Narrative Choice Paper

Decision-making has always been an important component in the realm of interactive medium; games from all eras have strongly stressed skills involving resource management and allocation as well as quick thinking. As games have grown into a storytelling medium, likewise developers have been seeking ways to make the stories they tell through games interactive in themselves; to bring player choice into the process of crafting a narrative, usually by offering decision-trees. Relative to the rest of the gaming world, though, there's relatively few titles and very little academic material that explores this method of storytelling, the only consensus between developers and academics alike being that it seems as though narrative choice systems preclude a certain degree of explicit characterization or cohesion within a story in favor of giving the player control over its events. With so little material published on this subject, though, we can likely assume that there is an awful lot yet to be understood about what constitutes a strong narrative choice system in a digital game and therefore plenty of room to develop a decision-making paradigm that can act as a tool to help writers rather than impose on their ability to craft characters and situations. Fortunately we do have a thorough understanding of what constitutes a satisfying narrative as well as a satisfying game, and decades of psychological research can give us a strong insight into how the human mind makes decisions and expects them present themselves.

The bulk of my research thus far has been focused on the latter two of those three topics--that is, games and psychology, as the basics of narrative are well-established and relatively stable while the basics of decision-making and game design are more malleable and carry more immediate weight in terms of the direction of this thesis. My exploration of choice purely in games has branched mainly in two directions: first choice as a general gameplay element, IE breaking down a choice in gameplay systems to determine their level of interest; and second choice as an interactive narrative element, IE how developers have approached the focus of this thesis thus far. As a gameplay element choice is relatively well-outlined, tactically, strategically, and managerially. The foundation of this branch of decision-making is basic risk and reward; whether players are evaluating what weapon to use and how best to use it or crunching numbers on equipment in a role-playing game, the consensus among designers is that gamers will inevitably try to maximize their reward; therefore inconsequential or minor decisions should be minimized or avoided while necessary and important decisions--IE ones with direct and visible impact on the flow of the game--are what designers should strive to create; critical decisions, IE life-and-death situations with a high degree of pressure--can become just as cumbersome as inconsequential decisions as players can be just as easily frustrated by frequent fail states as they are by frequent inconsequential choices slowing down the flow of the game. High numbers of critical decisions become an exercise in trial-and-error (most space shooters) or punishment for curiosity (King's Quest, Space Quest) rather than engaging decisions. If we were to translate this onto a narrative-based paradigm, we could imagine a decision tree where the player is given choice A or B, and B results in an immediate fail state or being denied a major branch of the game while A progresses normally. Even if choice B is a realistic option, it's either a frustrating penalty or else a choice that common sense would dictate that nobody would ever choose if they understood the consequences; therefore there's no reason to include the decision.

Rather, what designers should strive for is dilemmas, or choices that require careful evaluation and have ambiguous circumstances and data surrounding them. As a small-scale example, say we have an action-RPG with a focus on managing equipment. The designers narrow down their ideas to two options: either A: represent every possible facet of equipment on a person's body--boots, greaves, gloves, shirt, armor on top of the shirt, helmet--so that players can mix and match as they choose, plus one accessory; or B: roll the more minor elements of equipment into the accessory slot and just represent major pieces of equipment; chestplate, greaves, helmet. Depending on the game's execution either of these methods may be fine, however, the designers decide that boots consistently add progressively larger movement speed bonuses, with some occasionally designed for specific character classes and granting bonuses to their primary stats. For the player, if setup A is used, this isn't interesting as they simply find a new set of boots, compare its speed bonus to the previous' boots speed bonus, see whether it helps maximize their primary stat, and equip whichever set of boots is higher. The decision is obvious and therefore almost inconsequential but for the fact that they have to stop the game for a moment in order to equip the new boots. If boots are rolled into accessories, however, there's a dilemma as the player now has to decide between either the speed-boosting boots or any number of other items that present equally attractive bonuses--amulets with strength boosts, rings with magic spells attached to them, et cetera. The decision isn't immediately obvious and depends largely on the player's current situation and preferences, making it much more interesting.

