Interactive Narrative

Beyond Branching

Marc Canter, father of the computer program now called Macromedia Director, recently presented his CD-ROM Meet the Media Band at the MIT Media Lab. While presenting one component of the CD-ROM, an interactive music video where, with the help of the viewer, the lead singer explores various dating options, Canter quickly apologized for the piece having "only sixteen endings."

This notion of "multiple pathway" stories where a high variability of plot is the ideal, is extremely prevalent within the multimedia industry, academia, and popular culture at large.

[1] Laurel, B. "A Taxonomy for Interactive Movies" New Media News. Vol. 3, No. 1, 1989. The view seems rooted in popular conceptions of the ideal narrative experience involving being "inside the story." Brenda Laurel introduces her essay A Taxonomy for Interactive Movies, by citing one such conception, Star Trek's Holodeck.[1] In the Star Trek universe, the Holodeck is a large empty room into which a computer synthesizes landscapes and characters. Star Trek characters may enter the space, exploring and generally experiencing their synthetic environment exactly as if it were reality.
The inherent problem with looking at depicted representations of interactive experiences is that these representations are embedded in already familiar narrative forms. The ideas behind the depictions often exist merely as devices to further the intentions of the more or less traditional storyteller. Furthermore, while the observation of a character placed in some presumably interactive experience might be entertaining, the literal experience might not.

The Holodeck, which Laurel hails as "the best developed, and incidentally the most realistically doable, model for the interactive entertainment environment of the future," is a case in point. The Holodeck is a microcosm of the Star Trek experience itself. It's a place where the familiar Trek characters are placed in fantastic environments to explore and discover. Within this context, characters become involved in a "micro-narratives" which invariably get entangled in the framing narrative of the ship.

To the viewer, the Holodeck is just another setting for the Trek characters. The events that occur in this fantasy space are presented to the viewer in exactly the same way the rest of the Star Trek "reality" is presented, namely with sets, lighting, actors, and a choreographed camera. Rather than being the ultimate model for an interactive experience, the Holodeck is really the ultimate story device. Thus the Holodeck is to "ideal storytelling" what the Transporter is to "ideal travel" -- to get from here to there you simply have the computer "put you there."

The Narrative Prime Directive

In the universe of Star Trek, the members of the Starship Enterprise are bound to the principle of the Prime Directive. This fictional doctrine dictates that the crew members' actions must not fundamentally interfere with the cultures they encounter.

Star Trek's explicit premise is the exploration of new worlds and people, "to boldly go..." and so on. To that end, the Prime Directive stands not only to protect the worlds that are explored from being altered by the explorers, but presumably to maximize the meaning of that exploration by preventing such an alteration. In other words, when the actions of the observer significantly alters that which he observes, how can the observer know whether his observations are inherent to the subject or the result of his own actions?

The above situation exists for any observer. Cinema verité filmmakers must accept the fact that their own presence and the presence of the camera inherently alters the situation they attempt to faithfully record. To this end, the filmmaker is obliged to admit this influence and develop techniques to minimize the significance of its effect.

In storytelling we find a similar "split between worlds." The world created by the narrative, the diegesis, and the act which produces it, the narrating, are necessarily distinct.

[2] Genette, G. Narrative Discourse. Cornell University Press. 1980. p. 228 Genette defines the idea of narrative levels by stating that "any event a narrative recounts is at a diegetic level immediately higher than the level at which the narrating act producing this narrative is placed." [2] Thus the foundation or bottom-most level of any narrative consists of a narrating act. It is at this level that the intended reader, or narratee, receives the narrative. Events which occur within the world of the narrative, or diegesis, are termed diegetic. Events occurring at the level of the narrating act are considered extradiegetic.
Genette goes on to define any shift between two levels of narrative, such as between the diegetic and non- or extradiegetic, as a metalepsis.

In its most straightforward form, a metalepsis occurs when the narrator tells the story of a character telling a story. At that point, the character assumes the role of narrator in a new higher level or meta-narrative. Genette designates any other form of metalepsis transgressive. In these cases, metalepses occur unexpectedly as an element initially perceived at one level is revealed to be or treated as if it occurs at another. Genette sites Laurence Sterne's Tristram Shandy as a prime example of this latter use.

