Black Boxes, Ourselves

An important decision in setting out to write a game is to decide on scope: what do you want to tackle? How broad or narrow should the world of the game be? Do you wish to focus in closely on fine details, or build on a set of abstract concepts?

We sometimes want to model a system very particularly, and to capture many details, which can then interact in all kinds of ways. The level of detail can vivify the world, allowing all kinds of interesting and unexpected events and conditions.

From Atoms, Life

The term “emergent gameplay” captures the idea that the right ingredients, allowed to mix more or less freely, will here and there combine to form new experiences, not imagined by the writers but hopefully adding expressiveness to their ideas.

A relatively early example of this occurred in one of the Creatures games of the mid-1990s, in which the player raised and interacted with beings designed in a “bottom-up” approach to allow more freedom in their ultimate behavior and genetic expression. At one point, developers were apparently surprised to discover that two of these creatures had learned to play catch — throwing and catching had been programmed-in, but the feedback loop of a game of catch had not. It was possible, however, within the systems of the game, and represented exactly the kind of lifelike expression that was originally intended.

A more recent and particularly striking instance is found in almost the entire game of Dwarf Fortress — a building / strategy game in which the player controls an encampment of dwarves hoping to create a great fortress and amass wealth. The game is modeled down to the very physics of bones, dust, and elemental substances, and the gameplay itself layered on far, far above the vast data surrounding how objects fall, fly through the air, melt, and so on — and in between is a rigorous psychological structure determining the responses of animals and persons to their environment, from habit-forming to grief.

As a result, Dwarf Fortress showcases emergent behavior throughout each game — from friendships forming over time between workmates to catastrophic architectural failures, none mandated in a script-like way, but rather coalescing out of the immense detail alive and flowing behind the scenes.

Unknown Unknowns

There is a caveat in building worlds like this: in order to model systems in profound detail, and to enjoy the knock-on effects of that detail, one must understand the systems and actually know what those details should involve.

The thoughtful, freeform game UnReal World, a daily-life adventure in the Finnish Iron Age, thrives on its creators’ vast and personal knowledge of Finnish geography, folk arts, and mythology. The detail that brings the player’s experience to life comes from the weaving-together of actual history, skill, and stories.

If we do not know the density of various metals, or that those densities should be considered when chunks of two different metals slam into each other, or just how resistant bone is to wood, we cannot accurately model the kind of physics Dwarf Fortress thrives upon. Without a clear understanding of DNA and gene expression, the progress of generations in Creatures could not have been convincing enough to engage players for the time needed to nurture and influence those generations. We can simplify or condense a system to tell a story or express certain behaviors, but we must understand it to build it.

From the Outside In

And so we come to systems that are intrinsically difficult to model, for the reason that we cannot easily observe their mechanisms. One of my great interests in game design is the modeling of inner experience. The role-playing genre has provided a vast and intriguing array of models representing strengths, weaknesses, the progress of ability over time, the loss and recuperation of health, and the way in which people benefit by studying one specialty or taking up many hobbies. We have modeled friendship and enmity, and have begun exploring the web of relationships that forms as persons across a region come to know and engage with each other.

But thus far we have not looked too far inward, to observe what happens when the outside world (from shopping to dragon-slaying) affects the inner world (of preference, familiarity, and so on). To do so, we would need to find a good way to model that inner world effectively, and then to be able to observe it again as an expression in behavior.

Dwarf Fortress has carried this quite a long way via a beautifully-detailed mechanistic psychological approach. Beings in the game can seek out interests, form new aversions, and befriend each other. One can actually spend time simply reading various dwarves’ psychological and relational profiles as the game progresses, and know that these profiles are being built dynamically upon a minutely-drawn sketch of positive and negative mental and emotional factors.

What I am proposing as an area of study is something more abstract and yet familiar. In our own lives, we can observe an interesting cycle: we prefer to do or avoid certain things; these preferences and our experiences lead to certain states of mind, which then inform our future choices and evolving preferences. Can we model this organic and yet intangible cycle in the world of a game? Would it be useful to do so, if we could?

The Mystery of No Mystery

We have learned some interesting things about ourselves over the past handful of decades, where our choices and feelings are concerned:

When we begin to peer more closely at what we might call our deliberate, considered engagement in our daily activities and experiences, we find that there may be little substance there after all. Not that we are mere machines, but that behind the curtain of our inner identities, there may be no solid structure after all.

We have discovered that when we think and remember, new sets of neural connections form, comprising multidimensional structures before they vanish again — like castles of sand built and washed away.

Inside the black boxes of ourselves, we may find a kind of living emptiness — constant change fenced in here and there by lines of rigid habit. No tangible structure of our preferences and beliefs, no card-file bank of memories, no visible projection of conscious will into the sea of sensations and experiences we move through every moment of our lives.