If we open this up to a narrative-based paradigm, we see narrative choices granting systemic rewards and therefore instantly becoming less interesting as players invariably select for whichever reward is greater based on their preference and play style rather than whichever narrative choice they find more agreeable. We see a prime example of this in the game inFamous, where the player is granted something on the order of 40 narrative choices that grant points for either being good or evil; if the player can balance the scale in one direction far enough, they'll unlock stronger powers and eventually gain unlimited energy if they max out on either side. Being closer to the middle grants no bonuses. Players are interested less in the moral ramifications of their decisions and more in obtaining the systemic reward, and therefore will bias entirely in one direction or another with no sense of middle ground or personalization. Another example still is evident in Star Wars: Knights of the Old Republic, where early on in the game players are asked to seek out a bounty on a woman wrongfully accused of murder. While at this stage the good/evil biases are relatively weak, taking the time to investigate the incident further yields a greater experience point total. Additionally, should the player choose to spare the woman's life and simply lie to the crime boss who wants her head, they still claim the reward for killing her and obtain a gift from her in return for their kindness; the reward for the "good" path in this narrative is easily and measurably greater, with the only factor being the amount of time the player wants to spend on the quest--which isn't a systemic factor. While some players may be ignorant and will overlook this choice, anyone aware of even the possibility of the systemic rewards involved will find it uninteresting, and it becomes more of a matter of punishing players for ignorance or lack of observational skills within the game than an interesting narrative choice. What would make this an interesting dilemma, on the other hand, is if the woman were, in fact, guilty as charged and clearly a bloodthirsty killer, offering no reward in return for sparing her life, but also had a family she was trying to escape with. Instead of a clear case of a victimized woman at the hands of a crime lord, we have an ambiguous situation; either leave a killer on the loose, leaving the possibility of her coming back later to do more wrong, or deprive a family of its mother and possibly incite them to revenge later. By dissociating this situation of a clear systemic reward we dissociate it from a clear "right" choice as well, making the matter more the player's evaluation of their own guilt versus their sense of justice than the matter of whether the player is more observant or less--or, more importantly, whether they want to kill her simply for the sake of maxing out their dark side points. To put it simply, we have a narrative dilemma. While we could introduce reward into narrative choice as a systemic dilemma--IE, create two different items that the player could receive for any of several choices, and they only get one of them--it still becomes less a dilemma based on the ramifications within the narrative and more a dilemma based on which item the player prefers; there may be interest present, but it won't be in the narrative choice.

Already we're able to apply a lot of fundamentals of choice in game design to choice in narrative, but readings on narrative choice as a subject itself are sparse at worst and ambiguous at best. Books on game development and game narrative tend to categorize this under "nonlinear" or "branching" narrative, and tend not to clearly outline the goal of implementing such a system as much as simply outline their existence, citing tree, module, or grid systems as a means of tracking them and simply stating that they put control of the story more in the hands of the player than in the hands of the writer, which seems an oversimplification and contrary to the point of a narrative-based system: ideally it seems that it should enrich one's ability to tell a story rather than limit it in favor of a superficial sense of control or authorship; make players feel more involved with the characters rather than distance them from each other or preclude a sense of character development. Still yet more ambiguous is the matter of how much "authorship" players are meant to have in these systems; whether we should lean towards smaller, more subtle decisions or larger, more direct ones; whether we should script major storylines around branching paths or try to incorporate choice as an entirely emergent function. One way or another the mainstream gaming industry has explored many methods from one end of the spectrum to another with seemingly a very small degree of certainty. In trying to fill in this gap I elected to explore psychological texts on judgment and decision-making, most particularly James G. March's A Primer on Decision-Making, which outlines the many ways that psychologists study and break down the everyday human decision-making process, the interest here being that, in the tradition of seeking immersion in games, if we know how people actually make decisions and expect them to present themselves in real life we can better replicate them in an interactive narrative. To that end, March handily breaks down decisions with respect to two different psychological frameworks: rationality and rule-following.

Game designers are familiar with Rationality. With respect to the name of this architecture we should not misunderstand the term "rational" to mean "intelligent" or "sound-minded" as much as we should understand it to mean "conscious" or "intentional." It simply holds that people consciously make decisions based on a risk-versus-reward analysis, intent on maximizing their reward. There is some disagreement, however, as to whether they do as they intend, and Rationality is divided into two subsets to account for this: Pure Rationality and Limited Rationality. Both hold that people make decisions based on probabilities, not certainty, and that principal characteristics of Rational decision-making include post-decision surprise--either good or bad--as well as regret--the awareness of a better choice--and risk--or variation in outcomes. Pure Rationality holds that people do not, in fact, maximize their reward with 100% accuracy but rather try to maximize an expected value based on probabilities and risk factors. Expected value analysis involves developing an imaginary decision-tree, with branches representing either acts on the part of the decision-maker or "acts of nature," or acts outside the decision-maker's control. We see, then, that there's a clear parallel in this model of Rationality and in game design as many games do represent their stories with decision trees; almost never do we see an "act of nature" represented in these games, however. What would constitute a reasonable "act of nature" in a branching narrative remains to be seen, but we can suppose upon the previous example regarding the woman in Knights of the Old Republic, where the player constructs the narrative in their mind as possibly involving her returning to do more damage or possibly not; her family possibly taking revenge on her death, or possibly not; and also within what timeframe either of these events might happen. Those would each be examples of potential acts of nature in the game's decision tree, which would influence players' expected value analysis of these choices.

Before we explore Limited Rationality, we must first explore the factors that define risk in greater detail. Those factors, as outlined by Pure Rationality, are knowledge, actors, and preferences. Knowledge simply refers to how much the decision-maker knows about the state of the world and any other actors in the decision-making process. Actors refers to, literally, the number of actors in the process, with each actor increasing the number of potential branches and weighing in on the probability of any given situation. Preference refers to the decision-maker's values; IE, what constitutes a good reward to them and what doesn't. Each of these factors can create a high degree of ambiguity; people can make decisions with a certain degree of ignorance about the world or a situation, experience uncertainty about other actors, or even have ambiguous or conflicting preferences.

