Interaction & Narrative

[1] Genette, G. Narrative Discourse. Cornell University Press, 1980. p. 3 Before considering the relationship of interaction to narrative, we need to be clear about our terminology regarding narrative. Gérard Genette begins his essay Narrative Discourse with the following distinctions: (1) the story is the signified or narrative content described, (2) the narrative is the signifier or form of the narrative discourse itself, and (3) the narrating is the "producing act." [1] Thus, narrative in a general sense is presented as a system where the act of storytelling, the narrating, produces a narrative through which the viewer constructs a story world, or diegesis.
The introduction of interactivity complicates this picture. In the novel and film, the form of the discourse -- the narrative -- exists as a static object to be presented the same way to all viewers. With interactivity, the assumption is that the narrative varies. Interactivity could thus be said to occur at the level of the narrating. Each interactive reading or viewing represents a separate "producing act" the result of which is a narrative specific to that viewing. Thus when we speak of the narrative of an interactive narrative, we refer not to a static form but to the potential and inherently variable viewer experience of its construction.

The Constraints of Mass Audience

Interactivity raises the question: who's telling the story? While the ultimate understanding of any narrative depends on both the author and the viewer, authors have generally exerted explicit control over the exact form of the narrative.
[2] Dove, T. "Theater without Actors -- Immersion and Response in Installation" Leonardo, V27, No.4, 1994, p. 281 Seen this way, a key question becomes: why should an author relinquish control over the narrative? Electronic media artist Toni Dove recalls her initial response to the idea of incorporating interaction in her work as "Why should I replace intellectual challenge with multiple choice?" [2] Conversely, one might ask: why should a viewer want to exert control over the narrative? While the idea of interactivity remains appealing to the general public, people aren't exactly restless in their theater seats for something to click on.
To begin, an author may be willing to release some control if in return such a loss provides additional benefits as prior constraints are removed.

For example, documentary films must often conform to a rigid structure imposed by the conditions of their presentation. A program might need to be a specific length or be structured in a particular way to facilitate television scheduling. At the same time, filmmakers generally gather much more content than they can fit into their allotted time slots.

The inherent necessity to produce one specific "cut" of a story places a limitation on the use of content. Structuring a story around a particular theme might prevent the full incorporation of available material about another. Furthermore, presenting only a single structure limits a story from potentially taking on a variety of meanings through alternative tellings. The variability of the interactive narrative's form enables a multiplicity of story meaning.

Even without constraints imposed by the presentation, additional constraints relating to the audience exist. Lacking knowledge about each specific viewer, the author is instead asked to gauge the range of their audience and, in the interest of broadest possible appeal, structure the narrative to please the lowest common denominator.

For instance, lacking knowledge of the amount of time a specific viewer has available or is willing to spend, the author is obliged to conform to some "practical" time limit given their content.

Lacking information about a specific viewer's interest or knowledge, the author is similarly asked to tailor story content to a vague notion of "broad interest" or appeal. The result is an absence of any significant depth on issues not believed to be of general interest.

In sum, lowest common denominator programming places depth of content in inverse relation to breadth of audience; the result is shallow and disjoint "sound bite" programming.

The converse of the above is that relinquishing certain aspects of authorial control enables viewers to form a more personal and meaningful connection to the story. In this way, interactivity may function to increase viewer engagement with the narrative by facilitating a specific viewer's knowledge and viewing situation. As a form that supports multiple meanings, the interactive narrative has the potential to tell more complex and personally meaningful stories than those delivered to a mass audience.


Many approaches to incorporating interactivity with narrative have been based on a literal notion of making viewer actions map into story actions -- placing the viewer in the role of a character. Consequently, interactive stories based on this model have been used to tell stories about making choices, exploring the results of the actions we choose to make. Such interactive stories presumably encourage the viewer to role-play and thus "learn by doing."
[3] Dove, p. 281 Unfortunately the experience of such stories often seems empty. Choices seem contrived and constrained and the experience seems flat and ultimately devoid of meaning. Toni Dove attributes this effect to a "fallacy of choice" and finds the process of making decisions within an author's preconceived "matrix" of possibilities analogous to the "false sense of choice" instilled by advertising. [3]
Furthermore, making a viewer's actions have literal consequences in a story may undermine a story's integrity. This point is discussed in the following section, Beyond Branching.
A simple graph-structure hypermedia experience.