On Sterne's use of metalepsis in the narration of Tristram Shandy, Genette notes that:

[3] Genette, pp. 234-235.
Sterne pushed the thing so far as to entreat the intervention of the reader, whom he beseeched to close the door or help Mr. Shandy get back to his bed, but the principle is the same:

[Any] intrusion by the extradiegetic narrator or [reader] into the diegetic universe ... produces an effect of strangeness that is either comical ... or fantastic. [3]

[4] This reference was suggested to me by Kevin Sawad Brooks. A quintessential set of examples of such an effect in film involves background music. If, for instance, as lush romantic music swells, one character says to the other "Sorry? I couldn't hear you over the music," or in the style of Tristram Shandy, a character asks that the sound be turned down (and then it is), the effect is typically comical. Wim Wenders' Wings of Desire presents an example tending more toward the fantastic when music in the soundtrack abruptly ceases as a character, an angel, clasps his hands over his ears. [4]
Summing up the subject of transgressive metalepsis, Genette states that:
[5] Genette, p. 236
All these games, by the intensity of their effects, demonstrate the importance of the boundary they tax their ingenuity to overstep, in defiance of verisimilitude -- a boundary that is precisely the narrating (or the performance) itself: a shifting but sacred frontier between two worlds, the world in which one tells, the world of which one tells.

The most troubling thing about metalepsis indeed lies in this unacceptable and insistent hypothesis, that the extradiegetic is perhaps always diegetic, and that the narrator and his narratees -- you and I -- perhaps belong to some narrative. [5]

Philosophical concerns aside, transgressive metalepsis are at the very least exceptional and disruptive points in a narrative. The repetition of such effects tends to distract the viewer from the depicted story and undermine the integrity of the diegesis.

"Branch Point" interactions, where the viewer acts to change the course of a narrative, are thus disruptive not only when they stop the story to wait for user input, but fundamentally by placing the consequence of an action performed by the viewer within the diegesis.

By presuming to break the diegetic barrier, such approaches invariably undermine a narrative system, calling into question the boundaries between the narrative, story, narrating act, and ultimately the reader. While such an action may well be a potent tool for the author interested in calling these relationships into question, the technique shifts the viewer's attention from the narrative to its disruption. Furthermore, if such "effects" are perceived as novelties, they cannot form the backbone of an experience; the novelty wears thin quickly, the comic or fantastic quickly become tedious.

Focusing only on this exceptional potential of interactivity needlessly limits its role to one of disrupting narrative. Instead, by accepting its inherently extradiegetic position, interactivity may be seen, like the background music or editing in film, to serve the narrative by focusing the viewer's attention on the product rather than the process of the narrating.

Hollow Stories

[6] Laurel In Laurel's taxonomy, one of the three "definitional variables" for describing interactive movies is personness. Laurel defines a first-person experience as one where the viewer "participates directly in the world of the movie... [where the viewer] is a character in the action". Second-person is where "the [viewer] stands outside the imaginary world but communicates to characters or entities inside it, or vice versa." In third-person experiences "the [viewer] stands outside the action altogether," a situation Laurel notes "is hopefully never achieved in interactive media!" [6]
Such a breakdown, by its very literal notion of being either inside (good) or outside (bad) the "action" or diegesis of the movie, leaves no place for immersive narrative experiences. In Laurel's second-person experience, every instance of a "communication" between viewer and story represents a transgressive metalepsis -- a disruption of the narrative. Laurel's notion of a first-person interactive experience denies the presence of a diegetic barrier altogether. In this case, there is no narrative; the viewer is put into an environment and acts -- his or her experience is the story. Removing the viewer from the "action" altogether (third-person) is presented as entirely undesirable as the experience reduces to nothing more than an ordinary cinematic experience, presumably something the interactive moviegoer would not tolerate.

The fundamental problem with Laurel's taxonomy is the implication that one cannot provide an "immersive experience" without literally making the viewer a character. Undeniably, having viewers experience a story as if they were present in the diegesis is powerful. At the same time, giving viewers a sense of presence need not bind their actions to consequences in the story.

[7] Laurel

[8] This situation is discussed by Glorianna Davenport in the article "Cinema with 'thinkie' appeal" Financial Times, Feb. 5, 1996.

