what are players learning from your systems? avoiding deceptive design
a conversation with a friend following a previous post led to a few thoughts regarding “designing for delusion” (and the ethics thereof).
I was prompted to reconsider Noah Wardrip-Fruin’s “sim city effect”: black-boxing your game’s core system (or systems) in order to promote a player experience where the central challenge is to iteratively fashion a mental model of this system. in other words: players, when confronted by your interactive black box, are led to develop and test hypotheses about its internal behavior, a cyclic and convergent procedure which I would certainly interpret as an authentic form of learning.
this is all pretty groovy. one mistake in practice, however, is constructing an empty black box. the problem with an empty black box is that its contents are impossible for the player to reverse-engineer (read: learn) because… there’s nothing inside!
this is probably best exemplified by the well-known phenomenon where players are found to “discover” spurious patterns in noise, or relate otherwise uncorrelated events to one another by assigning meaning to them (note that noise-based generative techniques do not fall under this heading). this process is akin to making hypotheses against an “empty” black box with no hope of proof, unless of course the hypothesis predicts random behavior. outside of examples where this effect is intentional (say, a game about gambling bias), in general, it is not a replacement for real learning and should not be encouraged. alarmingly, this is sometimes touted as a successful result of good design; this is, however, another example of an abuse of the phrase “let the player fill in the gaps”.
sometimes a black box is not quite empty, but can effectively be reduced to one that is, or can be seen as an empty black box in context of the overall system. let’s consider an example.
imagine a city builder, one where the player builds cities that are quite large, perhaps with thousands or millions of citizens. in this city builder, the level of abstraction selected by the designer is relatively high: agents are not modeled at the individual level, and events that occur within the city are sampled from some empirically-motivated distributions. however, the designer scatters in some representative “agents” anyway: anthropomorphic figures that follow basic crowd and path routines but otherwise have no effect on the simulation.
now, imagine that an apartment building is on fire. the player applies themselves narratively to this problem. upon investigation, they observe an “agent” poised at the foot of the building, conceivably staring up at the fire. they might theorize—for example—that this individual left a candle burning in their apartment, walked out to buy groceries, and returned to find their building on fire.
but! we already know that the game does not model at the individual level. therefore, any attempt to rationalize the behavior of an “agent” at the individual level—holistically speaking, i.e. in the context of the whole model—is malformed. the behavior of the “agent” is irrelevant, and may as well be noise. ultimately, the player’s narrative is delusional, and blatantly falsifiable in full view of the evidence.
but is this actually harmful? one counter holds that the “agent”, while undeniably illusory, validates its own existence by functioning as a creative prompt. one might say that it—just by existing—is a creative constraint. after all, it prompted our player’s narrative to begin with, right?
perhaps. but this illusion simply cannot be maintained forever, and forcing the player to engage in the inevitable doublethink is creatively destructive and terribly dissonant, breaking any kind of immersion. I would like to emphasize that there is nothing intrinsically wrong with the player’s narrative from above. fanciful elaborations of high-level events are fine… more than fine: positive work. but when these interpretations are contravened: instead of learning, the player is left with an empty feeling of dissonance. there is nothing to learn! only farce.
again, all these tricks may work for awhile—the effects are indeed tantalizing when a skilled metagame player appears at playtest—but unfortunately, when it comes to games, “impression of depth” must truly move beyond the mere impression. eventually, delusion will face its day of judgement: either the player will see behind the curtain, or they will realize that no hypothesis comes to fruit with any real consistency. this procedure, carried to its terminal point, is more or less irreversible and radicalizing.
the problem with this hypothetical city builder is not its level of abstraction: it doesn’t have to model agents individually in order to be a good game. it just needs to stop offering invalid representations below its level of abstraction. doing so would transform delusions into legends: stories that—while perhaps finer grained than the system’s model and therefore unprovable—fortify the player’s necessary creative role in the work.