Ryan Heuser
This paper examines the formal and aesthetic patterns of AI-generated poems through a series of computational experiments. Through analyses of rhyme and rhythm, it reveals how large language models (LLMs) exhibit a stubborn, formal stuckness in their outputs. The paper demonstrates that LLMs often ossify poetic forms by producing formally conservative texts that adhere more rigidly to traditional poetic conventions than even the most formally strict periods of literary history. The paper interprets these findings as evidence of a computational logic of idealization that privileges the satisfaction of formal expectation over its artful frustration, regularity over variation, and conformity over contradiction. It proposes “generative formalism” as a critical framework that extends traditional as well as quantitative and formalist methods to understand how generative systems process, flatten, and reify cultural production.