Discussing the significance of interactivity, Laurel poses the question, "Are the user's choices simply cosmetic, or do they fundamentally affect the plot?" [7] In a narrative, this notion of significance seems inversely defined, since the ability to alter events in the plot actually works to diffuse the significance of the story. If viewers can change characters' actions with the wave of their hands, why should they care about the story? What indeed then is the story? [8]
Genette provides a more useful description of the conditions for what might be called Narrative Immersion near the end of his essay:
[9] Genette, p. 260
[The] more transparent the receiving instance & the more silent its evocation in the narrative, so undoubtedly the easier, or rather the more irresistible, each reader's identification with or substitution for that implied instance will be. [9]
In other words, the ideal is not for the viewer to map directly into the story, but rather into the telling of that story.

Counterexample: Interactive Theater

In John Krizanc's interactive theater piece Tamara, actors performed scenes in parallel throughout the many rooms of a real house. While audience members were free to follow characters throughout the house and observe their performances, the audience's presence was never acknowledged by the performers. Thus the "diegetic barrier" remained firmly in place even though the audience and actors shared the same literal space.

In this scheme, the audience was continuously interactive, in that they were free to move their bodies normally. The range of their interaction was quite limited -- an audience member could only leave or enter a room when characters did so. The significance of the interaction was high as the author designed the room's performances in such a way that each audience member only gets a portion of the total story depending on what rooms they visit. In this way, audience members are encouraged to speak to each other during a dinner held at the "intermission," to try and piece their disparate experiences together.

[10] This meeting occurred at the Interactive Screen conference held in January, 1995, at the Banff Center for the Arts in Alberta, Canada. I never saw Tamara performed. My understanding of its content and operation comes from informal discussions with the author whom I met at a workshop in Banff. [10] In these same discussions, amid an audience of primarily other writers, Krizanc described his recent discussions with a developer about producing Tamara in CD-ROM form and showed the lengthy design document the developer had produced outlining an approach to the project. Amid concerns over bandwidth, the document suggested that the play be recast as a game with the viewer choosing plot lines resulting in one of many possible endings. For Krizanc, the very idea that his main character, the Count, might live instead of die depending on the actions of the "player" was unacceptable; it "completely destroyed the whole point of the story" as he had written it.

In addition to underscoring the strong negative potential of handing control of events in the story over to the viewer, this incident demonstrates how bound the idea seems to be to a notion of interactivity. Even when presented with a narrative with a ready-made model for interactivity, the CD-ROM developer felt obliged to replace it with a branching narrative structure.

Fundamental Properties of Interactive Narrative

narrative intention
narrative immersion
narrative structure
narrative response
narrative guidance

Five Fundamental Properties of Interactive Narrative

In response to Laurel's taxonomy, this section describes a set of five "Fundamental Properties of Interactive Narrative." As aspects of each property are discussed, any sense of an ideal refers only to the ideal capabilities of a general purpose Storytelling System. In other words, authors should clearly be allowed do whatever they feel best serves their story. The aspects discussed represent the kinds of controls and capabilities a Storytelling System should have in order to facilitate a range of potential uses.

Narrative Intention

Narrative intention refers to the point of the narrative, or in the case of an interactive narrative, the viewer experience. The narrative intention should answer the viewer's question, "Why am I viewing this story?" and the author's question, "Why am I telling it?" Ultimately, the narrative intention depends on the author and the specific story he or she wishes to tell.

Ideally, a Storytelling System communicates a sense of purpose to viewers while providing authors with a sufficient level of control over those aspects they feel are essential to the telling of the piece. Note that an explicit notion of intention need not imply an explicit or "heavy-handed" articulation of that intention in the narrative. Indeed, an author's intention might be that viewers perceive an absence of intentionality to their experience.

Narrative intention is the broadest and most fundamental property of an (interactive) narrative as ultimately the decisions made concerning the other properties are made in its context.

Narrative Immersion

Narrative immersion refers to the quality of the viewer's reception of the narrative. Recalling Genette's definition of the ideal narrative reception, narrative immersion is achieved when the "receiving instance" is transparent, and "silent [in] its evocation." In other words, narrative immersion concerns how well the narrating functions as a background element to the narrative, silently conducting its construction.

