Top 10 Life Lessons I Learned from Roguelikes
Let’s say you’ve been playing some Roguelikes. And reading Less Wrong. And thinking about the distributive effects of artificial intelligence, crypto currencies and fintech?
It feels like we can’t escape the discussion of crushing global inequality and the coming wave of unemployment as technology automates everything but emotional labor. So you try to turn up the escapism by playing some video games. But instead, your brain starting making weird connections and suggesting that these games can be good models for thinking about society. How could procedurally generated worlds, Bayesian systems thinking and AI-built dystopias fit together to create a moral compass?
Good thing there’s Medium to write down your crazy thoughts. I’ll go into the rationale further in this essay, but let’s start with the conclusions. Here are my top life lessons from playing and thinking about Roguelike video games:
(1) The world you face is entirely out of your control, hostile, and sometimes downright impossible to succeed in by chance of where and who you are. And that’s in a narrow closed system designed to make people happy. Corollary: take no personal credit for success that is granted by being provided a favorable path.
(2) The person you are — including your attributes and abilities — define how you will travel through the world you are given. Playing skillfully is important for optimizing rewards from a journey, but sometimes you are the wrong person for a particular path.
(3) The information you have about the world is the largest controllable factor in contributing to performance over the long run. If you know how the world works and what risks you can take, your outcomes will be an order of magnitude better than if you are foolish with choices.
(4) Getting better takes consistent practice and learning from the failures of others. Empathy and hard work are a pre-requisite. Corollary: make sure the lessons you learn from others come from a world in which you are operating.
(5) If you only have one shot, on average you will fail. Oops. If everyone only has one shot, their run will be determined primarily by the roll of the dice and not by their earned skillset.
(6) In a closed system with deterministic rules, a machine that can make simulations according to those rules will beat a human agent 100% of the time. This is because simulations are like imagined extra lives. This, also, is why human mental modeling sets us apart from other animals in evolutionary success.
(7) In a system that is open and changing, other people’s information may not be helpful for your run. Everyone’s failure rate is likely to be higher than in a closed, known system.
(8) In a system that is open and changing, an agent (human or machine) must have tools to gather information about the nature of the system prior to risk-taking, and such tools could be experimental actions or research observations. Ask for advice more if you are unsure!
(9) We don’t have time to reason everything from first principles, which implies the need for a skill to tell whether other people are credible on their area of expertise. Thus the value of social proof (e.g., Harvard), being quoted by Forbes, and wearing a lab coat. But people have gotten good at growth hacking the impression of credibility while not having the underlying skill set. This leads to bad odds on risky bets, making decisions based on bad information, like electing the wrong person to be president of the United States.
(10) Success can snowball into an unstoppable Godmode where everything is too easy, and you don’t remember how hard things actually were at the beginning. Corollary: Being given success without going through the process of earning it robs the recipient of “the flow” in that particular field. Corollary 2: Going from rags to riches is the biggest rush.
So let’s dive into how I got there.
Videogames Teach Us About Fairness and Systems
Philosophy and arguments about fairness are boring. They are unintuitive and people don’t believe them. Especially if the listener is in a position of relative power, or is biased in some way against race, gender, ethnicity or income level (which we all are). Or is maybe distracted by Facebook on their phone.
But video games are pretty great! They are just escapist mindless fun right?
Actually, we can learn a lot from video games, and particular video game design. And, I think, we can learn a lot about distributions of social outcomes from the experience of playing a certain subgenre of game, called a Roguelike. For the purposes of this essay, when I mention a video game, this is the only kind to which I am referring.
Here are some Roguelikes, you Monsters
Maybe you’re not a geek like me, and have never heard of any of this. Brief intro then. Before you judge, remember Mary Meeker’s takeaways in the 2017 Internet Trends report was that most of today’s technology entrepreneurship is just recreating video games mechanics and SciFi of the 1980s into business.
Roguelikes are inspired by an ’80s game called Rogue, which set the premise for the genre — and has since been copied, splintered, and remixed. The defining features are that (1) each time you start fresh and if you die, you lose permanently, (2) the world is randomly generated within a set of constraints but different every time, (3) things are hard. You make many “runs” at the world, learning the possible variables, so that you can make better decisions in the future. Because the current run you will most likely fail.
This can be applied to different gameplay. In something like the Binding of Isaac, the player’s moveset is up, down, left, right shooting tears at enemies in a room. You track health, resources (coins, keys, bombs, etc), and a variety of role playing attributes that determine your speed, attach, size and others. Progress is determined both by how you play and the things you find.
