January 29, 2008
Earlier I mentioned that this book will discuss three “effects” that arise in the relationship between system operations, surface presentation, and audience experience. These will serve as the major waypoints for the remainder of this volume. The first — “the Eliza effect” — is the well-known phenomenon in which audience expectations allow a digital media system to appear much more complex on its surface than is supported by its underlying structure. However, I will consider what most authors have ignored: during playful interaction with the simulated therapist for which the Eliza effect is named, the illusion breaks down rapidly. One alternative to breakdown, with a system of this sort, is to severely restrict interaction. Another is that pursued by many modern games: never building up the Eliza illusion, and instead clearly representing the operations of a simple system on the work’s surface. But these simple systems prove too limited for the fictional experiences games seek to make available to their players, resulting in breakdowns of a different type. This leaves only one option for those seeking to create ambitious playable fictions: more developed system models of story and character.
Expressive Processing’s next waypoint is the first major story generation system, Tale-Spin. Though created decades before modern games, it represents an initial step toward a more flexible model of fiction. Its characters have goals, make plans to satisfy them, examine their relationships with other characters, speculate about the plans of others, and so on. But little of this is visible to audiences in the resulting story, and little can even be deduced through repeated interaction. It is this that leads to a new coinage — “the Tale-Spin effect” — for works that fail to represent internal system richness on their surfaces, in an inversion of the Eliza effect that is not uncommon in digital media. From there Expressive Processing looks at a different model of character planning (in the first-person shooter game F.E.A.R.), the alternative of statistical AI approaches, and the development of systems for story generation over the next several decades. While the examined story generation systems demonstrate a variety of interesting models of fiction, most focus on this to the exclusion of a rich audience experience, resulting in the continued presence of the Tale-Spin effect in even the most advanced work in this area.
The book’s third waypoint is a work famous for making a relatively complex system into a rich, enjoyable experience of play: the city planning game SimCity. “The SimCity effect” is my term for systems that shape their surface experience to enable the audience to build up an understanding of a relatively complex internal structure. This approach underlies the most successful game built around simulated human characters: The Sims. It also provides a way of understanding the development of systems for believable animated characters and interactive drama, which eschew the most direct representations of internal system state. Finally, the SimCity effect provides an important design guideline for those who seek to make another powerful technology for fiction — language — playable in new ways.
At the same time, in parallel with these waypoints, another set of arguments unfolds. The focus of this strand of Expressive Processing, as this chapter has outlined, is interpreting software. As I have discussed, our lives are lived surrounded by software — from everyday Google searches and payroll systems to the high stakes of Diebold voting machines and government surveillance. We need to be prepared to engage software critically, accustomed to interpreting descriptions of processes, able to understand common pitfalls, and aware of what observing software’s output reveals and conceals about its inner workings. To this end, this book uses the same systems for fiction and games mentioned above to provide specific grounding for ideas about software that are often described only in the abstract. For example, while it is a truism that software can initially appear to act with more intelligence than is present in its internal models, it is more vivid and helpful to see the specifics of how Eliza plays on our expectations to create its illusion. Similarly, while we might say, abstractly, that problems of “common-sense reasoning” limit the possibilities for artificial intelligence systems meant to emulate human thought, it is more memorable to examine the story generator Minstrel as it portrays a hungry knight eating a princess. At the same time, systems such as SimCity not only create rich experiences of play but also demonstrate another route — in addition to procedural literacy and legible examples — to building the necessary broad public understanding of software.
Overall, both strands of Expressive Processing grow from the same fundamental move: looking at digital media as software. It is my hope that — by examining games and fictions not just for their appearances, but also for their operations — this book provides new insights into how software can be used for authorial expression, as well as what is expressed by the shape of software processes themselves.