In the interactive narrative, immersion relates to how well the Storytelling System engages the viewer in the diegesis rather than in the mechanisms of its construction. To this end, a primary concern for a Storytelling System is how to manage interaction with the viewer without distracting them from or otherwise disrupting the narrative.

Narrative Structure

Narrative structure relates to the form, shape and rhythm of the narrative. As such, narrative structure is inherently a function of the temporal nature of the narrating; it relates to the time of the telling.
[11] Genette, p. 88 Discussing the uses of temporality in narrative, Genette notes that "a narrative can do without anachronies, but not without ... effects of rhythm." [11] In other words, while events in a narrative might be presented in a strictly chronological way, a sense of variation in the pace of the telling is practically indispensable.
Interactivity fundamentally frees the narrative from a fixed and pre-determined structure. This does not mean, however, that Interactive Narrative must lack a sense of structure. Indeed, the perceived "flatness" of many CD-ROMs and hypertexts, where the viewer keeps exploring until they either exhaust the system or themselves, tends to emphasize the importance of structure. As Toni Dove likes to say about many interactive experiences, "closure is boredom."

Consider the function of chapters in the novel. Short of providing absolute closure, the existence of each chapter at least provides the means for a satisfying "breaking off" of one's reading experience. As one reads, there exists a motivation to "at least finish off" the current chapter. The end of a well-designed chapter leaves the reader either wanting to read more, or in a position to put the book down with the pleasant sensation of wanting to return later to pick up from that point.

A crucial function for a Storytelling System is the ability to manage the narrative structure as it is constructed, to provide a sense of shape, pace, and rhythm to the experience. In addition, structure might include providing a sense of closure to the viewer's experience -- even though the story might be ongoing.

In order to provide structure without needlessly limiting the potential for responsiveness to the viewer, storytelling systems must clearly go beyond the use of rigid template structures. Instead, the storytelling system should be capable of forming an emergent structure over the course of a particular experience.

To facilitate an emergent structure, some persistence of structural knowledge over the course of the experience is required. At the very least, a storytelling system should be aware of what materials have already been shown to the viewer and avoid repeating them. Sadly, this basic functionality is missing in any hypermedia based on static "state-free" graph structures. In addition, a "structure engine" might take into account the order of previously shown materials to determine how to structure the experience in the future.

In addition to providing structure, a storytelling system might have special knowledge about how to articulate that structure to the viewer. In cinema, conventions for visual transitions have emerged to represent short temporal ellipses with dissolves, larger ellipses with fade outs, and larger possibly conceptual structural divisions with titles. A storytelling system might be capable of dynamically presenting such conventional transitions to articulate structure while maintaining immersion.

Narrative Response

Narrative response encompasses the fundamental question for any interactive narrative, "How does the narrative respond to the viewer?" or conversely, "How does the viewer influence the narrative?"

In a storytelling system, the answer to these questions relies on what types of control or handles the system is capable of providing. It is then for the author to decide exactly how viewers' actions are mapped to these controls.

Narrative response might occur at the level of the diegesis, as when the viewer literally chooses what happens in a story. Such a use however, as discussed previously, may fundamentally work against both narrative intention and immersion. As such, the most useful potential for narrative response occurs in extradiegetic roles: namely in the mechanics of material selection and the conditions of material presentation.

[12] I was first exposed to the idea of building responsive interactive environments while part of a team working on the "Wheel of Life." This project is described in:

Davenport, G. and Friedlander, L., "Interactive Transformational Environments: Wheel of Life" chapter in Contextual Media, ed. Barrett and Redmond, MIT Press, 1995.

In considering the potential for viewer influence on material selection, we come to the very heart of the storytelling system and its model for making editing decisions. Depending on that model, a storytelling system might provide a range of editing controls or handles. In the Automatist Storytelling System, keywords are one type of handle. The viewer's actions might then map into tending toward or away from materials described by those keywords.

A less direct type of handle a storytelling system might provide is a notion of the pacing or speed of the story presentation. Mapping viewer action to this control might be simply literal ­ a slider control for a "faster" or "slower" presentation. It might also be derived from other kinds of viewer interaction. For instance, perhaps the pacing could be based on the amount or level of viewer interaction; as the viewer interacts more, the pace of the story increases.

