From Marcel Duchamp's notion of The Creative Act, to Experiments in Art and Technology's Pepsi Pavilion, to Myron Krueger's assertion "Response is the Medium", the paradigm of spectator participation in interactive art presents a challenging problem of metacreation to artists. With emerging perceptual user interfaces and participatory sensing technologies providing multimodal data streams from an audience, the state-of-the-art creative strategies lack a meta-level organization of audience responses. Synthesis of new interactive experiences and narratives using audience sensing interfaces requires an abstract audience representation. The absence of such abstract audience representation could be characterized as a lack of models concerning reciprocal actions between artworks and audiences in interactive art.

This dissertation contextualizes the problem of metacreation through an original survey of the pioneering works using crowd-sourced sensing in interactive art practice, electronic music composition, human computer interaction, and ubiquitous computing. It addresses the significance of abstract audience representation in creative process through systems thinking, and provides a novel Symbolic Audience Model as the engine of the Audience Response Synthesis (ARS). Demonstrated by software implementations and interactive artworks, ARS considers humans as samplers and provides a meta-level technique for synthesis using the audience data, with an aim to fully exploit the creative potential provided by the emerging ecosystems of mobile and wearable sensing. The result is an analysis of the author's practice that yields an unique creative strategy: Collective Expression. Collective Expression manifests data with its emergent properties as a new language for crowd-based interactive artworks. This dissertation concludes with the applicability of ARS in synthesis of new interactive experiences using audience responses, and outlines the requirements of a model for multi-scale emergent behaviors to guide further investigation in expressiveness using ARS.