We might see ourselves not as definite figures, but instead as a collection of rules and stories, some fixed and some ever-changing, which inform what we feel and what we decide at any given moment.

And we can model rules, and write stories.

A Box of Patterns

If there is not necessarily a fixed structure within us, maybe we don’t need to model one to see ourselves reflected.

What if we only need to recreate the patterns we can easily see, and make some inferences about them?

For example, I might want to model the feelings an adventurer has after thirty days of swinging a sword-arm and collecting the teeth of disappointingly small fire-breathing lizards. Many games involve the “grind” of fighting easy foes for long periods, and I would like to examine what the hero actually thinks of this … and what they might do about it.

Feelings are hard to define at a kind of atomic level, but current research suggests that we may not need to do this anyway — it might even (at our present level of knowledge) be impossible. So what are some things we know about feelings from our own ordinary lives?

  • We often have a predisposition towards one or another: one person might more often feel excited, or bored.
  • Feelings gain some inertia with time. If we have been feeling disappointed for a long time, it might take more to cheer us up.
  • Our feelings are strongly informed by how our preferences meet our experiences. If we are doing something we like on an ongoing basis, we feel happier.
  • Our feelings are also informed by obstacles or a lack thereof. We often like to be challenged, but not too much.

So let us return to our adventurer. Perhaps this person is a cheery sort, and really does like sword-swinging. But on the other hand: thirty days of lizards.

Perhaps we can combine the list of ideas above with game values that are commonly modeled (skill and difficulty) to draw inferences, or derive a useful model of emotion over time.

  1. Our adventurer has a high sword-swinging skill — let’s assume for now that they became so good because they do enjoy this activity. (Of course, it’s possible to be very good at something you hate, and we could easily note “preferred/disliked skills” too).
  2. The small fire-breathing lizards are very easy to dispatch.

For this example, let’s propose a matrix: skilled vs. unskilled activities, and whether one is currently facing simple or difficult challenges with them. At the intersection will be a probable feeling to experience.

— — — — — —— — Easy | Hard
Skilled ……..…….. Bored | Contented
Unskilled.. ……….. Frustrated | Excited

In our simple model, our adventurer is doing something very simple using an ability they are already quite proficient at — they are likely to be rather bored.

Since we are dealing with a mechanical model we can express in math or code, we can easily add other dimensions (like whether an activity is liked or whether the person is currently failing at it, difficulty aside) and multiply our set of emotions dramatically with the same kind of rules.

However, let’s leave it as is for now and see what this simple model might actually be good for.

Agency Soup

We have only added one value here — current feeling. Within the parameters of an adventurer tackling various adventure-type tasks of varying difficulty and getting better at various abilities, we can now say how they might feel about it.

If we choose to look at the degree to which people are affected by routine and the way they enforce their preferences, we can use our new characteristic to:

  • Track feelings over time (“has been bored lately”) and allow a “tipping point” for sudden change, as well as a steady regression to the “default” or most prevalent state.
  • Represent the effect of mood on learning — increasing skills faster when enjoyably challenged, more slowly when bored or frustrated.
  • In situations where one or another activity might be chosen, influence decisions in favor of tasks that will result in positive feelings.
  • During long-term discontent, a character might begin to refuse certain choices or act more unpredictably in order to shake things up and have more pleasant (or at least interesting) experiences.
  • Alter interpersonal reactions according to mood, or the relationship between two parties’ moods.

Thus with a minor adjustment to mechanics (factoring together skill and task difficulty), we can straightaway begin to model a system in which characters can begin to act with greatly increased agency — even though we have not modeled the mind or self very deeply at all.

Not only can we use this to strengthen the relationship between a character and their environment, but also look more deeply into the effect of game situations like “the grind” — might a character refuse to continue, or seek out dangerous challenges early?

If we can easily model a prototype for preference, we can also revisit the relationship between a player and a character. What about an adventurer the player does not control at all, but can only influence? Or a “zero-player” game made up entirely of an adventurer’s diary? (“Too many lizards. Going home.”)

From the Inside Out

If we are constructed not just of neurons but of narratives, then we can leverage our understanding of narratives (which we may find simpler to organize) in order to build out the cycles of change by which we permit the world to inform us and which we, in turn, use to influence our world.

In these interactions we may also find worthy experiments in finding out whether our systems and those of the world can reflect each other in constructive and meaningful ways. What happens when a town — or a mountain — gets bored? Indeed, is that what brings the dragons down?

The above is only a sketch of this idea, and the demonstration is a simple one — but I hope it serves to suggest that we can use the narratives we discover in our own lives in order to evoke entirely new dimensions of experience in the world of games.

Originally published at on November 29, 2018.