We have somewhat of an agreement in the game design community that discussing and analyzing structures and mechanisms in game design is valuable and useful, however, there are some commonalities and differences between analysis I think we need to pay more attention to. Much how mathematicians can compare and analyze the quality and rigor of mathematical theory, we can, and indeed we should be, analyzing game design theories as well. Better understanding the common or competing structures of game design theories will help us to be more diligent in crafting these theories, and reduce confusion while debating them.
I want to make it clear I’m just throwing stuff at the wall here, I may be overlooking important aspects of theoretical meta analysis, and the terms I’m choosing may not be ideal. However we rarely talk about theoretical meta analysis at all, so this is relatively early days for our field and this post should be seen as more an encouragement to think about this stuff more, rather than authoritative and fully fleshed out proposals.
Organizing our thoughts
When talking about game design we often say something like “games where you have a binary win or loss condition result in clearer feedback”. The general structure of that statement is “Observable A effects condition B”. The idea here is that if we can observe A within a game, we can have some confidence in it resulting in B. There’s a degree of nesting which is possible here, and these statements can get reasonably complex. Even more so the evidence in defense of or against a particular statement can be extremely complex. A statement like this we can call a “Theory” or “Proposal”. There’s a few other colloquial names we use for this as well. For the remainder of this article, I’ll refer to this as a Theory.
In game design, it’s rare for Theories to be decisively proved the way mathematical theories are, although presumably, it’s possible for us to establish a small body of theories which we can reasonably argue are laws. It would be good for game designers to spend some time attempting to prove theories for which it seems possible, but it’s likely the vast majority of our operational theories will be analyzed primarily with bodies of evidence gathered from player experience (possibly measured in some way, but still, with a human in between input and output) and mental simulation. This being said we should avoid the trap of assuming that simply thinking hard about a theory is an acceptable substitute for testing the theory with evidence, regardless of the inability to objectively prove a theory.
Theories can connect together to create models and frameworks. A model is a group of theories which work together to describe a specific game mechanic reasonably comprehensively, where as a framework is a group of theories which work together to help a designer create a better game (usually, but not always, a game of some type/genre). From this point on in this article, we mostly want to talk about and compare the structure of theories.
To do or to observe
In university, I learned at length about the history and evolution of the field of linguistics. The science of linguistics started as grammar lessons which would be given to students to teach them to speak/read/write the “right” way. As linguistics evolved we learned some of these “right” ways were not very right, and we learned that some of them actually had good reasons for being thought of as “right”. Ultimately though this understanding was gained via greater understanding of how linguistics structures interacted with each other, and understanding these things, to some degree required study without judgement. I believe a similar process is occurring in the practice of game design.
The first major categorization which can be applied to all theories in game design, is prescriptive vs descriptive theory. Prescriptive theories are of the type “A game that does X is better” or “Strategy game designers should design games with binary win/loss states”. Prescriptive theories describe things game designers should or should not be doing. They make recommendations on which behaviors will result in better quality games, or at least they make qualitative judgements which aren’t really objectively quantifiable such as “easier to design” or “more terrifying”.
In contrast, descriptive theories are of the type “If our game implements mechanic X it will result in state Y”. I recently read a design blog from a designer of a new game in the genre of dwarf fortress, they observed that NPC characters spend about the same amount of time performing self-maintenance tasks each day cycle (eating, sleeping, showering), so a shorter day cycle meant a logarithmic increasing percentage of the day was spent on self-maintenance. Additionally, base designs which required NPC to travel more distance for self-maintenance tasks would multiply the time required for self-maintenance tasks. Ultimately this meant that shorter day cycles could produce completely unproductive day routines for certain base designs. This outlines a set of descriptive theories describing relationships between designed variables which may not be immediately transparent on the surface. Descriptive theories abstract concrete things into more universal relationships and interactions. They describe relationships between game systems, but they are not making qualitative statements about those relationships. In our example our game designer doesn’t claim a less productive routine cycle is good or bad, she just observes that day length interacts with self-maintenance routines in a describable way, and she can choose which variables are the most sensible to modify to make things “better” based on her understanding of these interactions.
