# Co-Writing Screenplays and Theatre Scripts with Language Models: An Evaluation by Industry Professionals

**Authors:** Piotr Mirowski, Kory W. Mathewson, Jaylen Pittman, Richard Evans  
**arXiv:** [2209.14958](https://arxiv.org/abs/2209.14958)  
**Date:** 2022-09-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2209.14958v1/

## Key points

- Dramatron is an interactive co-writing tool that uses language models (LLMs) to generate scripts and screenplays
- LLMs lack long-range semantic coherence, so the researchers applied them hierarchically in Dramatron
- Prompt chaining and building structural context were used to generate complete scripts with titles, characters, story beats, location descriptions, and dialogue
- A user study involving 15 theatre and film industry professionals was conducted to evaluate the effectiveness of Dramatron as a co-creative system
- Participants co-wrote scripts with Dramatron and provided critical reflections through open-ended interviews
- Independent reviewers watched stagings of the works to gather feedback
- Results show that Dramatron can be a valuable tool for human-machine co-creativity in writing
- The study raises ethical considerations regarding the use of LLMs for creative purposes such as plagiarism and bias
- This research expands understanding of co-creativity and highlights the potential benefits and challenges of using hierarchical text generation tools like Dramatron in collaborative writing processes.

## Summary

The study introduces Dramatron, an interactive co-writing tool that utilizes language models (LLMs) to generate coherent scripts and screenplays. To address the limitation of LLMs lacking long-range semantic coherence, the researchers applied them hierarchically in Dramatron. This enabled the generation of complete scripts with titles, characters, story beats, location descriptions, and dialogue by incorporating prompt chaining and building structural context. To evaluate its effectiveness as a co-creative system, a user study involving 15 theatre and film industry professionals was conducted. Participants co-wrote theatre scripts and screenplays with Dramatron and provided critical reflections through open-ended interviews. Additionally, independent reviewers watched stagings of the works to gather feedback. The results demonstrate that Dramatron can be a valuable tool for human-machine co-creativity in writing. The study also raises important considerations regarding the ethics of using LLMs for creative purposes such as plagiarism and bias. Overall, this research expands understanding of co-creativity and highlights the potential benefits and challenges associated with employing hierarchical text generation tools like Dramatron in collaborative writing processes.