2018 Selected Workgroup Topics2018 Workgroup Topic Proposals

Constructing Emergence

Emergent gameplay holds a lot of fascination for many people, and yet we have a difficult time nailing down what it is — much less how to reliably create it. There are a lot of descriptions and attempts at definitions of emergence, but they’re often incompatible and fairly ad hoc in nature. As a quick and dirty definition (with thanks to John Holland) I’ll propose that something is an emergent effect when it creates a new “thing,” a new object based on the interactions of multiple parts, but without being dependent on any of them.

A simple example of this is the “glider” in Conway’s Life: it can be described by the way individual cells turn on and off, but is more compactly described one organizational level up from that as a thing in itself that moves on its own.

There are a lot of other examples too of course: Nicky Case’s fireflies is one of my favorites.

As systemic/procedural/rogue-like gameplay gains popularity, creating emergent effects reliably is becoming more important, and I believe is an important part of the game designer’s toolkit.

I’ve been working on this problem a fair amount, and would love to chew on this topic with others (I may well be lost in the Dark Forest). What I’d like to get to is a minimum set of requirements for reliably constructing emergence but without constraining the emergent result itself: “do these things and you’ll get emergent behavior/gameplay.” And, just as important, some examples of how that actually works in game-like situations.

What I have gotten to currently is that to construct emergence you need objects with internal state and behaviors at affect other objects (i.e., nouns and verbs). To further set the stage, the verbs must be:

  • Local: operating only on “nearby” objects, typically spatially but also in terms of level of organization. A single bird changing direction doesn’t on its own immediately change the flock, and a single cell in Conway’s Life doesn’t determine the state of an entire region.
  • Narrow: they have one or at most a few very well-defined, clear effects. They affect only one or two attributes in another object, not many all at once.
  • Modular: they have well-defined bounds and unambiguous effects.
  • Generic: they operate the same way no matter what, without special exceptions or needing to know the context in which they’re operating.
  • Hierarchical: going back to “local,” emergence appears at different levels of organization. At each level, the emergent effect creates a new object “one level up,” which can then interact with other objects at that level.

The examples I’ve poked at — Conway’s Life, flocking, firefly displays, slimes in Slime Rancher, etc. — all fit these requirements. Are these necessary? Sufficient? Does this help us understand and construct emergence? Are there aspects missing? Is this worth looking at for creating engaging gameplay? (I believe so, as it helps reduce reliance on the content treadmill but… maybe that’s an illusion?)

I’d love to get a group of like-minded people to dig into this, or at the very least all end up looking as baffled as I sometimes feel.