Descriptive theories are often preferable, as they are more objective and help us to understand complex systems more deeply. They don’t easily lead us into traps where our lack of ability to quantify a qualitative judgement leads us to inappropriately prioritize one mechanic over another because their fundamental nature is to measure some cause producing some result in a diligent way. Descriptive theories are also occasionally provable where prescriptive theories are universally impossible to prove because they will always decay into some subjective human quality. But while descriptive theories can help us understand shapes of things we might not recognize just by looking at the surface, they can’t generally tell us if that shape is good or not.
For that reason we cannot avoid the prevalence of prescriptive theories. In fact a prescriptive theory is generally just a descriptive theory with a value judgement bolted on top of it. “We should have longer day cycles to decrease the punishment of new players with less efficient base designs”, where “punishment” is a subjective human quality. In order to make better games more elegantly, efficiently, and quickly we need models and frameworks which prescribe useful techniques which will help us achieve such. And in fact, it’s generally just easier for humans to create content contextualized with prescriptions rather than descriptions.
Timeline of theory population
In most intellectual pursuits generally prescriptive theories dominate early on. We don’t understand overly complex details about mechanisms and mechanics of the field
we’re discussing, so we fall back on observable evidence and run tests. In linguistics a common prescription was “Double negatives are bad”. It comes about because we can observe speech with double negatives and it feels awkward and unclear. I call these prescriptive theories “prudence predictive theories” because they seem to come with a parental desire to keep the next generation from falling into common errors in a bit of an overbearing way. We should be clear that it’s not bad to propose prudence prescriptive theories, it’s part of the natural flow of social understanding of a topic, but we need to understand that prudence theories are dangerous to hold onto too aggressively because we risk eschewing deeper understanding in favor of a false sense of security that we’re doing things “the right way” simply because we’ve overgeneralized seeing the “right” way work well and the “wrong” way work poorly in some specific context.
Therefore the next step in the natural flow is a pushback against the early prescriptive theories. As we gather more data and gain more understanding we start to look at a system as smaller parts interacting in various ways to produce various states and results. Inevitably the less rigorous prescriptive theories produced early on fail to account for edge cases, and have many areas in which they are not optimal. In order to reach a level of more diligence we must be able to describe the system. At this stage we generally don’t deal with theory proven such that we can call the principle a law. But we begin to understand the shape of the theory and what implications the claim has.
However, after a system is well described we realize that like all other systems we care about, the system has to interact with a human, and humans are not well described. What this means is inevitably we fall back on prescriptive theories to understand what we have mapped out, which at this stage I call “academic prescriptive theories”. The difference at this point however is that prescriptive theories which are built on top of well developed descriptive theories mostly have qualitatives at the human interface layer, and those qualitatives is generally far more measurable. A fun one from linguistics is the english language’s use of pronouns. Linguistics tells us that we VERY rarely create new pronouns in normal language usage. For most of our history we have only had 7, with their grammatical variants. A grammar teacher in the 1950s would prescribe the utter incorrectness of attempting to create or use new pronouns. Our descriptive studies however taught us that we settled on the 7 pronouns we had because they were culturally sufficient at the time they were being developed to communicate, in a socially acceptable way, all the situations people within that culture had to communicate, thus they were settled on because they served the needed purpose, not because the configuration was the best possible configuration. In our current culture it is socially acceptable for an individual to not want to be defined by their gender for either ideological or mental health reasons. Thus our current pronouns fail to adequately address a legitimate edge case, and we’re beginning to see language shifts to allow for new usages of old pronouns and in some cases new pronouns (although it’s yet to be seen where the dust will settle, as is often the case with linguistic change).
Eventually some descriptive theories should be proven in the same sense mathematical theories are proven. If it can be shown that in every case, a theorem always holds true, (inductively, not experimentally) and there are no unexpected edge cases, we can say a theory is a law. To date I’m not aware of any game design laws, although that being said we’re probably more looking at extensions of the fields of mathematics like game theory or economics which will outline these structures, and I’m not on the pulse of what those fields are doing right now. We should keep in mind though, for most mathematical laws, people did not start looking for proofs without gluts of evidence that the claim was almost definitely true already, so it’s still useful to propose and test descriptive theories that we don’t yet know how to prove, with the understanding that someone else might be able to prove or polish them up to the point of provability some day.