Αʹ
My wife isn’t really sure how she feels about coconut water. She doesn’t hate it, but there’s an odd quality to it. It leaves a strange taste in her mouth — not an unpleasant one, but not an entirely pleasant one either.
She does like lots of other beverages, all of which are widely available next to the coconut water: lemonade, smoothies, juice; these are things she knows she enjoys.
And yet, more often than not, she chooses coconut water anyway. Why is this?
If humans are simple pleasure-optimising machines, this is an entirely illogical decision. She would derive more pleasure from a nice refreshing lemonade than from a carton of coconut water. Maybe she’s optimising for long-term utility by choosing a healthier option, but I think it’s hard to argue coconut water is clearly the healthiest choice from a fridge that also contains smoothies, fresh fruit juice and plain ol’ bottled water.1
Perhaps my wife just makes irrational beverage choices, then. But I think the curious case of the coconut water points to something more broadly applicable to the human experience.
For example, consider films. If you were a simple pleasure-optimising automaton, an overtrained AI seeking only to maximise reward, you’d choose something you found extremely pleasurable to watch — maybe a rollicking comedy, or perhaps porn. You certainly wouldn’t choose a graphic horror film, a complex piece of French arthouse cinema, or a harrowing true crime documentary.
In fact, why would you even waste your time watching a film at all? After all, you could be doing much more pleasure-maximising things with your time. If you care mostly about short-term pleasure, you could be having sex, gorging yourself on chocolate cake, or injecting recreational drugs; if you take a longer view, you might prefer to study a new skill to increase your future earnings potential, invest all your time in finding a long-term partner, or perfectly optimise your exercise routine. Whatever your time horizon, you wouldn’t be staring at pictures on a screen. And yet the screens are the things we can’t tear ourselves away from.
Maybe humans are just really bad pleasure-optimising machines. But if we’ve missed the target variable so badly, what’s the poorly correlated metric we’ve ended up optimising for instead?
I think we used to optimise for pleasure — not in our evolutionary past, but in our childhoods. After all, who pursues pleasure more effectively than a child? But as we age, we start optimising for something else. Not something worse, just something different.
Βʹ
Let’s return to taste in the most literal sense. You’re sat in a restaurant with an extensive menu, trying to choose your main course. The unaccompanied child at the next table, as children are wont to do, ignores all the chef’s carefully crafted dishes and orders the largest available bowl of ice cream, or maybe an entire cake. If pushed to choose a savoury dish, maybe they ask for a huge block of cheese, or maybe they just want chips.
Children are experts at choosing their favourite thing and sticking to it. This is why so many children are picky eaters who will eat nothing but chips and ketchup. It’s why they watch a really good episode of Power Rangers and suddenly want to watch nothing else. It’s why they can spend all day playing in the same play area and still not want to go home. This is what optimising for pleasure looks like.
But I’m willing to wager that you — if you are not yourself a child — wouldn’t choose ice cream for dinner, that you’d rather watch Power Rangers in moderation, and that a play area would start to bore you after a few goes on the slide. That’s because you’re optimising for something else.
You’re optimising for complexity.
Ice cream tastes sweet, and has a single flavour and a single texture. If you like sweet food, that flavour, and a creamy texture, you have all the ingredients for a perfectly pleasurable meal.2 Chocolate is the same; so are chips, ketchup and buttercream icing and all the other things children love. As a child, I used to think a whole sugar cube was the height of cuisine.
But it’s nothing like your grown-up meal, with its carefully considered mix of herbs, range of ingredients, and sprinkling of truffle oil.3 There’s probably an ingredient you like a bit more than the rest — let's say you're a big mushroom fan. But if the whole dish were replaced with mushrooms, you'd probably enjoy it less. Similarly, there might be an ingredient you don't like as much — I happen to dislike tomatoes — but if you take it away, the dish might be poorer for it. Not everything on your plate is your favourite food, but that's almost the point.
It’s this complexity of flavour that draws the adult palate. This is why you graduate from sweet, sugary beverages like Coke and Ribena to coffee, wine and (of course) coconut water, all of which — let’s be honest — aren’t actually that nice.4
Γʹ
Scary films and grisly true-crime podcasts work the same way. They don’t make you feel pleasant emotions — maybe relief at times, or curiosity, or a boost of adrenaline, but these aren’t the prevailing experience.
Watching something purely pleasurable would look something like this — a flat line:
But watching something tense and scary looks more like this, with troughs of fear preceding peaks of relief and adrenaline:
So you enjoy the experience less overall, but the texture of the experience is more complex, and I think this complexity is what your reward function is really optimising for.
Music works the same way. Children’s songs have a cheerful, catchy tune, simple lyrics, and — well, not much else. It’s fun to listen to and makes you feel happy, but no adult voluntarily listens to “Wheels on the Bus” of their own accord. Adults want layered sound, complex lyrics, and a weird, strangely compelling bridge.