Limited Rationality states that although people intend to be rational, there's simply too much ambiguity and too many bottlenecks in the human decision-making process to create a sense of perfect Rationality. People don't know all alternate choices, they don't consider every consequence, and not all preferences are immediately evoked. Only a few of each are known, and they're reviewed sequentially rather than simultaneously. As such, people tend towards satisficing, or selecting an option that's "good enough" rather than maximizing, or choosing an option that's "best possible." Additional constraints imposed by Limited Rationality include Attention--peoples' ability to deal with multiple signals or streams of data at once; Memory--peoples' ability to retrieve previous lessons; Comprehension--peoples' inherent difficulty in organizing information for use in a decision or recognizing its relevance; and Communication--a more organizationally-based constraint than an individual one, wherein knowledge is differentiated between specialized individuals with differing identities and preferences. Each of these offers designers a good set of tools for either building ambiguity where necessary and creating dilemmas or eliminating ambiguity where it isn't desired.

While Limited Rationality is a difficult study due to its inherent ambiguity, it yields for us a several ways in which people break down decisions, giving us an insight into possible mental processes that players can be applying to their decisions. The first method is editing, wherein people re-construct problems with respect to a set of core elements, ignoring the unexpected in favor of the expected and ignoring things that seem relatively minor or peripheral to the problem. The second is called deconstruction; IE, the re-organization of a problem into multiple, smaller problems and decisions in order to make each element easier to digest. The third tool for breaking down problems is heuristics, or shorthand stereotypes that people can substitute as variables in place of complete calculations. The fourth, framing, simply refers to frame of reference; the identity and preferences of the decision-maker biasing them to think about the problem in a specific fashion.

In addition to Pure and Limited Rationality March also presents us with Rule-Following, an alternate form of decision evaluation that states that people don't necessarily think in terms of rational risk versus reward but rather in terms of trying to adhere to an identity. They recognize a situation, ask "what kind of person am I?" and then ask, "what does a person like me do in a situation like this?" In other words, they follow a set of rules. While this sounds like a relatively simple architecture compared with Rationality, Rule-following is most assuredly a very conscious and deliberate process, and as with any knowledge the concepts of identity which drive this process can be ambiguous. People can possess multiple, conflicting identities--professional versus personal is a classic example--and can even misunderstand the identities that they possess; or they can not immediately recognize the relevance of a particular identity to a particular situation. The mechanisms by which identities are built or accessed are equally as complex as the ways in which people evaluate risks in Rational theory. Those processes include Experiential Learning--"learning on the job," so to speak, experiencing rewards and punishments for specific actions and learning to evoke an identity based on those lessons; Categorization--organizing and prioritizing rules and ideas around central concepts of an identity; Recency--a simple fundamental that states that people will tend to repeat identities that have recently been evoked; and Social Context--wherein the expectations and rules of an identity become highlighted by the presence, real or imagined, of other people in the system.

If we were to equate Rule-Following to a term in gaming, it would most certainly be role-playing as it very literally boils down to the construction of an assumed identity by the player, which makes Rule-Following a very fitting metric for understanding decisions from the player's perspective. What Rule-Following also highlights is that this role need not be constructed by the player themselves; much of it is understood with respect to professional identities within an organization, which are notably not constructed by people themselves but rather constructed for them by the organization they are a part of, with a variety of devices outlined in the reading: the providing of models, allowing people to learn by example; the use of cues, visual, audio, or otherwise to set professional context and outline appropriate behavior; and the providing of experience itself, as outlined under experiential learning. In the context of a strongly narrative-driven experience the player almost can't construct their own identity outside of a few constraints based on a given title's mechanics and flexibility; the stronger the sense of character development that the developer wants, the fewer means of self-determination that there can be. These devices, then, outline a means by which developers can help players to assume the identity that they need players to resonate with, or in the case of less constrained titles, tools which they can use to make suggestions and help players define and understand their identities more strongly.

While this research is strongly focused on a single psychological text, it nevertheless has revealed a great deal of information about how decisions are constructed in the real world, reinforcing earlier conclusions about the relationship between systemic rewards and narrative decisions as well as the development of dilemmas while also providing a set of tools with which to dissect any decision; a series of real-world metrics by which narrative choice can be measured. Additionally, it has helped to expose a direction to the narrative decision-making paradigm which we are trying to construct in this thesis, wherein we wish to preserve narrative integrity but still allow a strong sense of personalization; the key being "personalization" and not "non-linearity." I have therefore developed a rubric with which to evaluate choice systems based on the preferences of this thesis.

1 - Gameplay integrity. Is there a systemic preference for one choice or another in the decision system being evaluated?

2 - Narrative integrity. Are the basic pillars of a good narrative--IE, character development, plot, rising tension, et cetera--still intact within the title being evaluated?

3 - Personalization. Does the narrative choice system enhance the player's sense of personal involvement or investment?

4 - Discernability. Does the decision-making system exhibit clear consequences, strong dilemmas, and generally evoke the elements of human decision-making, or is it mostly superficial?

This rubric needs some additional substantiation and refinement, which will follow in the coming week, but in the meantime it at least offers a solid set of criteria.