Choose What You Want

The following column has been written by 3QD reader Ian McMeans, a game programmer by profession, who won the CPCP challenge that I posted last week. Thanks much, Ian.

How could we predict what a person will do next? We could assume their goal is to maximize their happiness, (a philosopher or economist would more precisely say maximize their utility), then predict that their next choice will do so. This is a common assumption in the real world, one that we make when interacting with other people, and especially in setting social policy: We argue people should be free to do as they like because their chosen actions must be the best for themselves. We like democracy because we assume the electorate will pick the candidates who will maximize their utility, and similarly we like free markets because they are driven to efficiency by our self-interested, utility-maximizing behaviour. If we want what’s best for people, we let them choose it.

Economics until the mid-20th century was developed under this assumption, by treating involved actors as Homo economicus: perfectly rational and self-interested. (What’s interesting to me is that even with perfectly rational behaviour you can find surprising strategies that maximize utility in subtle ways). But what about real humans? We’re nothing like our rational approximation – real humans are a giant mess. Not only are we unable to predict what makes us happy, but we’re unable to correctly make the choices that would achieve those goals, failing twice. It’s like we aim at the wrong target and miss it. We regret our worst decisions while we make them, procrastinate, hope to get rich with the lottery, and have our opinions swayed by content-free advertisements. This is not rational (rational behaviour is utility-maximization, by definition).

Why should people’s suboptimal utility-maximization matter to us, aside from the accuracy of economics? There are people who believe (perhaps I’m one of them) that utility is the only goal worth pursuing, as a moral prerogative. Not just for yourself, but for every creature that can perceive utility and benefit from your actions. Utilitarianism is simple in concept (morality is maximizing happiness over all people and all time), and as a bonus it lets you use phrases like “Felicific Calculus”.

Since Utilitarianism is now our moral goal as of writing that last paragraph, how could we achieve our goal and get more utility for everyone? There are a few different approaches we could take, here’s one:

1) Make more of the things that give us utility.

Technology and our ever-increasing standard of life do a large part to improving our happiness. But why would this be the case – what determines which objects and activities make us happier? There is a teleological argument here from evolution, where the things that make us happy are the ones that made our ancestors flourish: good food, sex, signifiers of social status, curiosity satisfaction, avoiding pain. The mechanism behind this is evident: if we are happiness-seeking agents, then evolution will slowly pressure our species towards being happily rewarded for those behaviours that are well-correlated to genetically flourishing.

Imagine that natural selection gradually put together a utility scorecard, by trial and error it’s the one that makes people the best reproducers:

Action:

Reward: (in Utils, our unit of happiness)

Eat something sweet

5

Talk to a pretty stranger

15

Eat something bitter

-5

All the cavemen who thought poison mushrooms were delicious died, and didn’t pass on their ‘scorecard’. Our ancestors, (who sought out sweet apples because they carried the genes to enjoy them) passed on their sugar-enjoying genes to us. Fitness-maximization over the course of many generations becomes utility-maximization for each individual, and our utility is our ancestors’ measure of how much each action improves fitness.

There is a new strategy open to us for chasing utility: if we want to be happy (without going to the trouble of satiating the traditional evolution-driven goals), we can simply trick our evolution-tuned brains into thinking we’re doing a good job when we’re not. We can trigger those evolutionary fitness-detectors without doing all the work of producing what the detector was designed to originally detect. For example, instead of real social contact, we can enjoy drama in movies. Instead of sugar, we can cook with nutrasweet. We can reward ourselves with books, videogames, and drugs. Is this cheating, or somehow misguided? I don’t think so, because our goal is happiness. Although we were ‘designed’ by evolution to pursue certain goals, why should we care about those goals? We can hijack those reward mechanisms for our own selfish ends. All we care about is the sensation of sweetness, not that sugar is available for digestion and its effects on our differential reproduction. Evolution hasn’t had time to build taste sensors that can differentiate between sugar and nutrasweet, and we can exploit that fact for our calorie-free enjoyment.

It seems the trick here was to follow the chain of causation, skipping the sweet apple and going straight to the sweetness-detector (the tongue) and interacting with it, to get the reward of sweetness. In fact, we can go further upstream with the same goal of avoiding unnecessary indirectness: we can travel along the nerves up to the brain. Why bother mucking about with nutrasweet and tongues? Instead of making things that give us utility indirectly, why not

2) Make more utility, directly.

This certainly seems promising, if a little metaphysically odd. What exactly is utility, and how would you go about manufacturing something as intangible as happiness? It’s not exactly like manufacturing cupcakes, (although that’s close). Wikipedia says:

In economics, utility is a measure of the relative happiness or satisfaction (gratification) gained by consuming different bundles of goods and services.

That seems a bit materialistic. Do we really only care about happiness from goods and services? We could certainly make people happier by manufacturing consumer goods and giving them away, but the economy is already hard at work on this. In fact, the economy is as efficient as it can get, barring the aforementioned lack of rationality on the part of participants. The Earth’s resources are already allocated towards our happiness.

