Hedonism Revisited

Thinking about what we want and like

Image from the opening credits of American Gods

There’s a rough cluster of concepts that intuitively feel like they should mean more-or-less the same thing:

• what I want

• what I like

• what feels good

• what I prefer

• what I intend or attempt

• what I’m drawn to

• what makes me happy

• what is good for me

Of course, when you examine these things more closely, and try to specify them, it turns out they’re very different.

And this produces lots of arguments.

• “Happiness-as-in-eudaimonia isn’t just pleasure, otherwise everyone should shoot themselves up with heroin!”

• “What people choose doesn’t always make them happy — so what’s better, coerced happiness or unhappy freedom?”

• “What if wealth makes people less happy because their expectations are higher — would that mean economic growth is bad?”

Hedonistic utilitarianism, in particular, doesn’t make any sense if we aren’t clear on what pleasure is.

Wanting and Liking

“Wanting” and “liking” are not the same.

There are separate neural circuits (in rats) for “wanting” (motivation to seek out a food) and “liking” (pleasure, evinced by behaviors like licking their lips). You can experimentally knock out the “wanting” circuit in isolation (leading to rats that still enjoy their favorite food but don’t seek it out) or knock out the “liking” circuit in isolation (leading to rats that eagerly seek out their favorite food but get no pleasure from it.)

In both mammals and humans, some activities (like eating) are both “pleasurable” and “craveable”, both liked and wanted. There is a direct sensation (hunger) of wanting-to-eat. Likewise, thirst is “wanting-to-drink”, sleepiness is “wanting-to-sleep”, restlessness is “wanting-to-move”, and libido is “wanting-to-have-sex.” And, of course, people and mammals can experience pleasure while they are eating, drinking, drifting off to sleep, engaging in some kinds of exercise, and having sex.

But some activities, like jhana meditation, are reportedly extremely pleasurable yet not at all craveable, to the point that people will claim that they will frequently forget to get around to an activity that they also describe as the most intense bliss they have ever experienced.

To a less extreme extent, I think it’s not at all uncommon to have activities that you know you enjoy, but you somehow never get around to doing. Pleasurable but not craveable.

In the other direction, there are activities that are unpleasant yet highly motivating.

Video games are famously intrinsically motivating — people sink countless hours into them, for no external reward. But hard video games are often frustrating, not pleasurable. The player keeps trying because he’s angry he didn’t get it the last time.

Other activities — getting into arguments; solving tough problems — can be similarly frustrating yet addictive. You keep going because you feel like you “gotta finish” or “gotta fix it”, but you aren’t exactly having a good time. Your body language is strained, your mood is irritable. Continuing the activity is craveable but not pleasurable. You want to continue but you don’t like the activity.

Wanting and liking are equally “immediate” or “experiential”. They are both directly felt subjective experiences, which we share with at least mammals (and maybe other animals as well.)

Wanting is Motor

Wanting, or motivation, in the sense that the rodent experiments refer to, is an urge to act.

I think the most primitive, immediate form of “wanting” is probably always directed at a motor action, not a state of the world.

When you say you “want to travel to DC”, that’s a derived quantity, an abstraction.

The “desire to go to DC” is built up out of many concepts: a concept of the city of DC, of the many individual steps you would take in planning a trip, etc. When you think of “the trip to DC” as something you want, that’s actually made up of many (imagined or predicted) component experiences and concepts.

“I want to go to DC” is not an atomic desire.

An atomic desire, something that had no “smaller” or “simpler” component parts, would have to be something more like “I want to move my hand just so.”

You can actually, introspectively, observe “wanting” in isolation, in its most atomic form, through meditation.

Sit perfectly still for a while and force yourself to resist any urges to move (to scratch your face or shift your weight or whatever.)

You’ll start to notice things that I call “ghost movements.”

If you resist an urge to scratch your face, and you keep resisting it as it gets itchier and itchier, you can feel your mind sort of “shaping the path” that your hand would take if you did raise it.

You can isolate that “ghost movement” all by itself, without the actual hand movement. It’s not the verbal thought “I want to scratch”. It’s not the discomfort from the itchiness. It’s a simulated movement that very closely resembles the experience of actually moving.