In this model, content is embedded in containers (documents, card, screens) that are explicitly linked to other containers.

A large part of the problem with successfully integrating interactivity and narrative stems from the models for constructing such experiences implicit in current "hypermedia" authoring tools like Macromedia Director and hypertext systems like HTML on the Web. In these systems, the author creates links between pieces of content forming a kind of flow chart or graph structure into which the viewer is placed to navigate. In this sense, the author works directly at the level of the viewer interaction. Thus, Dove's notion of feeling constrained to a pre-determined "matrix" of interaction is inherent to the model of hypermedia.

Another inherent limitation of hypermedia is the fact that the resulting structure is static and lacks "state," or capacity to store knowledge. The structure itself captures no sense of the history of the experience nor does it have any built-in competency for presenting itself short of the viewers direct use of links or branch points.

Scaling the Story, Scaling the Telling

Hypermedia systems constructed in this way are extremely difficult to "scale" to include large amounts of content. As each new piece is added, the author must consider the potential linkage of that piece of content to every piece of content already in the system. In this way, adding content is an exponentially complex task.

Furthermore, such systems place the author in the position of effectively pre-thinking all possible viewer pathways through the content. Every "hard-coded" link between two pieces of content in effect freezes the function or intention of the linkage into the structure of the navigation. Thus, the organization of the story is difficult to scale as any change in "retrieval" functionality necessitates large-scale changes to the pre-coded link structure.

Storytelling Systems

[4] Davenport describes the design of "adaptive storytelling systems" in:

Davenport, G. "Seeking Dynamic Adaptive Story Environments" IEEE Multimedia, Fall 1994, pp. 9-13

Storytelling Systems present an alternative to simple graph-structured hypermedia. The Storytelling System isolates the author from the process of explicitly linking a story's content. Instead, the system uses descriptions of the content and built-in editing competencies to select and sequence materials dynamically. The Storytelling System provides hooks to make this sequencing process responsive to a viewer. Thus the Storytelling System is a kind of "editor in software" or "narrative engine" -- a computer program capable of constructing responsive narratives based on the content and description provided by an author. [4]

The Automatist Storytelling System

The word Automatist is used to denote the specific model presented in this thesis. As such, the Automatist Storytelling System is one possible approach to building a Storytelling System.

Automatism is a word used to describe a branch of the surrealist movement; it represents the process of creating art based on a kind of "automatic" or unconscious free association. The intention is a "truer" experience as meaning emerges from the interactions of individual expressions rather than from a structure imposed from an "exterior consciousness."

The model presented in the Automatist Storytelling System is similarly content-driven and decentralized. Structure and meaning are considered emergent properties of the storytelling process. Rather than there being a central "conducting" process, sequencing decisions result from the interacting effects of individual material presentations.

The Automatist Storytelling System edits by association. Specifically, the approach uses keywords as a means of indirectly defining potential links between materials. During the presentation process, keywords function in parallel, pushing and pulling the narrative toward and away from specific pieces of content.

Thesis Structure

The Automatist Storytelling System is presented in this thesis in five stages:

Part 1, Interactive Narrative, presents a theoretical framework for discussing interactive narrative, provides a critique of "branch-structure" narrative, and establishes a set of five Fundamental Properties of Interactive Narrative.

Part 2, Approach, describes the approach of the Automatist Storytelling System by discussing related and influential research. This section also introduces the keyword-based knowledge representation scheme common to both the ConTour and Dexter systems.

Part 3, ConTour, describes the design and implementation of ConTour, a graphical demonstration of a simple Automatist Storytelling System. The system represents a potential "back-end" or "narrative engine" for an end-user storytelling system. In addition, the application is a kind of "digital editing assistant" capable of producing "steerable" presentations of keyword-annotated materials.

Part 3, Dexter, describes the design and implementation of Dexter, a generalized system for browsing collections of documents on the World Wide Web. Dexter represents an application of the principles of the Automatist Storytelling System to the problem of supporting "true browsing" on the web.

Part 5, Extensions, concludes the thesis by discussing possible extensions of the Automatist Storytelling System. This section also presents scenarios for future applications.