What else could make people happy if not the effects of goods and services? What is happiness? As enlightened (philosophical) physicalists, we believe it must be reducible to some state of the brain – happiness is not stored in the mental res cogitens of an ephemeral soul hovering above us, but in the meat in our heads in electrical impulses, chemical signalling, or some other neurological mechanism. Luckily, there are people who have already figured out parts of it, and we can manipulate it, albeit crudely:

In the 1950s, Olds and Milner implanted electrodes into rat Nucleus accumbens and found that that the rat chose to press a lever which stimulated it. It continued to prefer this even over stopping to eat or drink. This suggests that the area is the ‘pleasure center’ of the brain.

(An aside: Have you ever wondered why your own emotional state is autonomous? If we had control over our own mental states, we would just choose to be happy all the time, regardless of external circumstances. Any ancient humans with that ability would have blissfully starved to death, and not passed on their genes. The reward mechanism is only indirectly accessible to us, so to mediate our own happiness we’re forced to pursue the goal of evolutionary fitness. Supposedly with meditation you can learn to maintain a state of happiness without cause, but this must have been rare or subtle enough in our ancestors to not affect how many children they raised. The wirehead rat was given control over its own mental state, and it acted in a perfectly utility-maximizing way.)

“If it was possible to become free of negative emotions by a riskless implantation of an electrode – without impairing intelligence and the critical mind – I would be the first patient.”
— Dalai Lama (Society for Neuroscience Congress, Nov. 2005)

It’s not hard to imagine a near future where this can be done with more precision, and to humans. Would this really make us happy? There’s an argument here that utility derived from these strange unnatural sources (like the wire going into the rat’s brain) is somehow illegitimate, that it’s not “true happiness” unless it’s earned in the difficult ways we eke out our own happiness from the evolution-sanctioned sources. I strongly disagree with this – as far as your brain is concerned, the way signals get into it are irrelevant, they’re all the same. Happiness is happiness. Why does the possibility of manipulating our own mental state like this make us feel so uncomfortable? People are afraid of having their preferences changed and losing those goals, like with the story of the lotus eaters, or Huxley’s Soma, or even the wirehead rat. The vision of being stupored and ignoring the things you currently care about is terrifying.

So where did we go wrong with this thought experiment? If wireheading is so objectionable, and we want to maximize happiness for Utilitarianism, how can we reconcile them? I think the solution is that wireheading doesn’t lead to an increase is happiness. If it incapacitates you with joy, then you won’t help other people, and (like the wirehead rat) you might end up ignoring the future effects of your actions, like a junkie. Trading a few days of starved wireheading isn’t worth losing a lifetime of milder joy and acting as a moral agent.

Could it be done more carefully? Instead of junkie-like wirehead dependence, what about just making everyone merely happy? It happens naturally to some people, could we trigger it artificially in everyone? Imagine your utility scorecard becomes this:

Action:

Reward: (in Utils, our unit of happiness)

Eat something sweet

50

Talk to a pretty stranger

65

Eat something bitter

45

You aren’t in a wirehead daze, you’re just having a lot of fun doing ordinary things.

It seems far-fetched to discuss this so soon before the technology exists, but what could matter more than cessation of human (and animal) suffering? There is a lot we can do even with current technology (and without drugs) to make creatures less fortunate than ourselves happier across the world, but eventually it will be technologically feasible to make people happier by modifying the people, rather than modifying their environment. Remember our end goal is happiness, not satisfying the arbitrary cues natural selection has implanted in us. Instead of trying to get what we want, we can choose what to want.

This poses a problem for our conception of rationality, though. What does it mean to be a utility-maximizer who can change the rules of the game, and assign utility to actions at whim? It would change people’s behaviour in unpredictable ways, by letting them add incentive to tasks they wish they had more reason to do. Is this something we want in society? It’s conceivable that it could sow chaos: How many parents could resist the temptation of making their children enjoy exercise and diligent learning, and not enjoy fatty foods or gambling? How many dictators could resist the temptation to pacify a population? How many smokers and gamblers would choose to hate their old habits? What would you choose to want?

In terms of evolutionary teleology, we could choose to redirect the reward mechanism that guided our ancestors to flourish. This might not be a bad thing in terms of our species’ survival, because our preferences are already out of synch with what benefits us (A caveman who loved the taste of sugar would do well, but too much sugar harms modern man and leads to long-term disutility. Man hasn’t evolved distaste for too much sugar fast enough). We have the option of adapting people’s utility functions to the modern (and future) environment, to keep pace with technology without waiting for evolution to catch up and tweak us. This could be a great boon, or (if mishandled) a huge disaster. It seems like this is a risk we get with advanced enough technology – once we start intentionally modifying ourselves, things can change very very quickly (faster than cultural evolution has been driving us, because cultural evolution operates in the constraints of biology), and we take future change into our own hands. (Is this yet another possible answer to the Fermi paradox? All the aliens clever enough to make interstellar spaceships don’t live past the self-modification phase).

Non-sequitor 1: Why doesn’t runaway sexual selection eventually break itself? Peahens who make their selections based on other indicators of health (rather than plumage) would beget children who didn’t need to maintain expensive plumage, which is an advantage. Isn’t the evolutionary pressure to select for the traits that natural selection prefers, and sexual selection should track those good traits? It obviously doesn’t, why not? It seems to be circular logic that peahens are optimizing their childrens’ chances of being sexually-selected for the trait, because any trait could get runaway selection in that case.

Non-sequitor 2: Is there a noticeable difference in the effects of (lack of) advertising on Tivo users?