Artificial intelligence methods open up new possibilities in art and entertainment, enabling rich and deeply interactive experiences. At the same time as AI opens up new fields of artistic expression, AI-based art itself becomes a fundamental research agenda, posing and answering novel research questions that would not be raised unless doing AI research in the context of art and entertainment. I call this agenda, in which AI research and art mutually inform each other, Expressive AI. Expressive AI takes seriously the problem of building intelligences that robustly function outside of the lab, engaging human participants in intellectually and aesthetically satisfying interactions, which, hopefully, teach us something about ourselves. This thesis describes a specific AI-based art piece, an interactive drama called Fa�ade, and describes the practice of Expressive AI, using Fa�ade, as well as additional AI-based artwork described in the appendices, as case studies. An interactive drama is a dramatically interesting virtual world inhabited by computer-controlled characters, within which the player experiences a story from a first person perspective. Over the past decade, there has been a fair amount of research into believable agents, that is, autonomous characters exhibiting rich personalities, emotions, and social interactions. There has been comparatively little work, however, exploring how the local, reactive behavior of believable agents can be integrated with the more global, deliberative nature of a story plot, so as to build interactive, dramatic worlds. This thesis presents Fa�ade, the first published interactive drama system that integrates character (believable agents), story (drama management) and shallow natural language processing into a complete system. Fa�ade will be publicly released as a free download in 2003. In the Fa�ade architecture, the unit of plot/character integration is the dramatic beat. In the theory of dramatic writing, beats are the smallest unit of dramatic action, consisting of a short dialog exchange or small amount of physical action. As architectural entities, beats organize both the procedural knowledge to accomplish the beat s dramatic action, and the declarative knowledge to sequence the beat in an evolving plot. Instead of conceiving of the characters as strongly autonomous entities that coordinate to accomplish dramatic action through purely local decision-making, characters are instead weakly autonomous the character s behavioral repertoire dynamically changes as beats are sequenced. The Fa�ade architecture includes ABL (A Behavior Language), a new reactive planning language for authoring characters that provides language support for joint action, and a drama manager consisting of both a language for authoring the declarative knowledge associated with beats and a runtime system that dynamically sequences beats. Fa�ade is a collaboration with independent artist and researcher Andrew Stern. Expressive AI is not the mere application of off-the-shelf AI techniques to art and entertainment applications. Rather, Expressive AI is a critical technical practice, a way of doing AI research that reflects on the foundations of AI and changes the way AI is done. AI has always been in the business of knowing-by-making, exploring what it means to be human by building systems. Expressive AI just makes this explicit, combining the thought experiments of the AI researcher with the conceptual and aesthetic experiments of the artist. As demonstrated through Fa�ade and the other systems/artworks described in the appendices, combining art and AI, both ways of knowing-by-making, opens up new research questions, provides a novel perspective on old questions, and enables new modes of artistic expression. The firm boundary normally separating art and science is blurred, becoming two components of a single, integrated practice.

284 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by [email protected]