The relationship between artificial intelligence and the arts has traditionally been one of inspiration. Particularly in areas such as computational creativity, many AI systems have been inspired by particular genres or even individual works. This dissertation explores an alternate relationship, in which AI can be both inspired by and an inspiration to an artistic theory.

Using answer set programming, Dunyazad is a program that generates narrative choices like those found in Choose-Your-Own-Adventure books. Dunyazad depends on a theory of "choice poetics" which describes how to analyze narrative choices in terms of the goals that players have when making decisions. However, this theory is also informed by results from experiments that used choices created by Dunyazad. In other words, the development of a theory of choice poetics is informed by the technical work of building an AI system which in turn relies on that theory.

This dissertation describes both a novel technical system that intentionally constructs diverse choice structures and a theoretical approach to the analysis of choice poetics. It also presents the results from two online surveys that both validate Dunyazad's performance as a system and inform the underlying theory. Important caveats are discussed, such as the importance of relative as opposed to absolute option analysis and the effect of outcomes on the perception of options that lead to them. These experimental results inform the development of both the system and the underlying theory, and describing the methodology which makes this possible is the final contribution of this document.