From Feedback to Funnier Scripts: Evaluating AI Stand-Up Comedy Performance Through Prompt Refinement in a Two-Phase Experiment

Abstract

Humor generation has become an emerging topic in the development of creative artificial intelligence (AI). This study examines whether AI-generated stand-up comedy can be improved through prompt refinement based on audience feedback. Drawing on Computational Humor Theory and Prompt Engineering, a two-phase experiment was conducted to evaluate and optimize AI-generated comedy scripts. In the first phase, participants assessed an AI-generated stand-up performance and provided qualitative feedback. Based on these responses, prompts were refined accordingly.The second evaluation showed overall improvements in audience perception, but some limitations remained. The fi ndings highlight the potential of human–AI feedback loops in enhancing creative AI content.