Intelligent environments combine the promise of ubiquitous computing with artificial intelligence and are increasingly being used in public art. The agent-based approach to artificial intelligence (AI) uses the intelligence function to characterize agent-based behavior. The inputs to the intelligence function, perception of the environment and the agent's internal state, combined with the outputs of the function, actuation and changes in internal state, provides a lens with which to categorized AI-based public art. Such works can be classified as generative, reactive, interactive, learning, or static. To illustrate this taxonomy, this paper gives examples of public artworks that fit into each of the five categories and uses the taxonomy to suggest new areas of creative inquiry.