The “ghost movement” is a very pure isolated form of the “urge to move” or “intent to move” or “will to move.” It is just the will, without the actual motion.

The interesting thing, which you can directly perceive, is that the “will-to-move” is very tightly coupled to the motion itself, such that they almost never separate without extraordinary effort.

I’ve heard an aikido practitioner explain that the way you can throw an opponent with “qi”, aka without touching them, is connected to this “ghost movement” or “will to move.” It’s essentially a very precise feint. You “make like” you’re going to move in a particular direction, but then you don’t; and the opponent follows your “ghost movement” rather than your actual body, and falls over. The feint is so quick and subtle that people think they’re using “magic”, or just “willing” the opponent to fall over with their minds, but of course it’s a bodily motion.

An “urge to act” or a “drive to act” can also apply to mental “motions”. You can feel an urge to pay attention to something.

You can also feel an urge to have a particular kind of mental posture/tone/attitude, which is very akin to a craving for a physical posture or movement.

When you feel an urge to continue an activity, the activity is of course a complex abstraction; I feel “motivated” to write this blog post right now, and of course a “blog post” is a complex concept. But I think it’s very plausible that the “urge” or “drive” to write a blog post is built up out of an “urge” or “drive” towards some much simpler behavior components, which are very nearly kinaesthetic.

As I write, I can feel how the chain-of-words is an ever-lengthening string and I have to figure out at each point how to “solve the problem” to complete the string. And there’s a feeling of “wanting/seeking/urge” around clicking the little puzzle pieces into place. And then my brain goes “more! again! do another one!”

And, of course, the kinaesthetic aspect makes sense neurologically. Motivation, addiction cravings, and motor learning tend to localize to the dopaminergic parts of the brain — the striatum of the basal ganglia and its projections into the forebrain such as the nucleus accumbens. Dopamine-destroying conditions like Parkinsonism, as well as dopamine-inhibiting drugs such as antipsychotics, inhibit both motivation and movement.

Dopaminergic drugs such as amphetamines, cocaine, and L-Dopa famously make people more restless and can produce “punding”, the craving to engage in repetitive activities such as:

• collecting objects

• tidying, straightening up, reorganizing, sorting

• assembling and disassembling machines (watches, guns, cameras, fridges, bicycles, radios, etc)

• tinkering (with model trains, computers, garden)

• writing or drawing

• playing computer games

• fidgeting (pacing, shuffling papers, playing with hair)

These are obviously motor activities, even the “fine motor” or “metaphorically motor” ones like writing, drawing, or using the computer.

There is, I think, a single neurological and experiential phenomenon that is the punding urge, which incorporates tinkering, fidgeting, “puttering", sorting and organizing, collecting, etc.

Healthy people who aren’t on dopaminergic drugs have this urge too; it’s just that we call it “punding” when a meth head stays up all night digging holes in the yard, and we call it “having hobbies” when you spend weekends working in your garden. But mostly I think that’s a difference of degree and context, not of kind.

One thing to note is that punding is described as “purposeless” activity, as opposed to goal-oriented activity. A person with hypergraphia writes or draws continuously, but may throw all the output away. Assembling and disassembling, sorting and resorting, are repetitive activities that never end. The reason people go to a psychiatrist about punding is because their compulsive activities interfere with other priorities in life, like sleep, work, or relationships.

Similarly, there is a component of ordinary motivation that is not goal oriented, or at least not about weighing goals by their overall importance in life. How “motivating” an activity is, how much you feel inclined to start doing it and keep doing it, is often a very local, process-oriented thing about whether you’ve gotten “hooked” on the activity.

Motivation is, at its most fundamental level, the urge to do, which means the urge to move.

Liking is (Mostly) Painlessness

If “wanting” is driven by dopamine and the dopaminergic parts of the brain, "liking” is focused on a few hotspots in the orbitofrontal cortex and ventral pallidus and is enhanced by opioid injections. A rat given an opioid agonist in one of the brain hotspots will express more pleasure (via lip-licking and similar gestures) in response to sugar.

Rats, apes, and human infants make the same faces to express pleasure and aversion.