This also explains why you’d bother watching TV or listening to music in the first place. You want to optimise for short-term pleasure sometimes, by having sex or eating cake, and for long-term pleasure other times, by working on your career or looking for love. But there are also plenty of times you want to do something that isn’t perfectly aligned with either goal. You do this to give your experience some texture.
So maybe this is about variety, but I don’t think that’s quite right. Variety is a component of what we’re talking about, but it’s not the whole thing. After all, coconut water is a single experience that can’t really be broken down into component parts. It’s not variety as much as it’s the experience of not being quite sure whether you’re having a good time. Variety is the spice of life, but an appealing meal needs more than just spices. Complexity is the name of the game.
Δʹ
And who says this desire for increasing complexity has to stop in adulthood?
I have a very technical sort of job, and tend to spend my days focused intently on a single task. The work is fun and intellectually stimulating, but uses a relatively narrow set of skills; let’s call it widget-making. My line manager makes widgets too, but also has to devote time to managing me. Their line manager also makes a few widgets every now and then, but mostly manages a larger team and has to go to a lot of meetings about the team’s strategic direction. Their line manager is pretty much detached from widget-making, but has to manage a lot of people and make a lot of big long-term plans…
Now, I look at them and think their jobs seem worse than mine. I get to spend all day doing fun widget-making, while they go to a lot of boring meetings and spend a lot of time thinking about budgets and filling in forms. But they all seem pretty fulfilled, and would probably tell you they enjoy their jobs now more than they did when they were lowly widget-makers. What’s going on here?
Obviously the money helps, but let’s take them at their word and assume they really do prefer their jobs. To an extent, we can explain this with variety. I make widgets. My boss makes widgets and manages people. Their boss makes widgets, manages people, and makes plans… But their boss doesn’t make any widgets at all — they only manage people and make plans, yet they seem the most fulfilled of all.
The key is that these tasks are more complex than widget-making. That doesn’t mean widget-making is easy, but making a widget is a rule-based process you can reason about and only uses a relatively narrow set of skills. Compare that to the demands of managing an actual person with feelings and needs and annoying human unpredictability. And compare that to managing a whole organisation made up of lots of unpredictable people with their own feelings and needs, who might not be making as many widgets as the customers need…
The skills you need as you climb the corporate ladder become less technical, more social, and harder to pin down, and that's exactly why the people doing those jobs were willing to leave widget-making behind. It's also the reason career trajectories invariably move in that direction. Early in your career, you want a job you enjoy; eventually, when that gets stale, you move into management.
Management isn’t as fun as making widgets, but it’s delightfully complicated, and I’d wager the skills involved light up a lot more of the brain on an fMRI than widget-making does — and that the complex taste of coconut water lights up more of the brain than the simple sweetness of ice cream does.5
Ϛʹ
Maybe the brain gets used to a certain type of stimulation over time — be it the taste of chocolate or the challenge of widget-making — and needs increasingly more stimulating experiences to feel fulfilled. Maybe if you could just keep cranking up the pleasure dial, you’d stay fulfilled forever; but for that to work, you’d need to start off eating nothing but Brussels sprouts6 and gradually introduce more pleasurable food until you finally gain access to chocolate towards the very end of your life. This doesn’t sound like a utility-maximising world.
Alternatively, instead of stimulating the brain harder, we can choose to stimulate it differently. What you crave is new patterns of stimulation, patterns that light up parts of the brain that aren’t used to working together — a variety of complex experiences. Salted caramel works because the “salt” and “caramel” neurons don’t normally fire together; coconut water works because the nutty, slightly sour flavour profile isn’t typical of a beverage.
And it doesn’t really matter whether you like it. It feels good anyway.
It might be the healthiest choice, but I don’t think it’s clearly the healthiest choice by any wide margin.
Even the more interesting ice cream flavours are much maligned by the average child — I don’t think many of our younger brethren tend to opt for pistachio or rum & raisin.
Does anyone even like truffle? I don’t think I do, yet I keep ordering it anyway. If only that effect had a name…
Wine obviously has the alcohol thing, but people continue to choose wine in situations where getting drunk would be undesirable and alcohol-free wines sell reasonably well, so at least some people must be choosing it for the flavour.
Unfortunately, I have no idea how to interpret any of the studies that seem like they might address this hypothesis. It’s ultimately inconsequential to the overall argument, but would be a nice mechanistic explanation, so please do share anything you know in the comments.
I actually think Brussels sprouts are quite nice and don’t deserve the reputation they have, but they serve here as a placeholder for whatever food you happen to dislike most.
I'm not sure if some of the people in your examples really optimise for complexity, as opposed to signalling the ability to handle complexity?
Some of your examples sound like they have alternative explanations, related to humans-as-social-animals.
Thank you for the post, it was an interesting read :)