In a game like FTL, the player pilots a spaceship escaping a sinister enemy. You manage space battles using the ship’s functional systems, like engines, shield, air, medbay, and crew. The ship can gather resources, be upgraded, and so on. You’ve seen the movies.
In Slay the Spire, the player faces a turn-based battle with enemies that attack a character whose abilities are defined by a set of cards, which you draw at random from a deck. These cards represent actions your character can take, and can be upgraded and enhanced by various experiences and possessions.
In Nuclear Throne, your tiny murderous avatar is engaged in a continuous shooting frenzy with enemies trying to gun you down on a large, randomly generated level. Your resources are life, speed, and most importantly, the weapon you carry.
In Dungeon of the Endless, your team explores rooms on an alien planet searching for power crystals and builds defensive structures against incoming waves of predators. An electric grid underpins how much infrastructure you can create before the team is overrun. A few directly controllable characters can attack, heal and otherwise aid your progress.
I give these examples to show that a Roguelike mechanic (runs through a challenging world) can be applied to different gameplay variations. Meaning, different implemented systems can all be built on the same meta-design. Life, the Roguelike. So here’s what makes the thing tick.
The World in Difficulty Curves
The worlds of these games are procedurally generated. This means that the elements of which they are composed are shuffled and reshuffled each time you enter the world anew each run. There is no fixed slate, just a permutation of the system.
One permutation may be very difficult, with you getting poor odds for how the world is laid out. Access to resources is low, or there is no synergy between the things you do acquire. Another permutation may be very easy, snowballing you into a ball of synergies that makes everything trivial.
This is generally known as “the curve”. Sometimes the one you get sucks.
By playing, you make decisions as to how to optimize the journey through this world generated by an algorithm. This is a skill, and can be improved over time by understanding how the algorithm works.
If we think about reality for a second, it is easy to see that depending on your starting position in the world, the difficulty curve you get can vary wildly. You can start in a physical position in the world that makes life very easy, which has nothing to do with you or your personal attributes. Being born in the United States, rather than being born in North Korea, is one such example. GDP per capita by country combined with an equality measure like the Gini coefficient is a good shortcut for difficulty curves globally.
Another element of the game that can change between runs is You. Think of your character as the DNA base for the progress of your avatar. Maggie in Binding of Isaac has more health, but is slower. The Barbarian in Diablo II has a skillset available uniquely to him. An FTL ship populated by robots does not need an oxygen system. Gork in Dungeon of the Endless is good at fighting but not much else. And so on.
As you progress through these games, you learn about how the talents of your character map onto the procedurally generated world. Maybe you like one style of play over another and start with a different character on your next run. The connection between how the world interacts with your projected person is something you can use as information to succeed.
But you can change your projected person.
Good luck changing who you are in the real world. Being a Muslim immigrant in Trump’s America, or a Bitcoin aficionado in Venezuela, or a Palestinian in the modern day Middle East, or a Jew in early 20th century Europe — all these are pretty bad interactions between your arbitrary environment and the person you are piloting through life.
And that’s not to mention narrower conflicts between artists and capitalism or low income kids and failing high schools.
What you learn about the real world is useful, but you are who you are. You acquired wisdom can never be applied to another run. I cannot restart as an Artificial Intelligence PhD that’s better at high level math, or as a 7 foot tall basketball player. That’s not to say we can’t drift into new careers and open up new paths; but that’s like entering a different game with a new character, rather than re-doing the previous game better. Maybe that’s why we drone on to our children about the world — so they can replay its systems with our information!
Here be Loot
We have the world, and we have our character. How do we make progress?
In nearly all cases, by picking our path through conflict. This conflict is some sort of battle with variously generated enemies, whom we must defeat or by whom we must be defeated.
Getting through conflict provides a reward of in-game currency or resources. That may be gold — which can be picked up, accumulated, and used as money at a vendor. Or it may be the resources (like experience or items) that we need to become better, by upgrading our character’s attributes, skills and equipment. These upgrades are usually randomly generated rewards (thus the gambling-like mechanic) according to a pre-set probability distribution. Some things are more or less rare, and depending on the risks you’ve taken, you are more or less likely to get the good stuff.
To make it obvious, these metaphors come from real life. Gold is money. Resources are education, fitness, mental health, friends, a professional network. In life, we acquire resources in order to do better in the future. It’s just we don’t truly know how well they will serve us. The system in which skills are applied now changes faster than our acquisition of these skills.
Getting to Godmode
If you do get all the good stuff, the rest of the game map becomes easier. You may get ahead of the curve for a while and coast. And if you are powerful enough to do that, you can continue to take all the risks and access the high payoffs.