Of course, opioid drugs are primarily used for pain relief. And the mu opioid receptors in the brain and nervous system are responsible for nociception (the sensation of pain.)

What does a pleasurable sensation (like the taste of sweetness) have to do with pain?

The philosopher Epicurus wrote that the greatest pleasure is simply the removal of all pain. “Pleasure in the body admits no increase when once the pain of want has been removed; after that it only admits of variation."

Epicurus isn’t saying that all pleasure simply is the absence of pain. He’s saying something subtler: that the magnitude of pleasure tops out at “zero pain.” In other words, we have some subjective experience of hedonic qualities as being measurable or comparable; we can sometimes identify one thing as “feeling better” than another. And this comparative function has an absolute ceiling at the point of zero pain.

At first I didn’t take this seriously. “Merely” not being in pain sounds like a sort of neutral state, a state of “meh, not bad”, rather than a state of maximum pleasure.

Surely “positive” pleasure, enjoyment, delight, bliss, etc., is better than mere painlessness?

But actually, when you feel “meh”, that’s not painless.

Feeling "meh" means feeling an ordinary, mild amount of muscle tension, gastrointestinal discomfort, fatigue, worry, shame, boredom, disappointment, etc. It's too little pain to complain about, certainly, but definitely detectable if you admit it to yourself.

A state of total painlessness means utter ease, comfort, fearlessness, flexibility -- it's truly extraordinary.

You can also have “positive pleasures” on top of it — things like listening to a symphony or having an orgasm — but people don’t necessarily find those pleasures strictly better than the “plain” state of painlessness, once they’ve experienced it. Nobody wants to keep repeating a particular pleasant activity, like listening to symphonies or having orgasms, literally all the time without a break.

In fact, people who practice jhana meditation often report that states of thrilling bliss are fun to try, but ultimately they’d rather just hang out in a state of understated calm most of the time.

That sounds very much like Epicurus’ claim: once pain has been truly eliminated, pleasure doesn’t increase in magnitude any more. Pleasure only varies, allowing different kinds of pleasurable experiences from different activities, all about equally (or incommensurably) good in terms of subjective pleasantness.

This is why Epicurus believed that the most pleasant possible life was easily attainable and required only modest material wealth. You only need enough security and resources to ward off physical pain, and then a bunch of study and contemplation to rid yourself of mental pains.

The Epicurean model allows for the possibility that when pain is present, positive pleasures may to some extent compensate for pain. We might possibly experience “liking” a state with both great pain and great pleasure more than a state with less pain and less pleasure. But nothing feels better than the complete absence of pain — not even the complete absence of pain plus additional pleasures!

Is this ancient philosophical model borne out by modern evidence?

Pain relief and pleasure are clearly interrelated.

There is a well-documented phenomenon of “pleasure analgesia”, where both humans and animals are less sensitive to pain when exposed to pleasurable sensations such as pleasant smells, images, music, food, sexual stimuli, or maternal cuddles (in preterm infants). Pleasure causes pain relief.

Also, chronic pain sufferers experience high rates of anhedonia, the inability to experience pleasure from normally enjoyable experiences. Moreover, blocking opiate receptors (which causes greater pain sensitivity) reduces animals’ attraction to especially preferred stimuli (favorite foods, females in estrus) but not less-desirable stimuli (normal chow, females not in estrus.) Pain interferes with pleasure.

And, of course, pain-relieving drugs (opiates) can also produce pleasure, though interestingly enough they don’t always. An analgesic dose of opiates in healthy drug-naive humans doesn’t produce pleasure in all subjects.  In one study of patients given opiates prior to surgery, patients didn’t report increased levels of “feeling good” on the drug, and only reported decreased levels of anxiety on the drug if they had high anxiety at baseline.

So, we can conclude:

• analgesia isn’t always identical with positive pleasure, but:

• positive pleasures can produce analgesia

• analgesia can relieve the pleasure-blocking properties of pain.

This is consistent with the Epicurean model, though it doesn’t fully demonstrate it.

Positive pleasures can trade off against pains (that’s pleasure analgesia). Also, pain relief (through opiates, endogenous opioids, or brain stimulation) enhances positive pleasure precisely because pain weakens the ability to experience positive pleasure from external stimuli.

