Gaming in a Skinner Box: Roguelikes, Slot Machines, and Behavioral Psychology
Exploring the reasons why it’s hard to put down roguelike games
If you peek into my Steam library, you would find a good bit of roguelikes + lites. It’s hilarious because I’m terrible at them! I’m afraid to share how many hours I’ve put into games like The Binding of Isaac, Risk of Rain, Enter the Gungeon, and Hades compared to my low percentage of achievements and other measurable progress. Despite it all, there is something about roguelikes that keeps me coming back to them again and again and again.
These roguelike games carry with them a unique “attitude” that I find appealing, such as their combination of great artwork, mechanics, characters, and indie backgrounds. I, also, appreciate their mechanics and high levels of difficulty as compared to AAA titles that can sometimes feel more like interactive movies than games to me.
Finally, it’s because they hook me using the same elements of behavioral psychology that keep casinos in the deep black.
What is a Roguelike?
Before delving into the behavioral psychology behind these games, it’s important to work up a quick definition for them. In short, roguelikes are games that hearken back to a title from the 1980s called Rogue by leaning heavily on the following mechanics:
- Dungeon-crawling + Hack ‘n’ Slash + RPG
The player is trying to 1) escape a dungeon, comprising a series of interlocking rooms, corridors, and floors, by 2) killing a lot of monsters standing between them and the exit which is made possible by 3) leveling up character stats and abilities via experience points and item acquisition.
- Procedural generation of environments & randomization of contents There is a variety of challenges and environments in this genre. Level layouts are randomized but curated (there are some elements that are not random and impact the levels presented), while enemies, items, and rewards can be either curated or totally random. For challenges and rewards, they’re unique to each playthrough.
There are no checkpoints and no extra lives — once you die, you have to start over with a new randomly generated playthrough.
The roguelike purist might turn up their nose at the above definition. Having played “true” ASCII-graphic roguelikes such as NetHack, Cogmind, and even Rogue itself, I understand.
But when looking at the genre as an evolutionary process, and considering the many branches and creative interpretations of what a roguelike can be, the above three elements are the strongest and most common features of these games.
In no way do I want to forget the roguelite subgenre. I define the *lites as following the rough blueprint of the *likes, but with some level of player progression baked in — such as through the cross-run item unlocks or ability leveling available in games like Dead Cells and Hades.
But regardless of *like or *lite, or the opinion of the purist or the casual fan, there are two common elements of these games — procedural generation and permadeath — that can be examined through the lens of behavioral psychology, specifically through the study of operant conditioning.
What is Operant Conditioning?
To understand what is operant conditioning, we need to examine the work of one of the most influential names in behavioral psychology: B. F. Skinner.
Skinner’s research involved shaping various behaviors in animals through the use of rewards and punishments — also known as operants — to encourage (reinforce) desired behaviors and to discourage undesired behaviors.
One of the main ways he studied how operants can encourage behavior was through the use of a device called a Skinner box — a common version of which involved a light or a speaker (to act as a behavioral trigger), a button or bar (the pressing of which is the desired behavior), a food dispenser (used to provide rewards for said behavior), and a floor that could be electrified (used to provide punishments for behavior).
Behavior can be shaped in various ways via a Skinner box.
For instance, a rat placed in the box can be rewarded with a food pellet by pressing the bar — a process called positive reinforcement. Or, the rat could be taught to stay away from the bar through the use of punishment (i.e. electrifying the floor if the rat presses the bar).
Skinner also used this device to explore the power of negative reinforcement. Negative reinforcement can be the animal pressing a button to stop the shock being delivered to them. The reward associated with pressing the button is being removed from this unpleasant stimulus — stopping the shock.
What Is the Strongest Way to Instill Behavior?
So given a device like a Skinner box, what ends up being the “best” way to cement the desired behavior in a subject?
In order to figure that out, there are two things to look at when determining how well a behavior has been learned:
- Response Rate: How often the subject performs the desired behavior — the more times the subject performs the behavior, the better it has been reinforced.
- Extinction Rate: How quickly the desired behavior dissipates once the reward/punishment is no longer paired with the behavior — the longer the behavior persists, the better it has been reinforced.
Skinner found that positive reinforcement far outpaces the use of punishment or negative reinforcement for cementing new behavior. But what he also found is that a variable ratio schedule of positive reinforcement leads to the strongest instilling of a new behavior:
- A ratio schedule of reinforcement means that the behavior (e.g. lever-pulling) must be performed a certain number of times before it is rewarded — as opposed to an interval schedule, in which a certain amount of time must pass before the behavior can be rewarded again.
- A variable schedule of reinforcement means that the ratio or interval changes between each reward — whereas a fixed schedule means that the number of behaviors to be performed or the amount of time that passes is the same between each reward.
Skinner found variable ratio schedules of reinforcement yield both the highest response rates and the slowest extinction rates for behavior when compared to fixed ratio, variable interval, and fixed-interval schedules. So a rat, pressing a bar to receive food, would have to press the bar an unknown (to them) number of times before being rewarded — and even if the food stopped coming, they would continue to press the bar in the hopes of a reward for quite a while before giving up on it.
What Does This Have to Do With Roguelikes?
It turns out that operant conditioning — specifically the high-response low-extinction variable ratio schedule of reward — has a direct and familiar representation in the real world: slot machines.
Slot machines are Skinner boxes for humans. The randomized results act as a variable ratio schedule of reinforcement for gamblers. In casinos, the formula of keeping people glued to their slot machine has been perfected. The high response — insert quarter, press lever — and low extinctions — players playing for hours thanks to small to medium-sized wins — demonstrates operant conditioning at work. That they may eventually hit that big jackpot and the occasional small rewards keeps people motivated to pull the lever again.
Roguelikes end up working in a similar, though less sinister, way.
The randomization that occurs as part of each run’s procedural generation operates on the same type of variable ratio schedule that can be found in Skinner boxes and casinos alike.
Think of The Binding of Isaac, where a single, randomly generated item on the right floor can make the difference between a failed run, a marginally successful one, or a completely overpowered, off-the-walls super-run. Or Hades, where the right series of randomly generated Boons, Hammers, and Poms can synergistically catapult you right out of Hell.
The high difficulty ensures quick deaths for the unskilled or the underpowered — and quick deaths are likely to lead to another pull of the lever, another attempt at a playthrough hoping to hit the jackpot: the right combination of items, pickups, enemies, and layouts that lead to a win.
There are big differences between slot machines and roguelikes. After all, slot machines are built to extract as much money from the players as possible, while the latter is (hopefully) optimized to extract the most amount of fun out of players. But the underlying principles — behavioral psychology, operant conditioning, and Skinner boxes — are the same.
It’s an important thing to keep in mind, especially in those cases where gaming lends itself towards addiction rather than fun. Being aware of the psychology behind roguelikes can keep you focused on maintaining a level of fun, while also keeping you away from the 3 AM, zombified, “just one more run” ruts that are all too easy to fall into.
Gaming should be fun, but being a rat in a box does not fall into that definition.