In terms of how viewer interaction might map into material presentation, the range of possibilities are dependent on the conditions and form of the presentation. For a computer-based presentation, the visual appearance of elements on the screen and the quality and use of sound are both rich domains for exhibiting viewer responsiveness. In a theatrical installation space, qualities like the use of physical space and lighting are potent ways to express narrative response. [12]

Responsiveness need not only refer to a direct and immediate response to a viewer action. The operation of a storytelling system might be responsive to viewers by incorporating accumulated knowledge of their history with the story into the construction of the experience. Thus, to be responsive to viewers in this deeper sense, a storytelling system must maintain persistent knowledge gathered through an accumulation of interaction. Continuing the simple example above, a viewer's use of "keyword handles" might accumulate to form a kind of viewer interest profile on which the storytelling system could base its future material selections, even in the absence of future viewer interaction.

Narrative Guidance

Narrative guidance is the culmination of each of the preceding properties; it concerns the fundamental challenge of providing narrative structure and responsiveness while preserving narrative intention and immersion.
[13] Galyean, T. "Narrative Guidance of Interactivity" MIT PhD Thesis, 1995. pp. 58-59 Tinsley Galyean, in his doctoral thesis Narrative Guidance of Interactivity, utilizes the metaphor of a river to describe a guided interactive experience. In a river, Galyean notes, there is a "continuous flow," in which the viewer "steers" and generally never encounters "dead-ends." [13]
Narrative guidance represents the current to the river -- the pull of the story. While viewers might have a range of options for interacting at any given moment in their experience, narrative guidance dictates that those of particular relevance should be emphasized. The relevance of a particular interaction should be a measure of its ability to serve the narrative -- its intention, immersion, structure, and response.

In the interest of preserving immersion, for instance, options or choices might sometimes be made "silently" by the storytelling system. In this way, the system prevents a disruption to the narrative by isolating the viewer from having to make a conscious decision. Ideally a storytelling system should be capable of providing a sliding scale of interaction, from giving viewers a high degree of conscious control to letting them simply "sit back" and watch the story unfold, as they would with a non-interactive narrative.

This notion of level or degree of interactivity might also be a function or manifestation of the narrative structure. Just as a river might consist of calm stretches punctuated by white water -- a storytelling system might orchestrate an experience to have highly responsive segments where the viewer is free to explore then transition to fast-paced segments where the viewer is more or less pulled along and left to enjoy the ride (ideally the river metaphor drops off here and viewers are left not fearing for their lives).

Two Design Heuristics for Storytelling Systems

A common experience when viewing CD-ROMs seems to be an increasing frustration with having to use the story's interface to "get at" the content. Eventually, if you're actually interested in the content, you just want the thing to "play out by itself."

Indeed, the common click-to-go-forward paradigm for interaction coupled with static graph structure navigation schemes, seem to place the viewer in an adversarial position with the story. Instead of giving the viewer their experience, such a scheme requires that the viewer constantly push the story forward, a situation roughly analogous to listening to a narcoleptic storyteller.

The ability for automatic- or self-playout serves as a powerful design heuristic for building a Storytelling System. Designing around the potential absence of the viewer requires that a system be built with enough base-level competence to present its content. To add interactivity then poses an interesting challenge as the role and value of the interaction must always be gauged against its absence.

A similar heuristic for designing a storytelling system is something we might call the Finder Challenge. A former colleague, David Kung, developed a streamlined technique for evaluating new CD-ROMs (or seedy roms as he called them). Once he figured out the general scheme of the interface, he'd quit the program to view the contents of the CD directly, the way you'd view the contents of a hard disk (on a Macintosh, this outermost application is called the Finder). From there, he'd simply open up and view whatever materials he thought might be interesting before finally tossing the CD to a shelf to collect dust.

As with self-playout, the designer of a storytelling system might imagine a base-level functionality of giving direct and immediate access to all of the story's content. Any additional functionality or control given to the viewer must then be gauged against direct access. In this way, the piece must prove its value by enabling a method of construction better than simple random access.