Opioids and similar “pain-relief” mechanisms will not be experienced as particularly pleasurable to organisms that aren’t in pain (physical or psychological) to begin with.

Relating Motivation and Pleasure/Pain

Why do we (apparently) have a separate motivation system and pleasure/pain-relief system, implemented with different neurons using different neurotransmitters?

From the perspective of evolution, an organism “should” take actions to improve its inclusive fitness.

You want some kind of signal to tell the organism “do this” and “don’t do that.”

Pleasure/pain does seem to connect to motivation, in a complex way I don’t currently understand; animals are at least sometimes motivated to seek pleasant things and avoid painful things.

But we don’t have a single signal of value to tell us how desirable/undesirable a thing is— we have a system with at least two parts, the “wanting/motivation” and “liking/pleasure/pain” parts.

And “wanting” seems to be more about actions (specifically motions) while “liking” seems to be more about experiences (or sensations).

You could instead imagine a system that only has “wanting” — the organism is motivated to do things that are good for it.

Or you could imagine a system that only has “liking” — the organism scores beneficial experiences as “pleasant” and harmful experiences as “unpleasant” and takes whatever actions it thinks will be pleasant and not unpleasant.

So why don’t we do either of these simpler things?

In a “wanting-only” system, you run into the problem that the same action (in different circumstances) might be either good or bad for the organism’s inclusive fitness. And the effects of actions might change, or be unpredictable; it might not be practical to hard-code the “right” action for every situation, at least in more complex organisms. Any organism that has neurons is built to be capable of at least some adaptation.

You want the organism to be able to learn from outcomes and not only pay attention to actions — you want the fact that an action, in some circumstance, turned out well or turned out badly, to affect how the organism behaves in future. And so “wanting-only” can’t suffice for an organism that learns.

On the other hand, in a “liking-only” system, you have a very challenging learning problem. The organism has to learn from scratch which actions, in each situation, will result in good outcomes.

This is equivalent to reinforcement learning, which in computational contexts is challenging because there’s a huge range of possible actions and sequences of actions to compare relative to the relatively scanty data on outcomes.

RL agents in games are only tractable, as far as I understand, because they do a lot of simulated “self-play”, and it’s trivial to score an imaginary game of chess or Go, once it’s complete, to see who the winner would have been.

In nature the situation is more like “a squirrel has to predict which sequences of movements will result in it being fed, injured, or eaten”, based exclusively on its extremely limited past experience with these outcomes — and this does not seem at all tractable. It’s not plausible that squirrel brains actually work like pure RL agents.

Adding an analogue of “wanting” to an RL agent might look something like incorporating some kind of “prioritization” or “weighting” over possible paths.

The agent “wants” to go in some directions, or make some moves, preferentially. (This cuts down on the search space.)

When the agent gets feedback on how much it “likes” how things turned out, this might modify the “wanting” function. Including, but not limited to, updating the agent to “want” to do more of the thing that produced a result it “liked.”

But feedback could also change the “wanting” function in other ways. Like “I’m safe and well-rested — now I want to explore more!” That’s not limited to “increase motivation to do more of whatever got you to become safe and well-rested.”

Or, it might be in an organism’s interest to do something that “hurts” now but pays off later. (Or, even, something whose “payoff” goes to close genetic relatives rather than the individual itself.) A preexisting motivational “drive” to do that thing might override a predicted “that’ll hurt”.

Having “drives” to do things, in addition to “reward signals”, allows some things that look like “hard-coding” (eg an organism doesn’t need prior experience with sex to “learn” to have a sex drive) while still being somewhat modifiable by the results of experience.

A “drive” can also have something of a “momentum”-like effect — it can nudge or bias an agent towards persistence in the direction it started, even when the initial feedback is negative.

This is pretty speculative and I’m not claiming it’s anywhere near the whole picture, but it’s the most plausible guess I could initially come up with.

“Wanting” might add something like “hard-coded priorities”, “momentum”, or “synthetic feedback” to the “liking” system. Intuitively, this is like having some kind of “guardrails” or “scaffolds” to keep you from getting “lost” in the vast underspecified space of possible actions.