This is called “Godmode” and makes everything trivial. It’s not a hypothetical. In Binding of Isaac, you may get such a powerful damage upgrade that enemies melt in your laser beam. In FTL, you could find drones that do all the work for you. In Slay the Spire, a certain card combination may let you break the game. In Doom, you literally become invincible.
And so in life. For example, if you start life with American upper middle class parents, have access to great private schools, are a shoe-in at Harvard as a legacy, get your Stanford MBA, find yourself starting a company that raises $50 million, leverage your Ivy League network for clients, and sell the thing for $2 billion by the age of 40 — perhaps your initial roll of the die had something to do with it.
Having to never work again is a Godmode privilege available to a tiny fraction of living humans.
So does failure.
But being granted Godmode early is also a curse. It removes progression, learning and growth and replaces it with a flat, one-note experience. So be wary of large gifts.
Pulling it All Together
In the regular Roguelike progression, you will usually spend the first 10 hours (if not much much more) being absolutely terrible. You won’t know how to use the items you pick up, your character will be upgraded in a way that is useless in the world, enemies will trick and surprise you.
The next 10 hours will be about starting to master the permutations of the world, knowing how things work, how they fit together, and where the synergies lie. You start to understand what does not work. Not unlike real life! Take the often repeated wisdom of what it means to “follow your passion”. People can spend decades following their passion (for e.g., knitting cat hats or writing poetry about insects) and having pretty miserable lives, if capitalism thinks their passion generates no value. But if your passion intersects with a marketable skill with a non-trivial probability, that’s a jackpot. Also, social rewards from being good at something create passion for things you may not have expected. This is a lesson one may want to absorb early on.
Once you do learn the rules of the Roguelike, optimizing a path through a randomly generated map can be a cakewalk. Pick the right character, max out their abilities, and snowball into Godmode. But even that can fail now and again, if the map is unluckily hard. It becomes patently obvious, after spending some time running through these simulations, that we really have to respect the hand people are dealt.
And if things are so dire in systems constructed efficiently to create pleasure, what about systems that are constructed inefficiently with lots of contradictory goals?
Video games are designed by humans, and are closed systems with certain built-in incentives, rewards and mechanics. There are levels, players, equipment, experience curves, controllable and uncontrollable events, and discrete goals. Everything is neatly determined by game designers and developers. Rules are agreed upon, designed, tested, and created for others to experience.
Humanity, with its society and culture, is in many respects the same. It is designed and controlled by humans. The systems have loops and progressions — think educational levels, structured careers, family expectations, life and death. They are designed through politics and media, legislated, and enforced.
Of course there are big big differences. First, life systems are fuzzy and only loosely defined. They are much more likely to break and fail to fulfill their stated goals. A teacher can quit her job and your school, but a non-player character will never abandon her post. Second, they are open and interact with other systems, creating all sorts of reinforcing feedback loops of a larger complex dynamic system. And finally, you only get one shot.
Since games compete for human attention with everything else and cost money (usually), the experience must be compelling enough to keep people engaged. To do this, designers optimize for some version of utility. A destructive way to do this is the gambling effect, which preys on our dopamine receptors (think about the Ding! of your phone’s notifications).
But a more fundamental design approach is to optimize around the state of flow — a state of deep rewarding concentration — for the gamer. This is achieved through a balancing of difficulty and skill acquisition that keeps a person engaged, learning and enjoying their time.
Musicians, runners, writers, stock traders, and other professionals enjoying their work in deep concentration know exactly what this feels like in real life. How nice then to spend time in a well designed system that is optimized around this variable! It is a sure way to make a human satisfied.
So then, what does real life optimize around? The answer isn’t “nothing”, since human societies formed out of evolutionary clay through collaboration. What a person needs is something resembling Maslow’s hierarchy below, which may be incomplete but is directionally helpful. We want systems that generate happinesses in different spheres of our lives in a consistent and predictable way over time.
The human-built systems we make should generally try to achieve these things for members of society, but we know there is a lot of error, because, again, it’s so complicated. The overall thing doesn’t necessarily optimize for the wellbeing of any particular person. Your experience is much more likely to just be an anonymous school of hard knocks, and the earlier you learn this, the better.
All this to say, that a system that must create a happy person in order to exist (a videogame) is going to have some crispness that our unaccountable Kafkaesque nightmare of a reality does not. It is — in a way — the best we can do in designing a system that fulfills a human utility goal.
And the lessons of the simulation are pretty stark.