them with full information ahead of time.

LESSWRONG
is fundraising!
LW

LESSWRONG
is fundraising!
LW

10

Types of subjective welfare

10

10

Summary

Acknowledgements

The four types

Hedonic states

Felt desires

Belief-like preferences

Choice-based preferences

Welfare should be conscious

Moral plurality or unification

More on unification

Do choice-based preferences matter?

Welfare types can influence each other

Dissociation between welfare types

Variations